Publish NLLB SPM8k private no-Bible v3 best checkpoint (20260809-210523)

README.md

@@ -4,7 +4,6 @@ library_name: transformers
 pipeline_tag: translation
 base_model: facebook/nllb-200-distilled-600M
 language:
-- zh
 - ami
 - bnn
 - ckv
@@ -20,232 +19,163 @@ language:
 - tsu
 - xnb
 - xsy
+- zh
 tags:
 - translation
-- nllb
 - nllb-200
-- low-resource
-- endangered-languages
 - formosan-languages
-- sentencepiece
-- private-no-bible
+- low-resource
 metrics:
 - bleu
-- chrf2
+- chrf
 - ter
 model-index:
 - name: nllb200-formosan-zh-spm8k
   results:
   - task:
-      name: Machine Translation
       type: translation
+      name: Translation
     dataset:
-      name: FormosanBank Traditional Chinese private no-Bible hard split
-      type: custom
+      name: FormosanBank private no-Bible hard test
+      type: private-no-bible-hard-test
+      split: test
     metrics:
-    - name: BLEU
-      type: bleu
-      value: 10.5767
-      args:
-        direction: f2zh
-        samples: 61384
-        tokenize: zh
-    - name: chrF2
-      type: chrf2
-      value: 12.3912
-      args:
-        direction: f2zh
-        samples: 61384
-    - name: TER
-      type: ter
-      value: 115.7521
-      args:
-        direction: f2zh
-        samples: 61384
+    - type: bleu
+      name: sacreBLEU
+      value: 12.580739
+    - type: chrf
+      name: chrF2
+      value: 12.964706
+    - type: ter
+      name: TER
+      value: 82.854846
 ---
 
 # nllb200-formosan-zh-spm8k
 
-**Base model:** [`facebook/nllb-200-distilled-600M`](https://huggingface.co/facebook/nllb-200-distilled-600M)  
-**Direction:** **Formosan -> Traditional Chinese**  
-**Companion model:** [`FormosanBank/nllb200-zh-formosan-spm8k`](https://huggingface.co/FormosanBank/nllb200-zh-formosan-spm8k)  
-**Release:** private no-Bible SPM8k flight `20260712-232900`, validation-selected step `270,000`
+**Direction:** Formosan to Traditional Chinese<br>
+**Base model:** [`facebook/nllb-200-distilled-600M`](https://huggingface.co/facebook/nllb-200-distilled-600M)<br>
+**Recipe:** `nllb200-spm8k-directional-v3`<br>
+**Release:** `20260809-210523`, validation-selected step 220,000
 
-This directional checkpoint replaces the earlier release with the strongest `private_no_bible` model from the
-fully rebuilt FormosanBank MT pipeline. It uses an 8,192-piece Formosan-aware SentencePiece extension and explicit
-direction, source-language, source-domain, and dialect control tags.
+This is a directional model for 15 Formosan languages. It uses the
+`private_no_bible` leakage-controlled corpus, Formosan-aware 8k SentencePiece extension, balanced
+language/source sampling, and direction/domain/dialect control tags. The model
+weights are public, but the private training corpus is not included.
 
-## Supported Languages
+## Model details
 
-| Language | NLLB code |
+| Item | Value |
 |---|---|
-| Traditional Chinese | `zho_Hant` |
-| Amis | `ami_Latn` |
-| Bunun | `bnn_Latn` |
-| Kavalan | `ckv_Latn` |
-| Rukai | `dru_Latn` |
-| Paiwan | `pwn_Latn` |
-| Puyuma | `pyu_Latn` |
-| Thao | `ssf_Latn` |
-| Saaroa | `sxr_Latn` |
-| Sakizaya | `szy_Latn` |
-| Tao / Yami | `tao_Latn` |
-| Atayal | `tay_Latn` |
-| Seediq | `trv_Latn` |
-| Tsou | `tsu_Latn` |
-| Kanakanavu | `xnb_Latn` |
-| Saisiyat | `xsy_Latn` |
-
-## Input Format
-
-Prefix every source with:
-
-`<to_zh> <src_LANG> <dom_BUCKET> <dialect_DIALECT>`
-
-Example:
-
-`<to_zh> <src_ami> <dom_unknown> <dialect_default> Pa'araw cingra to demak nira.`
-
-Use `<dom_unknown>` and `<dialect_default>` when metadata is unavailable.
+| Base revision | `f8d333a098d19b4fd9a8b18f94170487ad3f821d` |
+| Training rows | 702,520 |
+| Effective batch size | 64 |
+| Maximum sequence length | 384 |
+| Learning rate | 2e-05 |
+| Precision | `bf16` |
+| Checkpoint selection | Human validation `chrF2` |
+| Formosan text | `kindOf=standard`, `formosan-mt-standard-v3` |
+| Corpus SHA-256 | `238b96ea87716a749391d7157ce99f1611c1d4a489c4c108c315553505ca1f29` |
+| Training profile SHA-256 | `34f45832bdedc1b8322b26936dd88667dfecb360d941a7f54f60ae31a1fb4e10` |
 
 ## Usage
 
-Use the slow `NllbTokenizer` (`use_fast=False` with `AutoTokenizer`). These checkpoints were trained with
-`transformers==4.56.1`; fast-tokenizer added-token IDs can differ from the slow tokenizer IDs used in training.
-NLLB generation must start with the tokenizer EOS ID and force the target-language BOS ID.
-
 ```python
 import torch
 from transformers import AutoModelForSeq2SeqLM, NllbTokenizer
+from formosan_mt_inference import normalize_formosan
+
 
 model_id = "FormosanBank/nllb200-formosan-zh-spm8k"
-tokenizer = NllbTokenizer.from_pretrained(model_id)
+tokenizer = NllbTokenizer.from_pretrained(model_id, use_fast=False)
 model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
 model.to("cuda" if torch.cuda.is_available() else "cpu")
-
-FORMOSAN_TO_LID = {
-    "ami": "ami_Latn", "bnn": "bnn_Latn", "ckv": "ckv_Latn", "dru": "dru_Latn",
-    "pwn": "pwn_Latn", "pyu": "pyu_Latn", "ssf": "ssf_Latn", "sxr": "sxr_Latn",
-    "szy": "szy_Latn", "tao": "tao_Latn", "tay": "tay_Latn", "trv": "trv_Latn",
-    "tsu": "tsu_Latn", "xnb": "xnb_Latn", "xsy": "xsy_Latn",
-}
-
-def translate_formosan_to_chinese(text: str, lang_code: str, source_bucket: str = "unknown", dialect: str = "default") -> str:
-    tokenizer.src_lang = FORMOSAN_TO_LID[lang_code]
-    prompt = f"<to_zh> <src_{lang_code}> <dom_{source_bucket}> <dialect_{dialect}> {text}"
-    inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=384).to(model.device)
-    outputs = model.generate(
+NLLB_LIDS = {'ami': 'ami_Latn', 'bnn': 'bnn_Latn', 'ckv': 'ckv_Latn', 'dru': 'dru_Latn', 'pwn': 'pwn_Latn', 'pyu': 'pyu_Latn', 'ssf': 'ssf_Latn', 'sxr': 'sxr_Latn', 'szy': 'szy_Latn', 'tao': 'tao_Latn', 'tay': 'tay_Latn', 'trv': 'trv_Latn', 'tsu': 'tsu_Latn', 'xnb': 'xnb_Latn', 'xsy': 'xsy_Latn'}
+
+def translate(text, lang_code, source_bucket="unknown", dialect="default"):
+    text = normalize_formosan(text, lang_code)
+    tokenizer.src_lang = NLLB_LIDS[lang_code]
+    prompt = (
+        f"<to_zh> <src_{lang_code}> <dom_{source_bucket}> "
+        f"<dialect_{dialect}> {text}"
+    )
+    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+    output = model.generate(
         **inputs,
-        forced_bos_token_id=tokenizer.convert_tokens_to_ids("zho_Hant"),
         decoder_start_token_id=tokenizer.eos_token_id,
-        max_new_tokens=128,
+        forced_bos_token_id=tokenizer.convert_tokens_to_ids('zho_Hant'),
+        max_new_tokens=256,
         num_beams=4,
-        no_repeat_ngram_size=3,
-        repetition_penalty=1.15,
-        early_stopping=True,
     )
-    return tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
+    return tokenizer.batch_decode(output, skip_special_tokens=True)[0]
 
-print(translate_formosan_to_chinese("Pa'araw cingra to demak nira.", "ami"))
+print(translate("Pa'araw cingra.", "ami"))
 ```
 
-## Checkpoint Selection
-
-The published checkpoint was selected **only on validation chrF2**, not on the hard test set.
-
-| Selection step | Validation samples | Validation loss | Perplexity | BLEU | chrF2 | TER |
-|---:|---:|---:|---:|---:|---:|---:|
-| 270,000 | 1,920 | 2.4207 | 11.25 | 23.58 | 22.94 | 103.94 |
-
-Validation generation sampled 128 rows per Formosan language every 10,000 updates. The full hard test was evaluated
-only after selection.
-
-## Training Setup
-
-| Setting | Value |
-|---|---|
-| Corpus | `private_no_bible` (Traditional Chinese) |
-| Base model | `facebook/nllb-200-distilled-600M` |
-| Maximum updates | 300,000 |
-| Published best step | 270,000 |
-| Microbatch / accumulation | 16 / 4 |
-| Effective batch | 64 |
-| Maximum length | 384 |
-| Learning rate | `2e-05` |
-| Precision | `bf16` |
-| Easy-source weight | 0.05 |
-| Language sampling alpha | 0.5 |
-| Metadata control tags | enabled and validated as single tokenizer IDs |
-
-## Corpus and Split Integrity
-
-| Total | Train | Test | Validate | Minimum per-language test | Minimum per-language validate |
-|---:|---:|---:|---:|---:|---:|
-| 791,330 | 708,496 | 61,384 | 21,450 | 7.5% | 2.5% |
-
-The exact `Formosan-Taiwan-Bible-Society-Bibles` repository is excluded. Lexical entries are train-only. Independent
-validation found zero normalized source, target, or pair overlap; zero punctuation/spacing skeleton overlap; and zero
-one-edit source or target conflicts across train and evaluation. Connected similarity groups are assigned as units so
-variants cannot be split independently merely because they are not exact duplicates.
-
-All hard-test references in this Chinese corpus are original rather than pivot-generated.
-
-## Hard-Test Results
-
-SacreBLEU was computed with `zh` tokenization; chrF uses beta 2; TER is lower-is-better.
-
-| Direction | Samples | BLEU | chrF2 | TER | Exact match | Empty output |
-|---|---:|---:|---:|---:|---:|---:|
-| Formosan -> Traditional Chinese | 61,384 | 10.58 | 12.39 | 115.75 | 1.40% | 0.00% |
-
-### Per-Language Results
-
-| Language | Code | Samples | BLEU | chrF2 | TER |
-|---|---:|---:|---:|---:|---:|
-| Amis | `ami_Latn` | 10,929 | 7.91 | 10.07 | 115.26 |
-| Bunun | `bnn_Latn` | 5,450 | 9.24 | 10.91 | 120.58 |
-| Kavalan | `ckv_Latn` | 2,852 | 15.76 | 16.77 | 108.05 |
-| Rukai | `dru_Latn` | 5,524 | 8.14 | 10.05 | 127.33 |
-| Paiwan | `pwn_Latn` | 5,519 | 8.81 | 10.99 | 111.15 |
-| Puyuma | `pyu_Latn` | 4,105 | 13.10 | 15.12 | 107.76 |
-| Thao | `ssf_Latn` | 1,670 | 14.87 | 16.29 | 110.56 |
-| Saaroa | `sxr_Latn` | 1,539 | 8.57 | 11.04 | 109.41 |
-| Sakizaya | `szy_Latn` | 2,147 | 12.58 | 15.79 | 110.10 |
-| Tao / Yami | `tao_Latn` | 1,802 | 8.32 | 11.30 | 109.37 |
-| Atayal | `tay_Latn` | 6,395 | 8.92 | 10.24 | 125.92 |
-| Seediq | `trv_Latn` | 6,689 | 17.66 | 18.01 | 107.97 |
-| Tsou | `tsu_Latn` | 1,889 | 8.00 | 10.40 | 115.47 |
-| Kanakanavu | `xnb_Latn` | 2,927 | 17.48 | 18.51 | 111.07 |
-| Saisiyat | `xsy_Latn` | 1,947 | 9.94 | 12.25 | 125.58 |
-
-Full source-bucket and length-bin breakdowns are in [`eval/metrics.json`](eval/metrics.json).
-
-## Intended Use
-
-- Research, teaching, and prototyping for Formosan-language machine translation.
-- Draft translation assistance where knowledgeable speakers can review the output.
-- Comparative low-resource MT evaluation on the documented leakage-controlled split.
+The control tags are part of the training contract. Use `unknown` and `default`
+when source bucket or dialect metadata is unavailable.
+
+## Evaluation
+
+The best checkpoint was selected on human validation chrF2. Test references
+are human sentence pairs; synthetic pivots and lexical entries are train-only.
+The headline result uses `default` metadata controls, so it does not
+assume access to test-set domain or dialect labels.
+
+| Split | Rows |
+|---|---:|
+| Train | 702,520 |
+| Test | 64,356 |
+| Validate | 21,450 |
+
+| Scope | BLEU | chrF2 | TER |
+|---|---:|---:|---:|
+| Hard test | 12.58 | 12.96 | 82.85 |
+| Selection validation | 9.41 | 11.26 | 90.38 |
+
+Test empty-output rate: 0.0140%.
+
+
+| Language | Samples | BLEU | chrF2 | TER |
+|---|---:|---:|---:|---:|
+| `ami` | 11,271 | 13.19 | 13.38 | 80.51 |
+| `bnn` | 6,112 | 11.10 | 11.85 | 88.68 |
+| `ckv` | 3,038 | 15.17 | 14.81 | 76.72 |
+| `dru` | 6,109 | 5.94 | 7.04 | 102.81 |
+| `pwn` | 5,718 | 8.52 | 9.67 | 93.46 |
+| `pyu` | 4,280 | 19.80 | 18.77 | 73.94 |
+| `ssf` | 1,673 | 15.88 | 17.04 | 70.99 |
+| `sxr` | 1,781 | 13.11 | 13.68 | 76.03 |
+| `szy` | 2,055 | 12.49 | 14.56 | 75.39 |
+| `tao` | 1,838 | 10.60 | 12.10 | 86.38 |
+| `tay` | 6,723 | 10.03 | 10.98 | 78.01 |
+| `trv` | 6,983 | 14.43 | 15.24 | 80.84 |
+| `tsu` | 2,164 | 11.90 | 12.90 | 80.67 |
+| `xnb` | 2,728 | 19.40 | 18.49 | 69.28 |
+| `xsy` | 1,883 | 11.52 | 12.48 | 77.75 |
+
+The corpus gate enforces standard-tier Formosan text, at least 7.5% test and
+2.5% validation per language, human sentence-only evaluation, and zero exact,
+skeleton, one-edit, configured high character n-gram, or document
+train/evaluation conflicts. This release passed all gates: exact
+0, skeleton
+0, one-edit
+0, character n-gram
+0, and document
+0.
+
+See `eval/metrics.json` for sacreBLEU signatures, per-language, source,
+dialect, and length diagnostics. `publication.json` records the corpus,
+profile, run, and checkpoint hashes used for this release.
+
+## Intended use
+
+This model supports research, corpus development, and assisted translation for
+the 15 included Formosan languages. It is designed for the exact prompt and
+generation contract shown above.
 
 ## Limitations
 
-- Output may be incorrect, ungrammatical, incomplete, or culturally inappropriate.
-- Formosan generation is draft-only and requires speaker review.
-- Aggregate scores across 15 languages conceal substantial per-language variation.
-- This model is unsuitable for legal, medical, safety-critical, or authoritative community-facing use without expert review.
-- Hard-split scores are not directly comparable with earlier evaluations that allowed stronger train-test similarity.
-
-## License
-
-Released under `cc-by-nc-4.0`. Underlying corpus sources may impose additional restrictions. Confirm the rights needed
-for your use case.
-
-## Citation
-
-```bibtex
-@misc{formosanbank_nllb200_formosan_zh_spm8k_2026,
-  title  = {nllb200-formosan-zh-spm8k: Directional NLLB-200 MT on the FormosanBank private no-Bible corpus},
-  author = {FormosanBank contributors},
-  year   = {2026},
-  url    = {https://huggingface.co/FormosanBank/nllb200-formosan-zh-spm8k}
-}
-```
+Outputs require knowledgeable speaker review. Aggregate metrics hide large
+differences among languages and domains. This model is not suitable for
+authoritative, medical, legal, or safety-critical translation.

added_tokens.json

@@ -1,182 +1,309 @@
 {
-  "<dae>": 259639,
-  "<dialect_central>": 259640,
-  "<dialect_coastal>": 259641,
-  "<dialect_dawu>": 259642,
-  "<dialect_default>": 259643,
-  "<dialect_deluvalley>": 259644,
-  "<dialect_dona>": 259645,
-  "<dialect_duda>": 259646,
-  "<dialect_eastern>": 259647,
-  "<dialect_fourseasons>": 259648,
-  "<dialect_hengchun>": 259649,
-  "<dialect_jianhe>": 259650,
-  "<dialect_junqun>": 259651,
-  "<dialect_kanakanavu>": 259652,
-  "<dialect_kaqun>": 259653,
-  "<dialect_kavalan>": 259654,
-  "<dialect_luanqun>": 259655,
-  "<dialect_malan>": 259656,
-  "<dialect_maolin>": 259657,
-  "<dialect_nanwang>": 259658,
-  "<dialect_northern>": 259659,
-  "<dialect_saaroa>": 259660,
-  "<dialect_saisiyat>": 259661,
-  "<dialect_sakizaya>": 259662,
-  "<dialect_sekolik>": 259663,
-  "<dialect_southern>": 259664,
-  "<dialect_tanqun>": 259665,
-  "<dialect_tegudaya>": 259666,
-  "<dialect_thao>": 259667,
-  "<dialect_truku>": 259668,
-  "<dialect_tsou>": 259669,
-  "<dialect_unknown>": 259670,
-  "<dialect_wanda>": 259671,
-  "<dialect_wanshan>": 259672,
-  "<dialect_wenshui>": 259673,
-  "<dialect_wutai>": 259674,
-  "<dialect_xiqun>": 259675,
-  "<dialect_xiuguluan>": 259676,
-  "<dialect_yami>": 259677,
-  "<dialect_yilanzeaol>": 259678,
-  "<dialect_zeaol>": 259679,
-  "<dialect_zhiben>": 259680,
-  "<dialect_zhuoqun>": 259681,
-  "<dom_classroom_context>": 259682,
-  "<dom_culture>": 259683,
-  "<dom_dictionary>": 259684,
-  "<dom_essays>": 259685,
-  "<dom_ethnography_journal_ritual_song>": 259686,
-  "<dom_formosan_100_paiwan_texts>": 259687,
-  "<dom_formosan_academia_sinica_oral_legends>": 259688,
-  "<dom_formosan_amis_adversative_constructions>": 259689,
-  "<dom_formosan_amis_kavalan_lin_interrogative_verbs>": 259690,
-  "<dom_formosan_amis_kuo_sung_comparative_constructions>": 259691,
-  "<dom_formosan_amis_lifok_dongi_saopo_kimakimad>": 259692,
-  "<dom_formosan_amis_myths_and_customs>": 259693,
-  "<dom_formosan_amis_pa_verbs>": 259694,
-  "<dom_formosan_amis_serial_verb_constructions>": 259695,
-  "<dom_formosan_amis_tung_chiou>": 259696,
-  "<dom_formosan_asai_sedik_language>": 259697,
-  "<dom_formosan_bunun_debusser_dissertation>": 259698,
-  "<dom_formosan_bunun_moriguchi_northern_texts>": 259699,
-  "<dom_formosan_bunun_topic_focus>": 259700,
-  "<dom_formosan_chen_fukuda_relabeling_ergative>": 259701,
-  "<dom_formosan_chen_fukuda_three_ways_steal>": 259702,
-  "<dom_formosan_cip_atayal_grammar_overview>": 259703,
-  "<dom_formosan_dean_johnson_year_clouds_smangus>": 259704,
-  "<dom_formosan_dorinda_tsai_hsiu_liu_neutral_imperfect>": 259705,
-  "<dom_formosan_egerod_origin_headhunting_atayal_text_v>": 259706,
-  "<dom_formosan_egerod_soren_word_order_word_classes_at>": 259707,
-  "<dom_formosan_elizabeth_zeitoun_chen_huei_wu_overview>": 259708,
-  "<dom_formosan_epark>": 259709,
-  "<dom_formosan_ferrell_taiwan_aboriginal_groups_proble>": 259710,
-  "<dom_formosan_gitbook_translations>": 259711,
-  "<dom_formosan_glosbe>": 259712,
-  "<dom_formosan_hala_saku_la_videos>": 259713,
-  "<dom_formosan_holmer_seediq>": 259714,
-  "<dom_formosan_huang_grammaticalization_squliq_atayal>": 259715,
-  "<dom_formosan_huang_mei_chin_atayal_reference_grammar>": 259716,
-  "<dom_formosan_huteson_rukai_survey>": 259717,
-  "<dom_formosan_ilrdf_42_language_practice_word_lists>": 259718,
-  "<dom_formosan_ilrdf_tousvusvutu_kita>": 259719,
-  "<dom_formosan_indigenous_language_literary_awards>": 259720,
-  "<dom_formosan_kanakanavu_texts>": 259721,
-  "<dom_formosan_kavalan_zhang_reference_grammar>": 259722,
-  "<dom_formosan_li_peirong_xu_weisheng_introduction_tru>": 259723,
-  "<dom_formosan_li_peizhen_wu_xingyu_shouhu_xing>": 259724,
-  "<dom_formosan_naomi_tsukida_correlative_clauses_seedi>": 259725,
-  "<dom_formosan_nowbucyang_truku_thesis>": 259726,
-  "<dom_formosan_old_texts>": 259727,
-  "<dom_formosan_paiwan_collart_zeitoun_time_reference>": 259728,
-  "<dom_formosan_paiwan_ho_five_dialects>": 259729,
-  "<dom_formosan_paiwanstories>": 259730,
-  "<dom_formosan_puyuma_chen_raising_to_object>": 259731,
-  "<dom_formosan_puyuma_katipol_kopfjagdriten>": 259732,
-  "<dom_formosan_puyuma_teng_pronominal_systems>": 259733,
-  "<dom_formosan_puyuma_teng_reference_grammar>": 259734,
-  "<dom_formosan_rik_bunun>": 259735,
-  "<dom_formosan_robert_blust_austronesian_homeland_ling>": 259736,
-  "<dom_formosan_rukai_mantauran_stories>": 259737,
-  "<dom_formosan_rukai_texts>": 259738,
-  "<dom_formosan_rukai_zeitoun_saying>": 259739,
-  "<dom_formosan_saaroa_pan_grammar>": 259740,
-  "<dom_formosan_sakizaya_affixes>": 259741,
-  "<dom_formosan_seals>": 259742,
-  "<dom_formosan_seediq_zhang_reference_grammar>": 259743,
-  "<dom_formosan_shigeru_tsuchida_kanakanavu_texts>": 259744,
-  "<dom_formosan_shigeru_tsuchida_reconstruction_proto_t>": 259745,
-  "<dom_formosan_shih_rukai_adverbial>": 259746,
-  "<dom_formosan_song_limei_introduction_kanakanavu_gram>": 259747,
-  "<dom_formosan_song_limei_introduction_seediq_grammar>": 259748,
-  "<dom_formosan_sung_kuo_descriptive_comparative_constr>": 259749,
-  "<dom_formosan_taoshan_elementary_school_atelier_hui_k>": 259750,
-  "<dom_formosan_tsukida_naomi_verb_classification_amis>": 259751,
-  "<dom_formosan_wei_matri_clan_lineage_system_ami>": 259752,
-  "<dom_formosan_wilang_yutas_videos>": 259753,
-  "<dom_formosan_wu_jinglan_introduction_amis_grammar>": 259754,
-  "<dom_formosan_yedda_palemeq_blog>": 259755,
-  "<dom_formosan_yeddas_blog>": 259756,
-  "<dom_formosan_yeh_meili_introduction_saisiyat_grammar>": 259757,
-  "<dom_formosan_zhang_yongli_pan_jiarong_introduction_t>": 259758,
-  "<dom_formosan_zheng_acl_2024>": 259759,
-  "<dom_formosan_zheng_data>": 259760,
-  "<dom_learning_vocab>": 259761,
-  "<dom_nine_level>": 259762,
-  "<dom_ntu>": 259763,
-  "<dom_picture_book>": 259764,
-  "<dom_picture_story>": 259765,
-  "<dom_presidential_apology>": 259766,
-  "<dom_reading_writing>": 259767,
-  "<dom_unknown>": 259768,
-  "<dom_youtube>": 259769,
-  "<src_ami>": 259770,
-  "<src_bnn>": 259771,
-  "<src_ckv>": 259772,
-  "<src_dru>": 259773,
-  "<src_eng>": 259774,
-  "<src_pwn>": 259775,
-  "<src_pyu>": 259776,
-  "<src_ssf>": 259777,
-  "<src_sxr>": 259778,
-  "<src_szy>": 259779,
-  "<src_tao>": 259780,
-  "<src_tay>": 259781,
-  "<src_trv>": 259782,
-  "<src_tsu>": 259783,
-  "<src_xnb>": 259784,
-  "<src_xsy>": 259785,
-  "<src_zh>": 259786,
-  "<to_ami>": 259787,
-  "<to_bnn>": 259788,
-  "<to_ckv>": 259789,
-  "<to_dru>": 259790,
-  "<to_eng>": 259791,
-  "<to_pwn>": 259792,
-  "<to_pyu>": 259793,
-  "<to_ssf>": 259794,
-  "<to_sxr>": 259795,
-  "<to_szy>": 259796,
-  "<to_tao>": 259797,
-  "<to_tay>": 259798,
-  "<to_trv>": 259799,
-  "<to_tsu>": 259800,
-  "<to_xnb>": 259801,
-  "<to_xsy>": 259802,
-  "<to_zh>": 259803,
-  "ami_Latn": 259624,
-  "bnn_Latn": 259625,
-  "ckv_Latn": 259626,
-  "dru_Latn": 259627,
-  "pwn_Latn": 259628,
-  "pyu_Latn": 259629,
-  "ssf_Latn": 259630,
-  "sxr_Latn": 259631,
-  "szy_Latn": 259632,
-  "tao_Latn": 259633,
-  "tay_Latn": 259634,
-  "trv_Latn": 259635,
-  "tsu_Latn": 259636,
-  "xnb_Latn": 259637,
-  "xsy_Latn": 259638
+  "<dialect_central>": 261372,
+  "<dialect_coastal>": 261373,
+  "<dialect_dawu>": 261374,
+  "<dialect_default>": 261375,
+  "<dialect_deluvalley>": 261376,
+  "<dialect_dona>": 261377,
+  "<dialect_duda>": 261378,
+  "<dialect_eastern>": 261379,
+  "<dialect_fourseasons>": 261380,
+  "<dialect_hengchun>": 261381,
+  "<dialect_jianhe>": 261382,
+  "<dialect_junqun>": 261383,
+  "<dialect_kanakanavu>": 261384,
+  "<dialect_kaqun>": 261385,
+  "<dialect_kavalan>": 261386,
+  "<dialect_luanqun>": 261387,
+  "<dialect_malan>": 261388,
+  "<dialect_maolin>": 261389,
+  "<dialect_nanwang>": 261390,
+  "<dialect_northern>": 261391,
+  "<dialect_saaroa>": 261392,
+  "<dialect_saisiyat>": 261393,
+  "<dialect_sakizaya>": 261394,
+  "<dialect_sekolik>": 261395,
+  "<dialect_southern>": 261396,
+  "<dialect_tanqun>": 261397,
+  "<dialect_tegudaya>": 261398,
+  "<dialect_thao>": 261399,
+  "<dialect_truku>": 261400,
+  "<dialect_tsou>": 261401,
+  "<dialect_unknown>": 261402,
+  "<dialect_wanda>": 261403,
+  "<dialect_wanshan>": 261404,
+  "<dialect_wenshui>": 261405,
+  "<dialect_wutai>": 261406,
+  "<dialect_xiqun>": 261407,
+  "<dialect_xiuguluan>": 261408,
+  "<dialect_yami>": 261409,
+  "<dialect_yilanzeaol>": 261410,
+  "<dialect_zeaol>": 261411,
+  "<dialect_zhiben>": 261412,
+  "<dialect_zhuoqun>": 261413,
+  "<dom_classroom_context>": 261414,
+  "<dom_culture>": 261415,
+  "<dom_dictionary>": 261416,
+  "<dom_essays>": 261417,
+  "<dom_formosan_100_paiwan_texts>": 261418,
+  "<dom_formosan_academia_sinica_oral_legends>": 261419,
+  "<dom_formosan_adlay_kun_long_liu_syntactic_interactio>": 261420,
+  "<dom_formosan_adlay_liu_lfg_relativisation_jianshi_sq>": 261421,
+  "<dom_formosan_amis_adversative_constructions>": 261422,
+  "<dom_formosan_amis_huang_english_phonology_grammar>": 261423,
+  "<dom_formosan_amis_kavalan_lin_interrogative_verbs>": 261424,
+  "<dom_formosan_amis_kuo_sung_comparative_constructions>": 261425,
+  "<dom_formosan_amis_myths_and_customs>": 261426,
+  "<dom_formosan_amis_pa_verbs>": 261427,
+  "<dom_formosan_amis_serial_verb_constructions>": 261428,
+  "<dom_formosan_amis_tung_chiou>": 261429,
+  "<dom_formosan_amy_lee_body_part_nomenclature_seediq>": 261430,
+  "<dom_formosan_anna_hsiou_chuan_chang_reference_gramma>": 261431,
+  "<dom_formosan_anton_quack_puyuma_pulingaw>": 261432,
+  "<dom_formosan_apay_tang_anger_happiness_truku_seediq>": 261433,
+  "<dom_formosan_asai_sedik_language>": 261434,
+  "<dom_formosan_atayal_you_nao_savak_raxayal>": 261435,
+  "<dom_formosan_bien_daniel_chiang_house_social_hierarc>": 261436,
+  "<dom_formosan_boyizhenu_narrative_oral_literature_tap>": 261437,
+  "<dom_formosan_bunun_debusser_dissertation>": 261438,
+  "<dom_formosan_bunun_topic_focus>": 261439,
+  "<dom_formosan_c_douglas_chretien_oceanic_material_in>": 261440,
+  "<dom_formosan_catherine_tseng_seediq_atayal_foods_the>": 261441,
+  "<dom_formosan_chang_kwo_tan_syncretic_objects_materia>": 261442,
+  "<dom_formosan_chaolin_li_motion_verb_grammaticalizati>": 261443,
+  "<dom_formosan_chen_fukuda_relabeling_ergative>": 261444,
+  "<dom_formosan_chen_fukuda_three_ways_steal>": 261445,
+  "<dom_formosan_cheng_sung_modality_kanakanavu>": 261446,
+  "<dom_formosan_chi_lu_chen_and_micheal_d_coe_an_invest>": 261447,
+  "<dom_formosan_chuming_wu_linking_constructions_mayrin>": 261448,
+  "<dom_formosan_chun_mei_chen_comparative_phonology_pai>": 261449,
+  "<dom_formosan_chunming_wu_two_types_of_noun_incorpora>": 261450,
+  "<dom_formosan_cip_atayal_grammar_overview>": 261451,
+  "<dom_formosan_claire_mcgill_a_brief_tayal_vocabulary>": 261452,
+  "<dom_formosan_david_blundell_ed_austronesian_taiwan>": 261453,
+  "<dom_formosan_dean_johnson_year_clouds_smangus>": 261454,
+  "<dom_formosan_der_hwa_victoria_rau_grammar_atayal>": 261455,
+  "<dom_formosan_der_hwa_victoria_rau_lexical_similarity>": 261456,
+  "<dom_formosan_dong_manv_yami_ode_to_taro>": 261457,
+  "<dom_formosan_dong_yi_lin_universal_quantifier_kavala>": 261458,
+  "<dom_formosan_dong_yijia_tjuwabar_paiwan_ritual_langu>": 261459,
+  "<dom_formosan_dorinda_tsai_hsiu_liu_neutral_imperfect>": 261460,
+  "<dom_formosan_ed_torjesen_amis_taiwan_word_list>": 261461,
+  "<dom_formosan_egerod_origin_headhunting_atayal_text_v>": 261462,
+  "<dom_formosan_egerod_soren_word_order_word_classes_at>": 261463,
+  "<dom_formosan_elizabeth_zeitoun_chen_huei_wu_overview>": 261464,
+  "<dom_formosan_elizabeth_zeitoun_lillian_huang_anna_ch>": 261465,
+  "<dom_formosan_elizabeth_zeitoun_lillian_huang_marie_y>": 261466,
+  "<dom_formosan_elizabeth_zeitoun_pronominal_system_man>": 261467,
+  "<dom_formosan_elizabeth_zeitoun_squib_dynamic_vs_stat>": 261468,
+  "<dom_formosan_epark>": 261469,
+  "<dom_formosan_ferrell_taiwan_aboriginal_groups_proble>": 261470,
+  "<dom_formosan_george_leroy_shelley_vudai_rukai>": 261471,
+  "<dom_formosan_gitbook_translations>": 261472,
+  "<dom_formosan_glosbe>": 261473,
+  "<dom_formosan_gujing_lin_tsou_serial_verb_constructio>": 261474,
+  "<dom_formosan_henry_chang_nominal_aspect_tsou>": 261475,
+  "<dom_formosan_henry_chang_tsou_causative_applicative>": 261476,
+  "<dom_formosan_henry_y_chang_formosan_speech_act_mood>": 261477,
+  "<dom_formosan_ho_dan_phonological_system_butonglu_pai>": 261478,
+  "<dom_formosan_holmer_seediq>": 261479,
+  "<dom_formosan_hsiu_chuan_liao_transitivity_ergativity>": 261480,
+  "<dom_formosan_huang_dongqiu_atayal_bunun_amis_languag>": 261481,
+  "<dom_formosan_huang_grammaticalization_squliq_atayal>": 261482,
+  "<dom_formosan_huang_hui_chuan_glide_formation_takitud>": 261483,
+  "<dom_formosan_huang_mei_chin_atayal_reference_grammar>": 261484,
+  "<dom_formosan_hui_chuan_j_huang_syllable_types_in_bun>": 261485,
+  "<dom_formosan_hui_shan_lin_squliq_atayal_reduplicatio>": 261486,
+  "<dom_formosan_huteson_rukai_survey>": 261487,
+  "<dom_formosan_ian_maddieson_and_richard_wright_the_vo>": 261488,
+  "<dom_formosan_ilrdf_42_language_practice_word_lists>": 261489,
+  "<dom_formosan_ilrdf_tousvusvutu_kita>": 261490,
+  "<dom_formosan_indigenous_education_resource_center_ke>": 261491,
+  "<dom_formosan_indigenous_language_literary_awards>": 261492,
+  "<dom_formosan_isabelle_bril_roots_and_stems_lexical_a>": 261493,
+  "<dom_formosan_jian_iwan_guo_qingliu_life_history>": 261494,
+  "<dom_formosan_jian_shilang_introduction_thao_grammar>": 261495,
+  "<dom_formosan_john_tu_er_wei_notes_on_the_religious_b>": 261496,
+  "<dom_formosan_john_u_wolff_the_proto_austronesian_pho>": 261497,
+  "<dom_formosan_jonathan_kuo_amis_manner_result_verbs>": 261498,
+  "<dom_formosan_josiane_cauquelin_aborigines_taiwan_puy>": 261499,
+  "<dom_formosan_joy_wu_clausal_modifiers_in_amis>": 261500,
+  "<dom_formosan_joy_wu_verb_classification_case_marking>": 261501,
+  "<dom_formosan_kanakanavu_texts>": 261502,
+  "<dom_formosan_kavalan_zhang_reference_grammar>": 261503,
+  "<dom_formosan_kumu_tapas_tribal_memory_oral_history_w>": 261504,
+  "<dom_formosan_kuo_j_c_argument_alternation_and_argume>": 261505,
+  "<dom_formosan_lei_shih_family_system_paiwan_at_su_pai>": 261506,
+  "<dom_formosan_li_chen_yeh_developing_a_hla_alua_learn>": 261507,
+  "<dom_formosan_li_may_sung_budai_rukai_exclamatives>": 261508,
+  "<dom_formosan_li_may_sung_clausal_nominalization_buda>": 261509,
+  "<dom_formosan_lillian_huang_atayal_participants_cross>": 261510,
+  "<dom_formosan_lillian_huang_ergativity_atayal>": 261511,
+  "<dom_formosan_liu_manyi_kaskun_ata_matas_i_ki_quaz>": 261512,
+  "<dom_formosan_malcom_d_ross_the_history_and_transitiv>": 261513,
+  "<dom_formosan_malcom_ross_proto_austronesian_verbal_m>": 261514,
+  "<dom_formosan_man_ni_chu_the_measurement_of_final_i_v>": 261515,
+  "<dom_formosan_matsuzawa_kazuko_the_social_and_ritual>": 261516,
+  "<dom_formosan_maya_yeh_blaq_uv_construction_atayal>": 261517,
+  "<dom_formosan_mayumi_oiwa_adverbial_verb_construction>": 261518,
+  "<dom_formosan_mayumi_oiwa_bungard_morphology_and_synt>": 261519,
+  "<dom_formosan_mayumi_oiwa_possessor_raising_truku_see>": 261520,
+  "<dom_formosan_mei_chin_huang_verb_classification_in_m>": 261521,
+  "<dom_formosan_melissa_shih_hui_lin_sakizaya_or_amis_a>": 261522,
+  "<dom_formosan_nanang_tadaw_naci_mowna_pgagu>": 261523,
+  "<dom_formosan_naomi_tsukida_correlative_clauses_seedi>": 261524,
+  "<dom_formosan_naomi_tsukida_ditransitive_causative_se>": 261525,
+  "<dom_formosan_nowbucyang_truku_thesis>": 261526,
+  "<dom_formosan_ochiai_izumi_2016_comparative_linguisti>": 261527,
+  "<dom_formosan_ochiai_izumi_bu_hwan_vocabulary_recorde>": 261528,
+  "<dom_formosan_ochiai_izumi_numerals_paran_seediq_rela>": 261529,
+  "<dom_formosan_ochiai_izumi_on_a_seediq_glossary_recor>": 261530,
+  "<dom_formosan_old_texts>": 261531,
+  "<dom_formosan_paiwan_ho_five_dialects>": 261532,
+  "<dom_formosan_paiwanstories>": 261533,
+  "<dom_formosan_pang_hsin_ting_reconstruction_proto_puy>": 261534,
+  "<dom_formosan_patricia_stanley_morphophonemics_of_ver>": 261535,
+  "<dom_formosan_paul_jen_kuei_li_comparative_bunun_dial>": 261536,
+  "<dom_formosan_paul_jen_kuei_li_conjunction_thao>": 261537,
+  "<dom_formosan_paul_jen_kuei_li_internal_relationships>": 261538,
+  "<dom_formosan_paul_jen_kuei_li_origins_of_the_east_fo>": 261539,
+  "<dom_formosan_paul_jen_kuei_li_review_of_taiwan_abori>": 261540,
+  "<dom_formosan_paul_jen_kuei_li_rukai_structure>": 261541,
+  "<dom_formosan_paul_jen_kuei_li_some_plant_names_in_fo>": 261542,
+  "<dom_formosan_paul_jen_kuei_li_the_preglottalised_sto>": 261543,
+  "<dom_formosan_paul_li_atayalic_final_voiced_stops>": 261544,
+  "<dom_formosan_paul_li_case_marking_four_formosan_lang>": 261545,
+  "<dom_formosan_paul_li_position_atayal_austronesian>": 261546,
+  "<dom_formosan_pingdong_county_government_kai_na_sepuc>": 261547,
+  "<dom_formosan_puyuma_chen_raising_to_object>": 261548,
+  "<dom_formosan_puyuma_katipol_kopfjagdriten>": 261549,
+  "<dom_formosan_puyuma_teng_pronominal_systems>": 261550,
+  "<dom_formosan_puyuma_teng_reference_grammar>": 261551,
+  "<dom_formosan_raleigh_ferrell_construction_markers_fo>": 261552,
+  "<dom_formosan_raleigh_ferrell_paiwan_phonology_and_pr>": 261553,
+  "<dom_formosan_raleigh_j_ferrell_intent_and_volition_i>": 261554,
+  "<dom_formosan_rik_bunun>": 261555,
+  "<dom_formosan_rik_de_busser_the_influence_of_christia>": 261556,
+  "<dom_formosan_robert_blust_a_note_on_covert_structure>": 261557,
+  "<dom_formosan_robert_blust_austronesian_homeland_ling>": 261558,
+  "<dom_formosan_robert_blust_some_remarks_linguistic_po>": 261559,
+  "<dom_formosan_robert_blust_thao_triplication>": 261560,
+  "<dom_formosan_robert_blust_the_austronesian_languages>": 261561,
+  "<dom_formosan_rukai_mantauran_stories>": 261562,
+  "<dom_formosan_rukai_texts>": 261563,
+  "<dom_formosan_rukai_zeitoun_saying>": 261564,
+  "<dom_formosan_sakizaya_affixes>": 261565,
+  "<dom_formosan_schetelig_mittheilungen_uber_die_sprach>": 261566,
+  "<dom_formosan_seals>": 261567,
+  "<dom_formosan_seediq_zhang_reference_grammar>": 261568,
+  "<dom_formosan_shigeru_tsuchida_and_isidore_dyen_proto>": 261569,
+  "<dom_formosan_shigeru_tsuchida_kanakanavu_texts>": 261570,
+  "<dom_formosan_shigeru_tsuchida_preliminary_reports_on>": 261571,
+  "<dom_formosan_shigeru_tsuchida_reconstruction_proto_t>": 261572,
+  "<dom_formosan_shih_chi_stella_yeh_stress_and_quantity>": 261573,
+  "<dom_formosan_shih_rukai_adverbial>": 261574,
+  "<dom_formosan_sihwei_chen_and_haowen_jiang_ways_of_ta>": 261575,
+  "<dom_formosan_sihwei_chen_lexical_aspect_atayal>": 261576,
+  "<dom_formosan_soren_egerod_statement_atayal_phonology>": 261577,
+  "<dom_formosan_soren_egerod_verb_inflection_atayal>": 261578,
+  "<dom_formosan_stacy_teng_malcom_ross_is_puyuma_primar>": 261579,
+  "<dom_formosan_stacy_wan_tin_huang_functions_yami_verb>": 261580,
+  "<dom_formosan_sung_kuo_descriptive_comparative_constr>": 261581,
+  "<dom_formosan_taoshan_elementary_school_atelier_hui_k>": 261582,
+  "<dom_formosan_teresa_m_chen_verbal_construction_and_v>": 261583,
+  "<dom_formosan_ting_chi_wei_parallelism_in_amis_sluici>": 261584,
+  "<dom_formosan_tingchun_chen_successive_cyclic_case_as>": 261585,
+  "<dom_formosan_truku_lowking_demonstratives>": 261586,
+  "<dom_formosan_tsai_hsui_liu_complementation_three_lan>": 261587,
+  "<dom_formosan_tsai_wei_tien_dylan_conjunctive_reducti>": 261588,
+  "<dom_formosan_tsou_descriptive_study>": 261589,
+  "<dom_formosan_tsukida_naomi_verb_classification_amis>": 261590,
+  "<dom_formosan_tu_wen_chiu_a_synchronic_classification>": 261591,
+  "<dom_formosan_tung_tsuchida_ting_li_pan_saaroa_texts>": 261592,
+  "<dom_formosan_victoria_rau_yi_hsin_wu_and_meng_chien>": 261593,
+  "<dom_formosan_wei_huilin_liu_social_structure_yami>": 261594,
+  "<dom_formosan_wei_matri_clan_lineage_system_ami>": 261595,
+  "<dom_formosan_wei_tien_tsai_subjecthood_temporal_adju>": 261596,
+  "<dom_formosan_wiedfelt_formosan_aborigines_atayal>": 261597,
+  "<dom_formosan_wilang_yutas_videos>": 261598,
+  "<dom_formosan_wu_chunming_adverbials_in_paiwan>": 261599,
+  "<dom_formosan_xie_fuhui_introduction_kavalan_grammar>": 261600,
+  "<dom_formosan_yedda_palemeq_blog>": 261601,
+  "<dom_formosan_yeddas_blog>": 261602,
+  "<dom_formosan_yeh_meili_introduction_saisiyat_grammar>": 261603,
+  "<dom_formosan_yi_ming_chou_object_control_saisiyat>": 261604,
+  "<dom_formosan_yi_ting_chen_a_minimalist_approach_to_a>": 261605,
+  "<dom_formosan_yi_ting_chen_amis_left_periphery>": 261606,
+  "<dom_formosan_yi_ting_chen_language_use_code_mixing_a>": 261607,
+  "<dom_formosan_yi_yang_cheng_kanakanavu_word_level_pro>": 261608,
+  "<dom_formosan_yi_yang_cheng_tense_aspect_agent_markin>": 261609,
+  "<dom_formosan_yoshiro_nihira_bunun_vocabulary>": 261610,
+  "<dom_formosan_yuehchien_chien_the_lexical_system_of_y>": 261611,
+  "<dom_formosan_yukihiro_yamada_ying_chu_liao_phonology>": 261612,
+  "<dom_formosan_zeng_shifen_reduplication_affixation_pa>": 261613,
+  "<dom_formosan_zhan_sujuan_taiwan_indigenous_peoples_h>": 261614,
+  "<dom_formosan_zhang_xiujuan_introduction_paiwan_gramm>": 261615,
+  "<dom_formosan_zhang_yongli_pan_jiarong_introduction_t>": 261616,
+  "<dom_formosan_zhao_shanhe_li_taiyuan_zhu_qingyi_abori>": 261617,
+  "<dom_formosan_zheng_acl_2024>": 261618,
+  "<dom_formosan_zheng_data>": 261619,
+  "<dom_formosan_zheng_yumei_zhu_qingyi_bo_hongming_abor>": 261620,
+  "<dom_learning_vocab>": 261621,
+  "<dom_nine_level>": 261622,
+  "<dom_ntu>": 261623,
+  "<dom_picture_book>": 261624,
+  "<dom_picture_story>": 261625,
+  "<dom_presidential_apology>": 261626,
+  "<dom_reading_writing>": 261627,
+  "<dom_unknown>": 261628,
+  "<dom_youtube>": 261629,
+  "<src_ami>": 261630,
+  "<src_bnn>": 261631,
+  "<src_ckv>": 261632,
+  "<src_dru>": 261633,
+  "<src_eng>": 261634,
+  "<src_pwn>": 261635,
+  "<src_pyu>": 261636,
+  "<src_ssf>": 261637,
+  "<src_sxr>": 261638,
+  "<src_szy>": 261639,
+  "<src_tao>": 261640,
+  "<src_tay>": 261641,
+  "<src_trv>": 261642,
+  "<src_tsu>": 261643,
+  "<src_xnb>": 261644,
+  "<src_xsy>": 261645,
+  "<src_zh>": 261646,
+  "<to_ami>": 261647,
+  "<to_bnn>": 261648,
+  "<to_ckv>": 261649,
+  "<to_dru>": 261650,
+  "<to_eng>": 261651,
+  "<to_pwn>": 261652,
+  "<to_pyu>": 261653,
+  "<to_ssf>": 261654,
+  "<to_sxr>": 261655,
+  "<to_szy>": 261656,
+  "<to_tao>": 261657,
+  "<to_tay>": 261658,
+  "<to_trv>": 261659,
+  "<to_tsu>": 261660,
+  "<to_xnb>": 261661,
+  "<to_xsy>": 261662,
+  "<to_zh>": 261663,
+  "ami_Latn": 261357,
+  "bnn_Latn": 261358,
+  "ckv_Latn": 261359,
+  "dru_Latn": 261360,
+  "pwn_Latn": 261361,
+  "pyu_Latn": 261362,
+  "ssf_Latn": 261363,
+  "sxr_Latn": 261364,
+  "szy_Latn": 261365,
+  "tao_Latn": 261366,
+  "tay_Latn": 261367,
+  "trv_Latn": 261368,
+  "tsu_Latn": 261369,
+  "xnb_Latn": 261370,
+  "xsy_Latn": 261371
 }

config.json

@@ -30,5 +30,5 @@
   "tokenizer_class": "NllbTokenizer",
   "transformers_version": "4.56.1",
   "use_cache": true,
-  "vocab_size": 259804
+  "vocab_size": 261664
 }

eval/metrics.json

@@ -1,511 +1,1411 @@
 {
-  "input": "/projects/prudlab/formosan_parallel_corpora/private_no_bible/big_corpus_zh_in_domain_hard.csv",
-  "model": "/scratch/scheppat/projects/mt/formosan_mt_experiments/runs/private_no_bible/v1_spm8192_f2zh_20260712-232900/best",
-  "tokenizer": "/scratch/scheppat/projects/mt/formosan_mt_experiments/runs/private_no_bible/v1_spm8192_f2zh_20260712-232900/best",
+  "schema_version": 3,
+  "complete": true,
+  "profile": {
+    "path": "training_profile.json",
+    "sha256": "34f45832bdedc1b8322b26936dd88667dfecb360d941a7f54f60ae31a1fb4e10",
+    "recipe_id": "nllb200-spm8k-directional-v3"
+  },
+  "mt_standardization": {
+    "id": "formosan-mt-standard-v3",
+    "sha256": "4bbded87eb4833b2d1cd5a88a1f2b059560416e0e69ecfa55433a77f418c903e",
+    "namespace": "formosan-mt"
+  },
+  "model_family": "nllb",
+  "contract": {
+    "input_sha256": "238b96ea87716a749391d7157ce99f1611c1d4a489c4c108c315553505ca1f29",
+    "corpus_manifest_sha256": "0179c1fa87c990f3e4b4fe9615d1a0a9265a1afbd222c7cabe2393f21eaffe25",
+    "validation_report_sha256": "7cba1a2c4661189e63e10fb16cdbc64c271fd32be3cab19771d148d5288c1afa",
+    "run_contract_sha256": "4cfe683caf4d25810d668f761921cb154a2d41e4e83822352faacc2d5b14293f",
+    "checkpoint_metadata_sha256": "31bfbf36e640caeaed0f3b61e7b3d8661cf77bdece1179561a6d21bf29b52859"
+  },
+  "input": "private_no_bible:chinese",
+  "model": "FormosanBank/nllb200-formosan-zh-spm8k",
+  "tokenizer": "FormosanBank/nllb200-formosan-zh-spm8k",
   "direction": "f2zh",
   "target_lang": "chinese",
   "bleu_tokenize": "zh",
+  "lowercase_bleu": false,
   "split": "test",
-  "samples": 61384,
+  "samples": 64356,
+  "headline_metadata_mode": "default",
   "global": {
-    "samples": 61384,
-    "BLEU": 10.576691446569214,
-    "chrF2": 12.391218687865106,
-    "TER": 115.75207512366899,
-    "exact_match_rate": 0.01397758373517529,
-    "empty_output_rate": 0.0,
-    "character_length_ratio": 0.6925750688613788
+    "samples": 64356,
+    "BLEU": 12.580738587412997,
+    "chrF2": 12.96470590849762,
+    "TER": 82.85484609070636,
+    "exact_match_rate": 0.0069457393250046615,
+    "empty_output_rate": 0.00013984710050344958,
+    "character_length_ratio": 0.9735837589431371,
+    "signatures": {
+      "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+      "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+      "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+    }
+  },
+  "metadata_modes": {
+    "default": {
+      "samples": 64356,
+      "BLEU": 12.580738587412997,
+      "chrF2": 12.96470590849762,
+      "TER": 82.85484609070636,
+      "exact_match_rate": 0.0069457393250046615,
+      "empty_output_rate": 0.00013984710050344958,
+      "character_length_ratio": 0.9735837589431371,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    }
+  },
+  "metadata_fallbacks": {
+    "default": {
+      "domain_fallback_rows": 64356,
+      "dialect_fallback_rows": 64356
+    }
+  },
+  "bootstrap_95_ci": {
+    "status": "skipped",
+    "samples": 0,
+    "seed": 42
   },
   "by_language": {
     "ami": {
-      "samples": 10929,
-      "BLEU": 7.907158388525328,
-      "chrF2": 10.070412232114371,
-      "TER": 115.26327734906945,
-      "exact_match_rate": 0.010613962851130022,
+      "samples": 11271,
+      "BLEU": 13.1918555311043,
+      "chrF2": 13.378851866289018,
+      "TER": 80.5114298774068,
+      "exact_match_rate": 0.0030165912518853697,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.638478344685773
+      "character_length_ratio": 0.9545612209503989,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "bnn": {
-      "samples": 5450,
-      "BLEU": 9.239755649716296,
-      "chrF2": 10.905172805141694,
-      "TER": 120.58014216922723,
-      "exact_match_rate": 0.011376146788990826,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6297844327645479
+      "samples": 6112,
+      "BLEU": 11.10412132497452,
+      "chrF2": 11.8546422628787,
+      "TER": 88.67577392432739,
+      "exact_match_rate": 0.005726439790575916,
+      "empty_output_rate": 0.0001636125654450262,
+      "character_length_ratio": 1.0428235373472847,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "ckv": {
-      "samples": 2852,
-      "BLEU": 15.762605478682843,
-      "chrF2": 16.76866562299388,
-      "TER": 108.04521276595744,
-      "exact_match_rate": 0.011921458625525946,
+      "samples": 3038,
+      "BLEU": 15.171470002087373,
+      "chrF2": 14.808325172166564,
+      "TER": 76.72054372037238,
+      "exact_match_rate": 0.005924950625411455,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7161089052750993
+      "character_length_ratio": 0.9807046979865772,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "dru": {
-      "samples": 5524,
-      "BLEU": 8.136977886054543,
-      "chrF2": 10.050861113586718,
-      "TER": 127.32970671712394,
-      "exact_match_rate": 0.014301230992034758,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6891913918113896
+      "samples": 6109,
+      "BLEU": 5.940996423160562,
+      "chrF2": 7.037003101226312,
+      "TER": 102.81002819723521,
+      "exact_match_rate": 0.0022917007693566867,
+      "empty_output_rate": 0.0004910787362907187,
+      "character_length_ratio": 1.0361885969307136,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "pwn": {
-      "samples": 5519,
-      "BLEU": 8.80996385561619,
-      "chrF2": 10.986609677668069,
-      "TER": 111.14936146585231,
-      "exact_match_rate": 0.010871534698314912,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6481772904433953
+      "samples": 5718,
+      "BLEU": 8.521265940455624,
+      "chrF2": 9.674811375535183,
+      "TER": 93.4649114743248,
+      "exact_match_rate": 0.0013990905911157748,
+      "empty_output_rate": 0.0003497726477789437,
+      "character_length_ratio": 1.0221799857015816,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "pyu": {
-      "samples": 4105,
-      "BLEU": 13.101948522565348,
-      "chrF2": 15.116067674782407,
-      "TER": 107.76247540690396,
-      "exact_match_rate": 0.01900121802679659,
+      "samples": 4280,
+      "BLEU": 19.798357966901257,
+      "chrF2": 18.771919267723025,
+      "TER": 73.93693575821013,
+      "exact_match_rate": 0.0308411214953271,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.680589070537938
+      "character_length_ratio": 0.9459577012994074,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "ssf": {
-      "samples": 1670,
-      "BLEU": 14.866297961190686,
-      "chrF2": 16.293901223953522,
-      "TER": 110.55618615209988,
-      "exact_match_rate": 0.015568862275449102,
+      "samples": 1673,
+      "BLEU": 15.879819541690464,
+      "chrF2": 17.03540529123015,
+      "TER": 70.98675877950488,
+      "exact_match_rate": 0.005977286312014346,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.724459557930532
+      "character_length_ratio": 0.9727298044172147,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "sxr": {
-      "samples": 1539,
-      "BLEU": 8.570268882776618,
-      "chrF2": 11.04003820559944,
-      "TER": 109.40737670949025,
-      "exact_match_rate": 0.010396361273554255,
+      "samples": 1781,
+      "BLEU": 13.114989144321365,
+      "chrF2": 13.68203020438861,
+      "TER": 76.028900676788,
+      "exact_match_rate": 0.005614823133071308,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.715325101193831
+      "character_length_ratio": 0.9551852007922127,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "szy": {
-      "samples": 2147,
-      "BLEU": 12.583295396175457,
-      "chrF2": 15.787293357235924,
-      "TER": 110.10262989095574,
-      "exact_match_rate": 0.021891010712622262,
+      "samples": 2055,
+      "BLEU": 12.490528817195182,
+      "chrF2": 14.561259511691485,
+      "TER": 75.3939821068499,
+      "exact_match_rate": 0.0038929440389294406,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6639051190372373
+      "character_length_ratio": 0.9021482336135377,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "tao": {
-      "samples": 1802,
-      "BLEU": 8.318439478906354,
-      "chrF2": 11.299683524234275,
-      "TER": 109.37062937062937,
-      "exact_match_rate": 0.009988901220865706,
+      "samples": 1838,
+      "BLEU": 10.60189532011366,
+      "chrF2": 12.097253808423574,
+      "TER": 86.38102865278519,
+      "exact_match_rate": 0.003808487486398259,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6603234122334193
+      "character_length_ratio": 1.0268809557673162,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "tay": {
-      "samples": 6395,
-      "BLEU": 8.915637694044362,
-      "chrF2": 10.23849545512742,
-      "TER": 125.91745218266679,
-      "exact_match_rate": 0.012197028928850665,
+      "samples": 6723,
+      "BLEU": 10.030658272767216,
+      "chrF2": 10.982138055853184,
+      "TER": 78.01421413446823,
+      "exact_match_rate": 0.007437156031533542,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7223793403290757
+      "character_length_ratio": 0.8772094926350246,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "trv": {
-      "samples": 6689,
-      "BLEU": 17.65835020427611,
-      "chrF2": 18.00769858472333,
-      "TER": 107.97406438402912,
-      "exact_match_rate": 0.023620870085214532,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.881449702993361
+      "samples": 6983,
+      "BLEU": 14.434006728137819,
+      "chrF2": 15.235502409976897,
+      "TER": 80.84272268857266,
+      "exact_match_rate": 0.011313189173707576,
+      "empty_output_rate": 0.00028640985249892595,
+      "character_length_ratio": 0.960600440397052,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "tsu": {
-      "samples": 1889,
-      "BLEU": 8.004987496560164,
-      "chrF2": 10.402536265430221,
-      "TER": 115.47044632086852,
-      "exact_match_rate": 0.009528851244044468,
+      "samples": 2164,
+      "BLEU": 11.895805977241574,
+      "chrF2": 12.90299438484368,
+      "TER": 80.67225218412307,
+      "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6691029709510202
+      "character_length_ratio": 0.9650275854408117,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "xnb": {
-      "samples": 2927,
-      "BLEU": 17.479049560336957,
-      "chrF2": 18.514675122197012,
-      "TER": 111.06577052328373,
-      "exact_match_rate": 0.01400751622822002,
+      "samples": 2728,
+      "BLEU": 19.39856378760136,
+      "chrF2": 18.488341275928676,
+      "TER": 69.2817191639682,
+      "exact_match_rate": 0.013196480938416423,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7461520246270423
+      "character_length_ratio": 0.9773734780253946,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "xsy": {
-      "samples": 1947,
-      "BLEU": 9.938882396702315,
-      "chrF2": 12.24802454126081,
-      "TER": 125.58068459657702,
-      "exact_match_rate": 0.01386748844375963,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7180976234053972
+      "samples": 1883,
+      "BLEU": 11.518316944569966,
+      "chrF2": 12.476203336723275,
+      "TER": 77.74666348903719,
+      "exact_match_rate": 0.003186404673393521,
+      "empty_output_rate": 0.0005310674455655868,
+      "character_length_ratio": 0.9228345005528935,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     }
   },
   "by_source_bucket": {
     "Formosan-Academia-Sinica-Oral-Legends": {
-      "samples": 4,
-      "BLEU": 13.708780144263425,
-      "chrF2": 15.041820333450085,
-      "TER": 100.0,
+      "samples": 260,
+      "BLEU": 1.5184047323368557,
+      "chrF2": 3.607752879476709,
+      "TER": 147.0039364338825,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6444444444444445
+      "character_length_ratio": 1.3767271742075318,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Amis-Lifok-Dongi-Saopo-Kimakimad": {
-      "samples": 1604,
-      "BLEU": 2.3379152308621682,
-      "chrF2": 5.912072736378735,
-      "TER": 146.6566113624937,
+    "Formosan-Boyizhenu-Narrative-Oral-Literature-Tapangu": {
+      "samples": 20,
+      "BLEU": 0.0,
+      "chrF2": 0.0,
+      "TER": 114.39999999999999,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6363730379426839
+      "character_length_ratio": 0.872,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Amis_myths_and_customs": {
-      "samples": 7,
-      "BLEU": 1.6263246843084138,
-      "chrF2": 4.5570794176045215,
-      "TER": 100.0,
+    "Formosan-Bunun-Moriguchi-Northern-Texts": {
+      "samples": 382,
+      "BLEU": 4.556482074148768,
+      "chrF2": 7.167874351913245,
+      "TER": 96.64069264069263,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.36831604150039904
+      "character_length_ratio": 0.9919803600654664,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-CIP-Atayal-Grammar-Overview": {
-      "samples": 56,
-      "BLEU": 17.073688928053834,
-      "chrF2": 18.201008148785526,
-      "TER": 93.24324324324324,
-      "exact_match_rate": 0.07142857142857142,
+      "samples": 33,
+      "BLEU": 26.084925543784756,
+      "chrF2": 35.72891315451452,
+      "TER": 60.1027397260274,
+      "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.8041692987997473
+      "character_length_ratio": 1.0418918918918918,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Formosan-Deng-Fangqing-Introduction-Puyuma-Grammar": {
+      "samples": 376,
+      "BLEU": 24.499437200984158,
+      "chrF2": 28.174278345762676,
+      "TER": 58.88787602552416,
+      "exact_match_rate": 0.06914893617021277,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9226395939086295,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-Glosbe": {
-      "samples": 21,
-      "BLEU": 3.3978952817507193,
-      "chrF2": 6.608750008028403,
-      "TER": 100.0,
-      "exact_match_rate": 0.0,
+      "samples": 2186,
+      "BLEU": 14.93117749734157,
+      "chrF2": 14.478334396414597,
+      "TER": 73.35618631458007,
+      "exact_match_rate": 0.0009149130832570906,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.8833333333333333
+      "character_length_ratio": 0.913203808685123,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-Hala-Saku-La-Videos": {
-      "samples": 24,
-      "BLEU": 3.871055448346106,
-      "chrF2": 9.345209439285906,
-      "TER": 108.8235294117647,
+      "samples": 47,
+      "BLEU": 5.061816442197886,
+      "chrF2": 8.42442769014718,
+      "TER": 83.09919114516816,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.5491445823549145
+      "character_length_ratio": 0.6130952380952381,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Formosan-Huang-Huijuan-Shi-Chaokai-Introduction-Bunun-Grammar": {
+      "samples": 716,
+      "BLEU": 12.189672284506477,
+      "chrF2": 13.740446665721755,
+      "TER": 76.70538348082596,
+      "exact_match_rate": 0.013966480446927373,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.7565281655444244,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Huang-Mei-Chin-Atayal-Reference-Grammar": {
-      "samples": 2,
-      "BLEU": 15.154686451352852,
-      "chrF2": 30.96466834342189,
-      "TER": 100.0,
+    "Formosan-ILRDF-42-Language-Practice-Word-Lists": {
+      "samples": 3,
+      "BLEU": 3.8815952835694683,
+      "chrF2": 3.424641860039361,
+      "TER": 88.88888888888889,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6216216216216216
+      "character_length_ratio": 0.6296296296296297,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-ILRDF-Tousvusvutu-Kita": {
-      "samples": 738,
-      "BLEU": 4.529526769930424,
-      "chrF2": 6.126303744736681,
-      "TER": 122.51297090006767,
-      "exact_match_rate": 0.0,
+      "samples": 497,
+      "BLEU": 5.152060142903115,
+      "chrF2": 6.809234748570317,
+      "TER": 104.84475110892065,
+      "exact_match_rate": 0.004024144869215292,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.004917888820585,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Formosan-Indigenous-Education-Resource-Center-Keelung-City-Serangawan-No": {
+      "samples": 23,
+      "BLEU": 28.15899781413618,
+      "chrF2": 23.667599175837104,
+      "TER": 53.50877192982456,
+      "exact_match_rate": 0.043478260869565216,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6800853427800814
+      "character_length_ratio": 0.9094827586206896,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-Indigenous-Language-Literary-Awards": {
-      "samples": 3206,
-      "BLEU": 1.7454225871296085,
-      "chrF2": 3.2831850026941756,
-      "TER": 177.2076961026147,
+      "samples": 821,
+      "BLEU": 2.412181696193584,
+      "chrF2": 4.5279484158761845,
+      "TER": 100.01695274422546,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7565487244648617
+      "character_length_ratio": 0.934447142092918,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Kavalan-Zhang-Reference-Grammar": {
-      "samples": 2,
-      "BLEU": 13.555499052342826,
-      "chrF2": 12.73179069019462,
-      "TER": 87.09677419354838,
+    "Formosan-Kucapungane": {
+      "samples": 12,
+      "BLEU": 2.892986805656571,
+      "chrF2": 5.051336155822737,
+      "TER": 103.76175548589342,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.9154929577464789
+      "character_length_ratio": 0.9752144899904671,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Li-Peirong-Xu-Weisheng-Introduction-Truku-Grammar": {
-      "samples": 122,
-      "BLEU": 23.717942859641965,
-      "chrF2": 28.708599449512906,
-      "TER": 94.44444444444444,
-      "exact_match_rate": 0.07377049180327869,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.863456090651558
-    },
-    "Formosan-Li-Peizhen-Wu-Xingyu-Shouhu-Xing": {
-      "samples": 6,
-      "BLEU": 1.9271815425189616,
-      "chrF2": 5.233801820077148,
-      "TER": 100.0,
+    "Formosan-Kumu-Tapas-Tribal-Memory-Oral-History-Wushe-Incident-II": {
+      "samples": 71,
+      "BLEU": 2.9635392138545416,
+      "chrF2": 6.484189254367516,
+      "TER": 88.0565138820437,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.5118483412322274
+      "character_length_ratio": 0.6026764377887526,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Nowbucyang-Truku-Thesis": {
-      "samples": 64,
-      "BLEU": 33.03227143741538,
-      "chrF2": 35.4877874425264,
-      "TER": 81.69014084507043,
-      "exact_match_rate": 0.109375,
+    "Formosan-Li-Peirong-Xu-Weisheng-Introduction-Truku-Grammar": {
+      "samples": 468,
+      "BLEU": 22.879843447542882,
+      "chrF2": 24.303648070257005,
+      "TER": 60.86889539362193,
+      "exact_match_rate": 0.042735042735042736,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.803030303030303
+      "character_length_ratio": 0.8839035333707235,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-PaiwanStories": {
-      "samples": 18,
-      "BLEU": 8.869948503140135,
-      "chrF2": 10.358317911576524,
-      "TER": 109.09090909090908,
+      "samples": 13,
+      "BLEU": 3.320115399578185,
+      "chrF2": 7.1048822760617,
+      "TER": 89.1891891891892,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.8608247422680413,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Formosan-Puyuma-Katipol-Kopfjagdriten": {
+      "samples": 38,
+      "BLEU": 3.015861276037805,
+      "chrF2": 4.76287474070312,
+      "TER": 92.85714285714286,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7548262548262549
+      "character_length_ratio": 0.7934782608695652,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-SEALS": {
-      "samples": 16,
-      "BLEU": 5.401492843871145,
-      "chrF2": 6.2423672065754525,
-      "TER": 175.0,
+      "samples": 47,
+      "BLEU": 4.759285283577419,
+      "chrF2": 5.930459034869349,
+      "TER": 111.32756132756131,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.814629258517034
+      "character_length_ratio": 1.1212903225806452,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Seediq-Zhang-Reference-Grammar": {
-      "samples": 15,
-      "BLEU": 10.731865242690281,
-      "chrF2": 17.52649006585302,
-      "TER": 88.23529411764706,
+    "Formosan-Shen-Wenqi-Introduction-Sakizaya-Grammar": {
+      "samples": 67,
+      "BLEU": 16.992614089854257,
+      "chrF2": 17.75684375216057,
+      "TER": 65.62889165628891,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7033898305084746
+      "character_length_ratio": 0.9242979242979243,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Song-Limei-Introduction-Kanakanavu-Grammar": {
-      "samples": 104,
-      "BLEU": 21.222612310827056,
-      "chrF2": 23.44610806619767,
-      "TER": 103.84615384615385,
+    "Formosan-Sing-Olam-Hong-Tinglan-Aboriginal-Education-World-Issue": {
+      "samples": 1,
+      "BLEU": 71.65313105737896,
+      "chrF2": 68.86227544910179,
+      "TER": 25.0,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.8052876027152555
+      "character_length_ratio": 0.75,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-Song-Limei-Introduction-Seediq-Grammar": {
-      "samples": 107,
-      "BLEU": 28.635676865278743,
-      "chrF2": 31.56948663323015,
-      "TER": 95.72649572649573,
-      "exact_match_rate": 0.08411214953271028,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.8722649319929036
-    },
-    "Formosan-Wilang-Yutas-Videos": {
-      "samples": 10,
-      "BLEU": 9.324906605075657,
-      "chrF2": 11.029154358634294,
-      "TER": 100.0,
-      "exact_match_rate": 0.0,
+      "samples": 442,
+      "BLEU": 19.971996882565136,
+      "chrF2": 23.122087305085547,
+      "TER": 64.06685236768801,
+      "exact_match_rate": 0.05203619909502263,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7066666666666667
+      "character_length_ratio": 0.8887895050685748,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-Wu-Jinglan-Introduction-Amis-Grammar": {
-      "samples": 34,
-      "BLEU": 14.528023877356814,
-      "chrF2": 23.12502283686413,
-      "TER": 102.27272727272727,
-      "exact_match_rate": 0.0,
+      "samples": 288,
+      "BLEU": 23.971418217708198,
+      "chrF2": 26.20025205918842,
+      "TER": 56.87943262411348,
+      "exact_match_rate": 0.04861111111111111,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.5162019593067069
+      "character_length_ratio": 0.8612731422774537,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Yeh-Meili-Introduction-Saisiyat-Grammar": {
-      "samples": 4,
-      "BLEU": 3.851546244568978,
-      "chrF2": 6.389276621044812,
-      "TER": 100.0,
-      "exact_match_rate": 0.0,
+    "Formosan-Xie-Fuhui-Introduction-Kavalan-Grammar": {
+      "samples": 33,
+      "BLEU": 27.56170123220992,
+      "chrF2": 34.8621449696623,
+      "TER": 63.52395672333848,
+      "exact_match_rate": 0.06060606060606061,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.3870967741935484
+      "character_length_ratio": 1.011378002528445,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
-    "Formosan-Zhang-Yongli-Pan-Jiarong-Introduction-Tsou-Grammar": {
-      "samples": 37,
-      "BLEU": 6.388565113016769,
-      "chrF2": 9.86522307741866,
-      "TER": 100.0,
-      "exact_match_rate": 0.05405405405405406,
+    "Formosan-Zhao-Shanhe-Li-Taiyuan-Zhu-Qingyi-Aboriginal-Education-World-Issue-16": {
+      "samples": 5,
+      "BLEU": 78.9773161836023,
+      "chrF2": 76.09937642638272,
+      "TER": 14.634146341463413,
+      "exact_match_rate": 0.4,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6717557251908397
+      "character_length_ratio": 1.0,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-Zheng-ACL-2024": {
-      "samples": 4947,
-      "BLEU": 20.15707801656887,
-      "chrF2": 20.056972103148482,
-      "TER": 101.21051636088518,
-      "exact_match_rate": 0.003840711542348898,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7978903211354796
-    },
-    "Formosan-amis_tung-chiou": {
-      "samples": 4,
-      "BLEU": 2.270450618468058,
-      "chrF2": 4.86818640558682,
-      "TER": 102.94117647058823,
-      "exact_match_rate": 0.0,
+      "samples": 11928,
+      "BLEU": 15.780769976983771,
+      "chrF2": 16.42245978564808,
+      "TER": 70.74925721061533,
+      "exact_match_rate": 0.009054325955734407,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.8408941485864563
+      "character_length_ratio": 0.9224855323138599,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-ePark": {
-      "samples": 5491,
-      "BLEU": 36.11130715740376,
-      "chrF2": 32.81108828189628,
-      "TER": 94.7163014109877,
-      "exact_match_rate": 0.07412128938262612,
+      "samples": 5700,
+      "BLEU": 38.26167160586107,
+      "chrF2": 32.28225904343755,
+      "TER": 42.84845553175316,
+      "exact_match_rate": 0.028771929824561403,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7846655150753769
+      "character_length_ratio": 0.9634230597287887,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "Formosan-gitbook_translations": {
-      "samples": 62,
-      "BLEU": 5.336287344219993,
-      "chrF2": 11.250725678537165,
-      "TER": 97.20496894409938,
+      "samples": 51,
+      "BLEU": 3.2141866387955265,
+      "chrF2": 8.521924203356665,
+      "TER": 122.24409448818898,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.59533527696793
+      "character_length_ratio": 1.1219110378912684,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "culture": {
-      "samples": 11335,
-      "BLEU": 4.230999106074682,
-      "chrF2": 6.803049833850721,
-      "TER": 103.05977030765361,
-      "exact_match_rate": 8.82223202470225e-05,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.5262328026781123
+      "samples": 5489,
+      "BLEU": 6.480767384550856,
+      "chrF2": 7.660942069794658,
+      "TER": 97.1531089786563,
+      "exact_match_rate": 0.00018218254691200583,
+      "empty_output_rate": 0.00018218254691200583,
+      "character_length_ratio": 0.9973183587863729,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "essays": {
-      "samples": 3176,
-      "BLEU": 16.22959311401136,
-      "chrF2": 16.18942996487404,
-      "TER": 117.39726027397259,
-      "exact_match_rate": 0.010390428211586901,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7892565422616429
-    },
-    "nine_level": {
-      "samples": 1933,
-      "BLEU": 38.94085170738614,
-      "chrF2": 35.54484444458279,
-      "TER": 91.82939362795479,
-      "exact_match_rate": 0.08277289187790998,
+      "samples": 3722,
+      "BLEU": 16.328664532508153,
+      "chrF2": 15.989650491484719,
+      "TER": 71.99628228108611,
+      "exact_match_rate": 0.008060182697474477,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.8508923757909372
+      "character_length_ratio": 1.013505336474708,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "ntu": {
-      "samples": 2884,
-      "BLEU": 13.473343908627347,
-      "chrF2": 14.92087756037663,
-      "TER": 108.14696485623003,
-      "exact_match_rate": 0.0038141470180305132,
+      "samples": 440,
+      "BLEU": 17.345798902609253,
+      "chrF2": 15.844215543415258,
+      "TER": 76.1023690126232,
+      "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7321011104617183
+      "character_length_ratio": 0.9769941225860621,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "picture_book": {
-      "samples": 13023,
-      "BLEU": 13.78815631073597,
-      "chrF2": 16.69686326560025,
-      "TER": 112.86555398764051,
-      "exact_match_rate": 0.014359210627351609,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7865734687720004
+      "samples": 12634,
+      "BLEU": 12.23364317472049,
+      "chrF2": 13.761258091926761,
+      "TER": 78.72619308231174,
+      "exact_match_rate": 0.002691150862751306,
+      "empty_output_rate": 7.915149596327371e-05,
+      "character_length_ratio": 0.9670261525343629,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "picture_story": {
-      "samples": 729,
-      "BLEU": 6.502675604326252,
-      "chrF2": 8.737286139710305,
-      "TER": 109.9591836734694,
-      "exact_match_rate": 0.0013717421124828531,
+      "samples": 863,
+      "BLEU": 9.080114094143601,
+      "chrF2": 9.65627900464358,
+      "TER": 90.16095665775266,
+      "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6134348528296911
+      "character_length_ratio": 0.9506388767429432,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "presidential_apology": {
-      "samples": 482,
-      "BLEU": 3.4850550479787836,
-      "chrF2": 5.319974289887845,
-      "TER": 108.20930232558139,
+      "samples": 108,
+      "BLEU": 4.400830341996858,
+      "chrF2": 6.4000097247865275,
+      "TER": 113.2823488045007,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.5924246272176469
+      "character_length_ratio": 1.1276301892923877,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "reading_writing": {
-      "samples": 4793,
-      "BLEU": 5.319686455411381,
-      "chrF2": 7.422908022519387,
-      "TER": 150.53610644619556,
-      "exact_match_rate": 0.0,
+      "samples": 1124,
+      "BLEU": 9.71120279354296,
+      "chrF2": 9.901041025507332,
+      "TER": 87.93712212817412,
+      "exact_match_rate": 0.0008896797153024911,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6066766246521526
+      "character_length_ratio": 0.981299672744273,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "youtube": {
-      "samples": 6324,
-      "BLEU": 7.76714599062836,
-      "chrF2": 9.71030705405905,
-      "TER": 113.357928884931,
-      "exact_match_rate": 0.001265022137887413,
+      "samples": 15448,
+      "BLEU": 6.083847184995035,
+      "chrF2": 7.481384089171102,
+      "TER": 101.0536648581735,
+      "exact_match_rate": 0.0004531330916623511,
+      "empty_output_rate": 0.0004531330916623511,
+      "character_length_ratio": 1.0220858782839062,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    }
+  },
+  "by_dialect": {
+    "Central": {
+      "samples": 1462,
+      "BLEU": 6.245837215434144,
+      "chrF2": 7.437307813519742,
+      "TER": 102.61820687279304,
+      "exact_match_rate": 0.002051983584131327,
+      "empty_output_rate": 0.0006839945280437756,
+      "character_length_ratio": 1.0643430520317765,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Coastal": {
+      "samples": 4090,
+      "BLEU": 12.62785293823784,
+      "chrF2": 13.352495811232142,
+      "TER": 82.80232427576915,
+      "exact_match_rate": 0.001466992665036675,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9847345307281669,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Dawu": {
+      "samples": 960,
+      "BLEU": 6.636383121958551,
+      "chrF2": 7.847467315350563,
+      "TER": 104.00590644946175,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.0849292635218124,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "DeluValley": {
+      "samples": 439,
+      "BLEU": 7.162521403162108,
+      "chrF2": 8.501476537286635,
+      "TER": 93.18593946157065,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9764104580302733,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Dona": {
+      "samples": 363,
+      "BLEU": 10.053733462553387,
+      "chrF2": 9.974232937576398,
+      "TER": 91.17067402443831,
+      "exact_match_rate": 0.008264462809917356,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.0028869084975525,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Duda": {
+      "samples": 677,
+      "BLEU": 8.54491075080391,
+      "chrF2": 9.038500417850894,
+      "TER": 90.9516517173485,
+      "exact_match_rate": 0.005908419497784343,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.8946756375114212,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Eastern": {
+      "samples": 2631,
+      "BLEU": 6.740811782877297,
+      "chrF2": 7.895764731761029,
+      "TER": 97.69556854469738,
+      "exact_match_rate": 0.0011402508551881414,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9951046317714778,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "FourSeasons": {
+      "samples": 76,
+      "BLEU": 4.424597259237414,
+      "chrF2": 6.383118262768753,
+      "TER": 105.2596089008766,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.0513530135301352,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Hengchun": {
+      "samples": 234,
+      "BLEU": 15.791912276224933,
+      "chrF2": 15.31039939820434,
+      "TER": 75.69772434521254,
+      "exact_match_rate": 0.008547008547008548,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.0211640211640212,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Jianhe": {
+      "samples": 286,
+      "BLEU": 25.567508607352632,
+      "chrF2": 22.7006312699475,
+      "TER": 61.663239661868865,
+      "exact_match_rate": 0.04195804195804196,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9455693446549096,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Junqun": {
+      "samples": 2663,
+      "BLEU": 10.662204649619605,
+      "chrF2": 11.780536944704087,
+      "TER": 88.93881419594904,
+      "exact_match_rate": 0.004506196019526849,
+      "empty_output_rate": 0.0003755163349605708,
+      "character_length_ratio": 1.0220821214435436,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Kanakanavu": {
+      "samples": 2728,
+      "BLEU": 19.39856378760136,
+      "chrF2": 18.488341275928676,
+      "TER": 69.2817191639682,
+      "exact_match_rate": 0.013196480938416423,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9773734780253946,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Kaqun": {
+      "samples": 172,
+      "BLEU": 16.774848869485286,
+      "chrF2": 15.542877633403249,
+      "TER": 68.89018351296242,
+      "exact_match_rate": 0.01744186046511628,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9372159090909091,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Kavalan": {
+      "samples": 3038,
+      "BLEU": 15.171470002087373,
+      "chrF2": 14.808325172166564,
+      "TER": 76.72054372037238,
+      "exact_match_rate": 0.005924950625411455,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9807046979865772,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Luanqun": {
+      "samples": 155,
+      "BLEU": 17.349484372027945,
+      "chrF2": 17.598349739237214,
+      "TER": 71.81554337310546,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.1005674653215636,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Malan": {
+      "samples": 1230,
+      "BLEU": 10.221167917101575,
+      "chrF2": 10.439882337669687,
+      "TER": 82.39746523499385,
+      "exact_match_rate": 0.0008130081300813008,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.8698444895199459,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Maolin": {
+      "samples": 410,
+      "BLEU": 4.273942963093996,
+      "chrF2": 5.276588781857147,
+      "TER": 100.75223319228961,
+      "exact_match_rate": 0.0024390243902439024,
+      "empty_output_rate": 0.004878048780487805,
+      "character_length_ratio": 0.9729977116704805,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Nanwang": {
+      "samples": 1892,
+      "BLEU": 14.57954039098228,
+      "chrF2": 14.258816665992901,
+      "TER": 85.16901657719309,
+      "exact_match_rate": 0.022198731501057084,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9574596552859809,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Northern": {
+      "samples": 2159,
+      "BLEU": 9.938683098210683,
+      "chrF2": 11.156971342176261,
+      "TER": 88.06784142789266,
+      "exact_match_rate": 0.0013895321908290875,
+      "empty_output_rate": 0.0004631773969430292,
+      "character_length_ratio": 1.0169857104340794,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Saaroa": {
+      "samples": 1781,
+      "BLEU": 13.114989144321365,
+      "chrF2": 13.68203020438861,
+      "TER": 76.028900676788,
+      "exact_match_rate": 0.005614823133071308,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9551852007922127,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Saisiyat": {
+      "samples": 1877,
+      "BLEU": 11.461679154597425,
+      "chrF2": 12.441476730582655,
+      "TER": 77.79412349409566,
+      "exact_match_rate": 0.002131060202450719,
+      "empty_output_rate": 0.0005327650506126798,
+      "character_length_ratio": 0.9228555617044826,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Sakizaya": {
+      "samples": 2055,
+      "BLEU": 12.490528817195182,
+      "chrF2": 14.561259511691485,
+      "TER": 75.3939821068499,
+      "exact_match_rate": 0.0038929440389294406,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9021482336135377,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Sekolik": {
+      "samples": 430,
+      "BLEU": 2.813092140199989,
+      "chrF2": 5.591458672900994,
+      "TER": 93.83398407788651,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.8395688321732405,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Southern": {
+      "samples": 780,
+      "BLEU": 15.796482686409094,
+      "chrF2": 14.845783907869976,
+      "TER": 75.59048898017221,
+      "exact_match_rate": 0.005128205128205128,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9534954639018068,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Tanqun": {
+      "samples": 184,
+      "BLEU": 16.428580485687892,
+      "chrF2": 16.320602885354223,
+      "TER": 71.26141552511416,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.0512963479230555,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Tegudaya": {
+      "samples": 1502,
+      "BLEU": 18.866407115218493,
+      "chrF2": 18.089019543909796,
+      "TER": 71.87598972446071,
+      "exact_match_rate": 0.02796271637816245,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9561662198391421,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Thao": {
+      "samples": 1673,
+      "BLEU": 15.879819541690464,
+      "chrF2": 17.03540529123015,
+      "TER": 70.98675877950488,
+      "exact_match_rate": 0.005977286312014346,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9727298044172147,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Truku": {
+      "samples": 3204,
+      "BLEU": 17.498129570998714,
+      "chrF2": 18.218042405574568,
+      "TER": 76.84859333997844,
+      "exact_match_rate": 0.009987515605493134,
+      "empty_output_rate": 0.0006242197253433209,
+      "character_length_ratio": 1.016496910372456,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Tsou": {
+      "samples": 2164,
+      "BLEU": 11.895805977241574,
+      "chrF2": 12.90299438484368,
+      "TER": 80.67225218412307,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9650275854408117,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Wanda": {
+      "samples": 40,
+      "BLEU": 9.759361343346518,
+      "chrF2": 12.427580196511215,
+      "TER": 84.28005284015853,
+      "exact_match_rate": 0.025,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9300613496932515,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Wanshan": {
+      "samples": 912,
+      "BLEU": 4.783936442989583,
+      "chrF2": 6.165013325791203,
+      "TER": 99.1668616896794,
+      "exact_match_rate": 0.0010964912280701754,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9984970624402241,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Wenshui": {
+      "samples": 35,
+      "BLEU": 12.232564716539688,
+      "chrF2": 12.825466615679723,
+      "TER": 79.42122186495176,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.0926517571884984,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Wutai": {
+      "samples": 2231,
+      "BLEU": 5.560392839668127,
+      "chrF2": 6.605242162459571,
+      "TER": 105.53900221560089,
+      "exact_match_rate": 0.002689376961004034,
+      "empty_output_rate": 0.00044822949350067237,
+      "character_length_ratio": 1.0327670191279554,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Xiqun": {
+      "samples": 634,
+      "BLEU": 38.94182683168665,
+      "chrF2": 32.7957607713137,
+      "TER": 42.73384947328341,
+      "exact_match_rate": 0.055205047318611984,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9568452380952381,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Xiuguluan": {
+      "samples": 2210,
+      "BLEU": 11.03007805840091,
+      "chrF2": 12.433794737348038,
+      "TER": 89.38989352712522,
+      "exact_match_rate": 0.007239819004524887,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9983890327879209,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Yami": {
+      "samples": 1838,
+      "BLEU": 10.60189532011366,
+      "chrF2": 12.097253808423574,
+      "TER": 86.38102865278519,
+      "exact_match_rate": 0.003808487486398259,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.0268809557673162,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "YilanZeaol": {
+      "samples": 390,
+      "BLEU": 2.9104309307826854,
+      "chrF2": 4.647373218475065,
+      "TER": 94.05347264904734,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.7690119760479042,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Zeaol": {
+      "samples": 308,
+      "BLEU": 3.7135631295484477,
+      "chrF2": 5.443722731848967,
+      "TER": 104.93009565857247,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.0196219191194065,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Zhiben": {
+      "samples": 834,
+      "BLEU": 32.380374722084525,
+      "chrF2": 30.384844655711102,
+      "TER": 49.310409975439256,
+      "exact_match_rate": 0.050359712230215826,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 0.9372444444444444,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "Zhuoqun": {
+      "samples": 703,
+      "BLEU": 10.560107321229836,
+      "chrF2": 10.98310054202614,
+      "TER": 91.74892256360351,
+      "exact_match_rate": 0.001422475106685633,
+      "empty_output_rate": 0.0,
+      "character_length_ratio": 1.0717540637965042,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
+    },
+    "unknown": {
+      "samples": 12906,
+      "BLEU": 12.861387188089601,
+      "chrF2": 13.547981744471688,
+      "TER": 78.81499768554235,
+      "exact_match_rate": 0.006121183945451728,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6960417107375285
+      "character_length_ratio": 0.9422104966725459,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     }
   },
   "by_length_bin": {
     "004_008": {
-      "samples": 1326,
-      "BLEU": 33.75734846323192,
-      "chrF2": 29.923056248729267,
-      "TER": 86.28858578607323,
-      "exact_match_rate": 0.14630467571644043,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.8196241382566815
+      "samples": 29612,
+      "BLEU": 15.800227037482845,
+      "chrF2": 15.308497380504566,
+      "TER": 76.3719293314718,
+      "exact_match_rate": 0.01472376063757936,
+      "empty_output_rate": 0.00027016074564365796,
+      "character_length_ratio": 0.9294510528432304,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "009_016": {
-      "samples": 16531,
-      "BLEU": 18.47308209896322,
-      "chrF2": 18.70713057005937,
-      "TER": 100.480836562252,
-      "exact_match_rate": 0.029096848345532635,
+      "samples": 23105,
+      "BLEU": 13.472340011569313,
+      "chrF2": 13.660536137778026,
+      "TER": 82.01923448653837,
+      "exact_match_rate": 0.0004760874269638606,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.7163221383373605
+      "character_length_ratio": 0.9809750373923358,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "017_032": {
-      "samples": 34591,
-      "BLEU": 11.382747289801152,
-      "chrF2": 13.351430361108669,
-      "TER": 113.17818330941152,
-      "exact_match_rate": 0.005290393454944928,
-      "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6985298702060786
+      "samples": 9642,
+      "BLEU": 10.597229992386678,
+      "chrF2": 11.753171184438896,
+      "TER": 85.23120014906445,
+      "exact_match_rate": 0.0,
+      "empty_output_rate": 0.00010371292263015972,
+      "character_length_ratio": 0.9915622408160732,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     },
     "033_plus": {
-      "samples": 8936,
-      "BLEU": 4.7159182063106675,
-      "chrF2": 7.239179763126493,
-      "TER": 136.46777442094663,
+      "samples": 1997,
+      "BLEU": 7.3139233977674705,
+      "chrF2": 9.211145132235528,
+      "TER": 95.05836715635215,
       "exact_match_rate": 0.0,
       "empty_output_rate": 0.0,
-      "character_length_ratio": 0.6668947584508006
+      "character_length_ratio": 1.0091243551633586,
+      "signatures": {
+        "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+        "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+        "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+      }
     }
   }
 }

experiment_metadata.json

@@ -1,183 +1,271 @@
 {
-  "step": 270000,
+  "model_family": "nllb",
+  "recipe_id": "nllb200-spm8k-directional-v3",
+  "mt_standardization": {
+    "id": "formosan-mt-standard-v3",
+    "sha256": "4bbded87eb4833b2d1cd5a88a1f2b059560416e0e69ecfa55433a77f418c903e",
+    "namespace": "formosan-mt"
+  },
+  "step": 220000,
   "best_metric": "chrF2",
-  "best_value": 22.942873146989022,
+  "best_value": 11.256583041753892,
   "direction": "f2zh",
+  "run_contract_sha256": "4cfe683caf4d25810d668f761921cb154a2d41e4e83822352faacc2d5b14293f",
   "validation": {
-    "mean_token_loss": 2.420696591592408,
-    "ppl": 11.253695815033172,
+    "mean_token_loss": 3.481121237640053,
+    "ppl": 32.49613754942568,
     "by_language": {
-      "ami": 2.5921270985999834,
-      "bnn": 2.1358773591268245,
-      "ckv": 1.9623805818177094,
-      "dru": 2.014338453068967,
-      "pwn": 2.489123042752967,
-      "pyu": 1.8734191935536901,
-      "ssf": 2.1260528645607564,
-      "sxr": 2.578650267016047,
-      "szy": 2.6909024219707973,
-      "tao": 2.8988929454858394,
-      "tay": 1.8934121017405425,
-      "trv": 3.2390189123141764,
-      "tsu": 2.578946331391228,
-      "xnb": 1.8004895268200158,
-      "xsy": 2.6114685328135936
+      "ami": 4.034636270919837,
+      "bnn": 3.5769809744689085,
+      "ckv": 2.955491952107692,
+      "dru": 3.4944524112298745,
+      "pwn": 3.7311584976757053,
+      "pyu": 2.6635534406284203,
+      "ssf": 3.8926185333821532,
+      "sxr": 2.5630961256169162,
+      "szy": 3.0292117092905326,
+      "tao": 3.5835890884774324,
+      "tay": 3.1955901890426928,
+      "trv": 3.134190788581884,
+      "tsu": 4.479148868581087,
+      "xnb": 1.7955884545588958,
+      "xsy": 4.265404218997709
     },
     "generation": {
       "samples": 1920,
       "bleu_tokenize": "zh",
       "global": {
-        "BLEU": 23.57852524101978,
-        "chrF2": 22.942873146989022,
-        "TER": 103.9412997903564,
-        "exact_match_rate": 0.020833333333333332,
+        "BLEU": 9.412108007110993,
+        "chrF2": 11.256583041753892,
+        "TER": 90.38131436978911,
+        "exact_match_rate": 0.010416666666666666,
         "empty_output_rate": 0.0,
-        "character_length_ratio": 0.7932179967578198
+        "character_length_ratio": 0.9647018635897825,
+        "signatures": {
+          "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+          "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+          "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+        }
       },
       "by_language": {
         "ami": {
           "samples": 128,
-          "BLEU": 19.55644041328223,
-          "chrF2": 18.577728465895298,
-          "TER": 104.81283422459893,
-          "exact_match_rate": 0.0078125,
+          "BLEU": 4.367336716716484,
+          "chrF2": 7.79001854487594,
+          "TER": 94.51874175927671,
+          "exact_match_rate": 0.0,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.7593659942363112
+          "character_length_ratio": 0.8488700564971752,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "bnn": {
           "samples": 128,
-          "BLEU": 25.380621452125194,
-          "chrF2": 23.79299268084961,
-          "TER": 103.28947368421053,
-          "exact_match_rate": 0.0078125,
+          "BLEU": 7.43278344662541,
+          "chrF2": 8.30436731110693,
+          "TER": 86.1777627576884,
+          "exact_match_rate": 0.0,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.7781674704077159
+          "character_length_ratio": 0.8770355599867066,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "ckv": {
           "samples": 128,
-          "BLEU": 27.113629785341512,
-          "chrF2": 25.146757668610668,
-          "TER": 103.82165605095541,
-          "exact_match_rate": 0.015625,
+          "BLEU": 13.913159689280391,
+          "chrF2": 14.232097621479383,
+          "TER": 76.30648330058939,
+          "exact_match_rate": 0.0,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.8045977011494253
+          "character_length_ratio": 0.9460856720827179,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "dru": {
           "samples": 128,
-          "BLEU": 26.22655487827295,
-          "chrF2": 25.694007229053373,
-          "TER": 105.22875816993465,
+          "BLEU": 7.257745876888657,
+          "chrF2": 11.379785118423582,
+          "TER": 100.21342186388429,
           "exact_match_rate": 0.0,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.7707253886010362
+          "character_length_ratio": 1.0559664201479113,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "pwn": {
           "samples": 128,
-          "BLEU": 21.141247720936498,
-          "chrF2": 19.70542420778234,
-          "TER": 101.65745856353593,
-          "exact_match_rate": 0.0078125,
+          "BLEU": 6.1170651870234245,
+          "chrF2": 7.937279691190215,
+          "TER": 94.11955815464587,
+          "exact_match_rate": 0.0,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.7858619264158744
+          "character_length_ratio": 0.921300516560316,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "pyu": {
           "samples": 128,
-          "BLEU": 28.010020387816798,
-          "chrF2": 26.26557564859005,
-          "TER": 113.88888888888889,
-          "exact_match_rate": 0.015625,
+          "BLEU": 10.846317334976424,
+          "chrF2": 15.366836715237858,
+          "TER": 78.68571428571428,
+          "exact_match_rate": 0.0078125,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.819854514334617
+          "character_length_ratio": 0.9635897435897436,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "ssf": {
           "samples": 128,
-          "BLEU": 28.980899971001712,
-          "chrF2": 27.362664735428773,
-          "TER": 99.32432432432432,
-          "exact_match_rate": 0.0234375,
+          "BLEU": 6.739578125938128,
+          "chrF2": 9.015721660009403,
+          "TER": 89.45256520492978,
+          "exact_match_rate": 0.0,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.8144778481012658
+          "character_length_ratio": 0.9204152249134948,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "sxr": {
           "samples": 128,
-          "BLEU": 21.355107092623804,
-          "chrF2": 22.270553009964367,
-          "TER": 97.97297297297297,
+          "BLEU": 13.977859965034813,
+          "chrF2": 15.815619449135317,
+          "TER": 70.94994087504926,
           "exact_match_rate": 0.015625,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.793171114599686
+          "character_length_ratio": 0.8838782412626832,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "szy": {
           "samples": 128,
-          "BLEU": 21.52598808800314,
-          "chrF2": 21.39925099180053,
-          "TER": 98.66666666666667,
-          "exact_match_rate": 0.03125,
+          "BLEU": 14.160740671900028,
+          "chrF2": 16.97155381389867,
+          "TER": 88.17733990147784,
+          "exact_match_rate": 0.046875,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.7829099307159353
+          "character_length_ratio": 1.0388719512195121,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "tao": {
           "samples": 128,
-          "BLEU": 22.207667963076098,
-          "chrF2": 23.056824267735124,
-          "TER": 109.39597315436242,
-          "exact_match_rate": 0.03125,
+          "BLEU": 6.998578356091424,
+          "chrF2": 16.04089062942359,
+          "TER": 90.4168922577152,
+          "exact_match_rate": 0.0,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.8163511880789368
+          "character_length_ratio": 0.9255461435577351,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "tay": {
           "samples": 128,
-          "BLEU": 28.009733813829975,
-          "chrF2": 26.444824918215147,
-          "TER": 102.29885057471265,
-          "exact_match_rate": 0.03125,
+          "BLEU": 11.001312009737747,
+          "chrF2": 10.895381601226708,
+          "TER": 84.46982055464926,
+          "exact_match_rate": 0.0,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.7627254509018037
+          "character_length_ratio": 0.940184537066497,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "trv": {
           "samples": 128,
-          "BLEU": 13.93836812194567,
-          "chrF2": 14.344449649929345,
-          "TER": 125.88235294117646,
-          "exact_match_rate": 0.015625,
+          "BLEU": 11.434671909068165,
+          "chrF2": 12.101809517498904,
+          "TER": 84.21642837118026,
+          "exact_match_rate": 0.0078125,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.7944753659039373
+          "character_length_ratio": 1.00375234521576,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "tsu": {
           "samples": 128,
-          "BLEU": 24.78776703981478,
-          "chrF2": 23.942995364265407,
-          "TER": 95.625,
-          "exact_match_rate": 0.03125,
+          "BLEU": 4.998406609073161,
+          "chrF2": 5.827140351620994,
+          "TER": 109.9028227672374,
+          "exact_match_rate": 0.0,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.7733105218135158
+          "character_length_ratio": 1.0106544901065448,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "xnb": {
           "samples": 128,
-          "BLEU": 33.44577807814948,
-          "chrF2": 32.87463813651187,
-          "TER": 92.90322580645162,
-          "exact_match_rate": 0.0625,
+          "BLEU": 25.945659871800135,
+          "chrF2": 24.697976362629582,
+          "TER": 55.62326869806095,
+          "exact_match_rate": 0.046875,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.8204047217537943
+          "character_length_ratio": 0.9163598106259864,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         },
         "xsy": {
           "samples": 128,
-          "BLEU": 22.10377012604304,
-          "chrF2": 22.747005676693703,
-          "TER": 103.18471337579618,
-          "exact_match_rate": 0.015625,
+          "BLEU": 6.906682364790633,
+          "chrF2": 8.045205239872994,
+          "TER": 111.64319248826291,
+          "exact_match_rate": 0.03125,
           "empty_output_rate": 0.0,
-          "character_length_ratio": 0.8155631986242476
+          "character_length_ratio": 1.1322869955156951,
+          "signatures": {
+            "BLEU": "nrefs:1|case:mixed|eff:yes|tok:zh|smooth:exp|version:2.5.1",
+            "chrF2": "nrefs:1|case:mixed|eff:yes|nc:6|nw:0|space:no|version:2.5.1",
+            "TER": "nrefs:1|case:mixed|tok:tercom|norm:yes|punct:yes|asian:yes|version:2.5.1"
+          }
         }
       }
     },
-    "step": 270000,
+    "step": 220000,
     "selection": {
       "metric": "chrF2",
-      "value": 22.942873146989022,
+      "value": 11.256583041753892,
       "improved": true,
-      "best_value_before_eval": 22.794952357181167
+      "best_value_before_eval": 11.132391003032263
     }
   }
 }

formosan_mt_inference.py

@@ -0,0 +1,79 @@
+#!/usr/bin/env python3
+"""Apply the corpus MT standardization profile to Formosan inference input."""
+
+from __future__ import annotations
+
+import argparse
+import sys
+from pathlib import Path
+
+SCRIPT_DIR = Path(__file__).resolve().parent
+PROJECT_ROOT = SCRIPT_DIR.parents[1]
+LOCAL_SCRIPTS = PROJECT_ROOT / "scripts" / "local"
+if LOCAL_SCRIPTS.is_dir():
+    sys.path.insert(0, str(LOCAL_SCRIPTS))
+
+from mt_standardization import (  # noqa: E402
+    StandardizationContext,
+    assert_idempotent,
+    load_profile,
+    standardize_text,
+)
+
+PACKAGED_PROFILE = SCRIPT_DIR / "mt_standardization_profile.json"
+PROJECT_PROFILE = PROJECT_ROOT / "config" / "mt_standardization.json"
+
+
+def default_profile_path() -> Path:
+    if PACKAGED_PROFILE.is_file():
+        return PACKAGED_PROFILE
+    return PROJECT_PROFILE
+
+
+def normalize_formosan(
+    text: str,
+    lang_code: str,
+    *,
+    row_type: str = "sentence",
+    source_repository: str = "inference",
+    source_path: str = "interactive",
+    profile_path: Path | None = None,
+) -> str:
+    """Return model-ready Formosan text or reject unresolved notation."""
+    path = (profile_path or default_profile_path()).expanduser().resolve()
+    profile = load_profile(path)
+    context = StandardizationContext(
+        language=lang_code.strip().lower(),
+        row_type=row_type,
+        repository=source_repository,
+        xml_path=source_path,
+    )
+    result = standardize_text(text, context=context, profile=profile)
+    if result.status != "accepted":
+        raise ValueError(
+            "Formosan input is not safe to normalize automatically: "
+            f"{result.reason or result.status}"
+        )
+    assert_idempotent(result, context=context, profile=profile)
+    return result.text
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("text")
+    parser.add_argument("--lang-code", required=True)
+    parser.add_argument("--row-type", default="sentence")
+    parser.add_argument("--profile", type=Path)
+    args = parser.parse_args()
+    print(
+        normalize_formosan(
+            args.text,
+            args.lang_code,
+            row_type=args.row_type,
+            profile_path=args.profile,
+        )
+    )
+
+
+if __name__ == "__main__":
+    main()

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:14395539c53293e66e69aef90ba1247992b9e5b89a083890d7cae56873a9a0e7
-size 2475092320
+oid sha256:4a4db8ac7fb122b76471ebe901e1ea16f17d0d43a9ad0390c32d37e357acbecb
+size 2482710880

mt_standardization.py

@@ -0,0 +1,458 @@
+"""Deterministic, source-aware Formosan standardization for MT corpora."""
+
+from __future__ import annotations
+
+import hashlib
+import html
+import json
+import re
+import unicodedata
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any, Iterable
+
+PROJECT_ROOT = Path(__file__).resolve().parents[2]
+DEFAULT_PROFILE_PATH = PROJECT_ROOT / "config" / "mt_standardization.json"
+
+CONTROL_RE = re.compile(r"[\x00-\x08\x0B\x0C\x0E-\x1F\x7F]")
+FORMAT_CHARACTERS = frozenset("\u200b\u200c\u200d\ufeff")
+WHITESPACE_RE = re.compile(r"\s+")
+WIKI_HEADING_RE = re.compile(r"(^|\s)==+\s*([^=\n]+?)\s*==+(?=\s|$)")
+ANGLE_CONTENT_RE = re.compile(r"<([^<>\s]{1,24})>")
+BRACE_CONTENT_RE = re.compile(r"\{([^{}\s]{1,24})}")
+OPTIONAL_CONTENT_RE = re.compile(r"\(([^()\s]{1,16})\)")
+SIMPLE_ALTERNATIVE_RE = re.compile(r"(?<!\S)([^\s/]+)\s*/\s*([^\s/]+)(?!\S)")
+SPACED_TILDE_RE = re.compile(r"(?<!\S)([^\s~]+)\s+~\s+([^\s~]+)(?!\S)")
+SPEAKER_PREFIX_RE = re.compile(r"^([A-Z][A-Za-z0-9_. -]{0,31}):\s+")
+URL_RE = re.compile(r"(?:https?://|www\.)", re.IGNORECASE)
+SUPERSCRIPT_FOOTNOTE_RE = re.compile(r"(?<=\D)[⁰¹²³⁴⁵⁶⁷⁸⁹]+(?=\s|$)")
+EXERCISE_BLANK_RE = re.compile(r"_{2,}")
+
+
+@dataclass(frozen=True)
+class StandardizationContext:
+    language: str
+    row_type: str
+    repository: str
+    xml_path: str
+
+
+@dataclass(frozen=True)
+class StandardizationResult:
+    text: str
+    status: str
+    confidence: str
+    eval_eligible: bool
+    transformations: tuple[dict[str, Any], ...]
+    unresolved_markers: tuple[str, ...]
+    reason: str = ""
+    speaker_label: str = ""
+
+
+def load_profile(path: Path = DEFAULT_PROFILE_PATH) -> dict[str, Any]:
+    try:
+        profile = json.loads(path.read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError) as exc:
+        raise SystemExit(f"Cannot load MT standardization profile {path}: {exc}") from exc
+    if profile.get("schema_version") != 1:
+        raise SystemExit(f"Unsupported MT standardization profile schema: {profile.get('schema_version')}")
+    if not str(profile.get("profile_id") or "").strip():
+        raise SystemExit(f"MT standardization profile has no profile_id: {path}")
+    if profile.get("unicode_normalization") != "NFC":
+        raise SystemExit("MT standardization currently requires NFC")
+    expected_implementation = str(
+        profile.get("implementation_sha256") or ""
+    )
+    actual_implementation = hashlib.sha256(
+        Path(__file__).read_bytes()
+    ).hexdigest()
+    if expected_implementation != actual_implementation:
+        raise SystemExit(
+            "MT standardization implementation hash mismatch: "
+            f"profile={expected_implementation or '<missing>'}, "
+            f"loaded={actual_implementation}"
+        )
+    if not isinstance(profile.get("policy"), dict):
+        raise SystemExit(f"MT standardization profile has no policy mapping: {path}")
+    if not isinstance(profile.get("source_overrides", []), list):
+        raise SystemExit(f"MT standardization source_overrides must be a list: {path}")
+    return profile
+
+
+def profile_sha256(path: Path = DEFAULT_PROFILE_PATH) -> str:
+    return hashlib.sha256(path.read_bytes()).hexdigest()
+
+
+def is_orthographic_character(character: str) -> bool:
+    if not character:
+        return False
+    return unicodedata.category(character)[0] in {"L", "M", "N"} or character in {"'", "’", "ʼ"}
+
+
+def insertion_content_is_safe(value: str) -> bool:
+    return bool(value) and all(is_orthographic_character(character) for character in value)
+
+
+def source_matches(selector: dict[str, Any], context: StandardizationContext) -> bool:
+    for key, actual in (
+        ("language", context.language),
+        ("row_type", context.row_type),
+        ("repository", context.repository),
+    ):
+        expected = str(selector.get(key) or "*")
+        if expected not in {"*", actual}:
+            return False
+    path_pattern = str(selector.get("path_regex") or "")
+    return not path_pattern or re.search(path_pattern, context.xml_path) is not None
+
+
+def effective_policy(profile: dict[str, Any], context: StandardizationContext) -> tuple[dict[str, Any], bool]:
+    policy = dict(profile["policy"])
+    reviewed_ambiguous = False
+    for override in profile.get("source_overrides", []):
+        if not isinstance(override, dict) or not source_matches(override, context):
+            continue
+        values = override.get("policy", {})
+        if not isinstance(values, dict):
+            raise ValueError(f"Source override policy must be a mapping: {override}")
+        policy.update(values)
+        reviewed_ambiguous = reviewed_ambiguous or bool(override.get("reviewed_ambiguous"))
+    return policy, reviewed_ambiguous
+
+
+class _StandardizationRun:
+    def __init__(self, text: str) -> None:
+        self.text = text
+        self.transformations: list[dict[str, Any]] = []
+        self.ambiguous = False
+        self.reason = ""
+        self.speaker_label = ""
+
+    def replace(self, rule: str, new_text: str, *, count: int = 1, ambiguous: bool = False) -> None:
+        if new_text == self.text:
+            return
+        self.text = new_text
+        self.transformations.append({"rule": rule, "count": int(max(count, 1))})
+        self.ambiguous = self.ambiguous or ambiguous
+
+    def regex_sub(
+        self,
+        rule: str,
+        pattern: re.Pattern[str],
+        replacement: str | Any,
+        *,
+        ambiguous: bool = False,
+    ) -> None:
+        new_text, count = pattern.subn(replacement, self.text)
+        if count:
+            self.replace(rule, new_text, count=count, ambiguous=ambiguous)
+
+
+def unwrap_infixes(run: _StandardizationRun, pattern: re.Pattern[str], rule: str) -> None:
+    count = 0
+
+    def replacement(match: re.Match[str]) -> str:
+        nonlocal count
+        content = match.group(1)
+        if not insertion_content_is_safe(content):
+            return match.group(0)
+        count += 1
+        return content
+
+    new_text = pattern.sub(replacement, run.text)
+    if count:
+        run.replace(rule, new_text, count=count)
+
+
+def include_optional_segments(run: _StandardizationRun) -> None:
+    original = run.text
+    working = original
+    total = 0
+    while True:
+        count = 0
+
+        def replacement(match: re.Match[str]) -> str:
+            nonlocal count
+            start, end = match.span()
+            previous = working[start - 1] if start else ""
+            following = working[end] if end < len(working) else ""
+            content = match.group(1)
+            previous_is_boundary = (
+                is_orthographic_character(previous)
+                or previous in {"-", "=", "+", "~"}
+            )
+            following_is_boundary = (
+                is_orthographic_character(following)
+                or following in {"-", "=", "+", "~"}
+            )
+            if not (
+                insertion_content_is_safe(content)
+                and previous_is_boundary
+                and following_is_boundary
+            ):
+                return match.group(0)
+            count += 1
+            return content
+
+        updated = OPTIONAL_CONTENT_RE.sub(replacement, working)
+        if not count:
+            break
+        total += count
+        working = updated
+    if total:
+        run.replace(
+            "include_optional_intraword_segment",
+            working,
+            count=total,
+            ambiguous=True,
+        )
+
+
+def remove_boundary_character(run: _StandardizationRun, marker: str, rule: str) -> None:
+    output: list[str] = []
+    removed = 0
+    index = 0
+    while index < len(run.text):
+        character = run.text[index]
+        if character != marker:
+            output.append(character)
+            index += 1
+            continue
+        end = index + 1
+        while end < len(run.text) and run.text[end] == marker:
+            end += 1
+        previous = run.text[index - 1] if index else ""
+        following = run.text[end] if end < len(run.text) else ""
+        if (is_orthographic_character(previous) or is_orthographic_character(following)) and not (
+            previous.isdigit() and following.isdigit()
+        ):
+            removed += end - index
+            index = end
+            continue
+        output.append(run.text[index:end])
+        index = end
+    if removed:
+        run.replace(rule, "".join(output), count=removed)
+
+
+def normalize_morphological_boundaries(
+    run: _StandardizationRun,
+    policy: dict[str, Any],
+) -> None:
+    while True:
+        before = run.text
+        if policy.get("remove_clitic_boundaries") and "=" in run.text:
+            count = run.text.count("=")
+            run.replace(
+                "remove_clitic_boundary",
+                run.text.replace("=", ""),
+                count=count,
+            )
+        if policy.get("remove_intraword_hyphens"):
+            remove_boundary_character(
+                run,
+                "-",
+                "remove_hyphen_boundary",
+            )
+        if policy.get("remove_intraword_tildes"):
+            remove_boundary_character(
+                run,
+                "~",
+                "remove_tilde_boundary",
+            )
+        if policy.get("remove_intraword_pluses"):
+            remove_boundary_character(
+                run,
+                "+",
+                "remove_plus_boundary",
+            )
+        if policy.get("include_optional_intraword_segments"):
+            include_optional_segments(run)
+        if run.text == before:
+            return
+        if len(run.text) >= len(before):
+            raise ValueError(
+                "Morphological normalization did not strictly reduce text"
+            )
+
+
+def select_simple_alternatives(run: _StandardizationRun) -> None:
+    if URL_RE.search(run.text):
+        run.reason = "url_in_formosan_standard"
+        return
+    slash_count = run.text.count("/")
+    if not slash_count:
+        return
+    if slash_count > 4:
+        run.reason = "complex_slash_notation"
+        return
+    working = run.text
+    total = 0
+    while True:
+        working, count = SIMPLE_ALTERNATIVE_RE.subn(lambda match: match.group(1), working)
+        total += count
+        if not count:
+            break
+    if total:
+        run.replace("select_first_slash_alternative", working, count=total, ambiguous=True)
+    if "/" in run.text:
+        run.reason = "unresolved_slash_notation"
+
+
+def select_tilde_alternatives(run: _StandardizationRun) -> None:
+    working = run.text
+    total = 0
+    while True:
+        working, count = SPACED_TILDE_RE.subn(lambda match: match.group(1), working)
+        total += count
+        if not count:
+            break
+    if total:
+        run.replace("select_first_tilde_alternative", working, count=total, ambiguous=True)
+
+
+def unresolved_markers(text: str, configured: Iterable[str]) -> tuple[str, ...]:
+    markers = {marker for marker in configured if marker and marker in text}
+    if EXERCISE_BLANK_RE.search(text):
+        markers.add("exercise_blank")
+    return tuple(sorted(markers))
+
+
+def standardize_text(
+    value: object,
+    *,
+    context: StandardizationContext,
+    profile: dict[str, Any],
+    contains_unclear: bool = False,
+) -> StandardizationResult:
+    raw = "" if value is None else str(value)
+    if contains_unclear:
+        return StandardizationResult(
+            text="",
+            status="ineligible",
+            confidence="none",
+            eval_eligible=False,
+            transformations=(),
+            unresolved_markers=(),
+            reason="contains_unclear",
+        )
+    if not raw.strip():
+        return StandardizationResult(
+            text="",
+            status="ineligible",
+            confidence="none",
+            eval_eligible=False,
+            transformations=(),
+            unresolved_markers=(),
+            reason="empty_source_standard",
+        )
+
+    policy, reviewed_ambiguous = effective_policy(profile, context)
+    run = _StandardizationRun(raw)
+
+    controls = CONTROL_RE.sub(" ", run.text)
+    controls = "".join("" if character in FORMAT_CHARACTERS else character for character in controls)
+    run.replace("remove_control_and_format_characters", controls)
+    run.replace("html_unescape", html.unescape(run.text))
+    run.replace("unicode_nfc", unicodedata.normalize("NFC", run.text))
+    run.replace(
+        "normalize_whitespace",
+        WHITESPACE_RE.sub(" ", run.text).strip(),
+    )
+
+    if policy.get("strip_wiki_heading_markup"):
+        run.regex_sub("strip_wiki_heading_markup", WIKI_HEADING_RE, lambda match: f"{match.group(1)}{match.group(2)}")
+
+    if policy.get("remove_null_morphemes"):
+        for marker in ("Ø-", "∅-", "-Ø", "-∅", "Ø", "∅"):
+            count = run.text.count(marker)
+            if count:
+                run.replace("remove_null_morpheme", run.text.replace(marker, ""), count=count)
+
+    if policy.get("unwrap_letter_infixes"):
+        unwrap_infixes(run, ANGLE_CONTENT_RE, "unwrap_angle_infix")
+        unwrap_infixes(run, BRACE_CONTENT_RE, "unwrap_braced_morpheme")
+
+    # Remove annotation characters before boundary and alternative handling so
+    # cleanup cannot expose notation that is processed only on a later pass.
+    if "_" in run.text and not EXERCISE_BLANK_RE.search(run.text):
+        count = run.text.count("_")
+        run.replace("remove_annotation_underscore", run.text.replace("_", ""), count=count, ambiguous=True)
+    if "\\" in run.text:
+        count = run.text.count("\\")
+        run.replace("remove_escape_backslash", run.text.replace("\\", ""), count=count)
+
+    normalize_morphological_boundaries(run, policy)
+
+    if policy.get("select_first_simple_alternative"):
+        select_simple_alternatives(run)
+        select_tilde_alternatives(run)
+
+    if policy.get("strip_trailing_superscript_footnotes"):
+        run.regex_sub("strip_superscript_footnote", SUPERSCRIPT_FOOTNOTE_RE, "")
+
+    squeezed = WHITESPACE_RE.sub(" ", run.text).strip()
+    run.replace("normalize_whitespace", squeezed)
+
+    speaker = SPEAKER_PREFIX_RE.match(run.text)
+    if speaker:
+        run.speaker_label = speaker.group(1)
+
+    markers = unresolved_markers(run.text, profile.get("unresolved_annotation_characters", []))
+    reason = run.reason
+    if markers and not reason:
+        reason = "unresolved_annotation_markers"
+    if not run.text or not any(unicodedata.category(character)[0] in {"L", "M", "N"} for character in run.text):
+        return StandardizationResult(
+            text=run.text,
+            status="ineligible",
+            confidence="none",
+            eval_eligible=False,
+            transformations=tuple(run.transformations),
+            unresolved_markers=markers,
+            reason="empty_or_nonlinguistic_after_standardization",
+            speaker_label=run.speaker_label,
+        )
+    if reason or markers:
+        return StandardizationResult(
+            text=run.text,
+            status="quarantine",
+            confidence="ambiguous",
+            eval_eligible=False,
+            transformations=tuple(run.transformations),
+            unresolved_markers=markers,
+            reason=reason,
+            speaker_label=run.speaker_label,
+        )
+
+    confidence = "ambiguous" if run.ambiguous else ("safe" if run.transformations else "unchanged")
+    return StandardizationResult(
+        text=run.text,
+        status="accepted",
+        confidence=confidence,
+        eval_eligible=(context.row_type == "sentence" and (not run.ambiguous or reviewed_ambiguous)),
+        transformations=tuple(run.transformations),
+        unresolved_markers=(),
+        speaker_label=run.speaker_label,
+    )
+
+
+def assert_idempotent(
+    result: StandardizationResult,
+    *,
+    context: StandardizationContext,
+    profile: dict[str, Any],
+) -> None:
+    if result.status != "accepted":
+        return
+    repeated = standardize_text(result.text, context=context, profile=profile)
+    if (
+        repeated.text != result.text
+        or repeated.status != "accepted"
+        or repeated.transformations
+        or repeated.speaker_label != result.speaker_label
+    ):
+        raise ValueError(
+            f"MT standardization is not idempotent for {context.repository}/{context.xml_path}: "
+            f"{result.text!r} -> {repeated.text!r}"
+        )

mt_standardization_profile.json

@@ -0,0 +1,32 @@
+{
+  "schema_version": 1,
+  "profile_id": "formosan-mt-standard-v3",
+  "implementation_sha256": "c78f092f826aed70468bb3d4276eb27bab482dc5dc36dedf9ed5726596e0dd31",
+  "unicode_normalization": "NFC",
+  "policy": {
+    "unwrap_letter_infixes": true,
+    "remove_clitic_boundaries": true,
+    "remove_null_morphemes": true,
+    "remove_intraword_hyphens": true,
+    "remove_intraword_tildes": true,
+    "remove_intraword_pluses": true,
+    "include_optional_intraword_segments": true,
+    "select_first_simple_alternative": true,
+    "strip_wiki_heading_markup": true,
+    "strip_trailing_superscript_footnotes": true
+  },
+  "source_overrides": [],
+  "unresolved_annotation_characters": [
+    "<",
+    ">",
+    "=",
+    "Ø",
+    "∅",
+    "*",
+    "_",
+    "{",
+    "}",
+    "|",
+    "\\"
+  ]
+}

publication.json

@@ -0,0 +1,30 @@
+{
+  "schema_version": 1,
+  "repo_id": "FormosanBank/nllb200-formosan-zh-spm8k",
+  "direction": "f2zh",
+  "run_stamp": "20260809-210523",
+  "checkpoint": "best",
+  "checkpoint_step": 220000,
+  "corpus": {
+    "name": "private_no_bible",
+    "target_lang": "chinese",
+    "rows": 788326,
+    "splits": {
+      "test": 64356,
+      "train": 702520,
+      "validate": 21450
+    },
+    "sha256": "238b96ea87716a749391d7157ce99f1611c1d4a489c4c108c315553505ca1f29"
+  },
+  "profile_sha256": "34f45832bdedc1b8322b26936dd88667dfecb360d941a7f54f60ae31a1fb4e10",
+  "base_model": {
+    "name": "facebook/nllb-200-distilled-600M",
+    "revision": "f8d333a098d19b4fd9a8b18f94170487ad3f821d"
+  },
+  "mt_standardization": {
+    "id": "formosan-mt-standard-v3",
+    "sha256": "4bbded87eb4833b2d1cd5a88a1f2b059560416e0e69ecfa55433a77f418c903e",
+    "namespace": "formosan-mt"
+  },
+  "run_contract_sha256": "4cfe683caf4d25810d668f761921cb154a2d41e4e83822352faacc2d5b14293f"
+}

sentencepiece.bpe.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0571a653d72d4d5a892f302084363ae86de93f9da19521a9bd3498b60819b960
-size 4898297
+oid sha256:d8ffca4dcbcf1655c9d4c9c4f3cd0ed2eaecf37fd689618fb9702a899021bc03
+size 4948342

special_tokens_map.json

@@ -217,8 +217,6 @@
     "tsu_Latn",
     "xnb_Latn",
     "xsy_Latn",
-    "<mask>",
-    "<dae>",
     "<dialect_central>",
     "<dialect_coastal>",
     "<dialect_dawu>",
@@ -265,81 +263,209 @@
     "<dom_culture>",
     "<dom_dictionary>",
     "<dom_essays>",
-    "<dom_ethnography_journal_ritual_song>",
     "<dom_formosan_100_paiwan_texts>",
     "<dom_formosan_academia_sinica_oral_legends>",
+    "<dom_formosan_adlay_kun_long_liu_syntactic_interactio>",
+    "<dom_formosan_adlay_liu_lfg_relativisation_jianshi_sq>",
     "<dom_formosan_amis_adversative_constructions>",
+    "<dom_formosan_amis_huang_english_phonology_grammar>",
     "<dom_formosan_amis_kavalan_lin_interrogative_verbs>",
     "<dom_formosan_amis_kuo_sung_comparative_constructions>",
-    "<dom_formosan_amis_lifok_dongi_saopo_kimakimad>",
     "<dom_formosan_amis_myths_and_customs>",
     "<dom_formosan_amis_pa_verbs>",
     "<dom_formosan_amis_serial_verb_constructions>",
     "<dom_formosan_amis_tung_chiou>",
+    "<dom_formosan_amy_lee_body_part_nomenclature_seediq>",
+    "<dom_formosan_anna_hsiou_chuan_chang_reference_gramma>",
+    "<dom_formosan_anton_quack_puyuma_pulingaw>",
+    "<dom_formosan_apay_tang_anger_happiness_truku_seediq>",
     "<dom_formosan_asai_sedik_language>",
+    "<dom_formosan_atayal_you_nao_savak_raxayal>",
+    "<dom_formosan_bien_daniel_chiang_house_social_hierarc>",
+    "<dom_formosan_boyizhenu_narrative_oral_literature_tap>",
     "<dom_formosan_bunun_debusser_dissertation>",
-    "<dom_formosan_bunun_moriguchi_northern_texts>",
     "<dom_formosan_bunun_topic_focus>",
+    "<dom_formosan_c_douglas_chretien_oceanic_material_in>",
+    "<dom_formosan_catherine_tseng_seediq_atayal_foods_the>",
+    "<dom_formosan_chang_kwo_tan_syncretic_objects_materia>",
+    "<dom_formosan_chaolin_li_motion_verb_grammaticalizati>",
     "<dom_formosan_chen_fukuda_relabeling_ergative>",
     "<dom_formosan_chen_fukuda_three_ways_steal>",
+    "<dom_formosan_cheng_sung_modality_kanakanavu>",
+    "<dom_formosan_chi_lu_chen_and_micheal_d_coe_an_invest>",
+    "<dom_formosan_chuming_wu_linking_constructions_mayrin>",
+    "<dom_formosan_chun_mei_chen_comparative_phonology_pai>",
+    "<dom_formosan_chunming_wu_two_types_of_noun_incorpora>",
     "<dom_formosan_cip_atayal_grammar_overview>",
+    "<dom_formosan_claire_mcgill_a_brief_tayal_vocabulary>",
+    "<dom_formosan_david_blundell_ed_austronesian_taiwan>",
     "<dom_formosan_dean_johnson_year_clouds_smangus>",
+    "<dom_formosan_der_hwa_victoria_rau_grammar_atayal>",
+    "<dom_formosan_der_hwa_victoria_rau_lexical_similarity>",
+    "<dom_formosan_dong_manv_yami_ode_to_taro>",
+    "<dom_formosan_dong_yi_lin_universal_quantifier_kavala>",
+    "<dom_formosan_dong_yijia_tjuwabar_paiwan_ritual_langu>",
     "<dom_formosan_dorinda_tsai_hsiu_liu_neutral_imperfect>",
+    "<dom_formosan_ed_torjesen_amis_taiwan_word_list>",
     "<dom_formosan_egerod_origin_headhunting_atayal_text_v>",
     "<dom_formosan_egerod_soren_word_order_word_classes_at>",
     "<dom_formosan_elizabeth_zeitoun_chen_huei_wu_overview>",
+    "<dom_formosan_elizabeth_zeitoun_lillian_huang_anna_ch>",
+    "<dom_formosan_elizabeth_zeitoun_lillian_huang_marie_y>",
+    "<dom_formosan_elizabeth_zeitoun_pronominal_system_man>",
+    "<dom_formosan_elizabeth_zeitoun_squib_dynamic_vs_stat>",
     "<dom_formosan_epark>",
     "<dom_formosan_ferrell_taiwan_aboriginal_groups_proble>",
+    "<dom_formosan_george_leroy_shelley_vudai_rukai>",
     "<dom_formosan_gitbook_translations>",
     "<dom_formosan_glosbe>",
-    "<dom_formosan_hala_saku_la_videos>",
+    "<dom_formosan_gujing_lin_tsou_serial_verb_constructio>",
+    "<dom_formosan_henry_chang_nominal_aspect_tsou>",
+    "<dom_formosan_henry_chang_tsou_causative_applicative>",
+    "<dom_formosan_henry_y_chang_formosan_speech_act_mood>",
+    "<dom_formosan_ho_dan_phonological_system_butonglu_pai>",
     "<dom_formosan_holmer_seediq>",
+    "<dom_formosan_hsiu_chuan_liao_transitivity_ergativity>",
+    "<dom_formosan_huang_dongqiu_atayal_bunun_amis_languag>",
     "<dom_formosan_huang_grammaticalization_squliq_atayal>",
+    "<dom_formosan_huang_hui_chuan_glide_formation_takitud>",
     "<dom_formosan_huang_mei_chin_atayal_reference_grammar>",
+    "<dom_formosan_hui_chuan_j_huang_syllable_types_in_bun>",
+    "<dom_formosan_hui_shan_lin_squliq_atayal_reduplicatio>",
     "<dom_formosan_huteson_rukai_survey>",
+    "<dom_formosan_ian_maddieson_and_richard_wright_the_vo>",
     "<dom_formosan_ilrdf_42_language_practice_word_lists>",
     "<dom_formosan_ilrdf_tousvusvutu_kita>",
+    "<dom_formosan_indigenous_education_resource_center_ke>",
     "<dom_formosan_indigenous_language_literary_awards>",
+    "<dom_formosan_isabelle_bril_roots_and_stems_lexical_a>",
+    "<dom_formosan_jian_iwan_guo_qingliu_life_history>",
+    "<dom_formosan_jian_shilang_introduction_thao_grammar>",
+    "<dom_formosan_john_tu_er_wei_notes_on_the_religious_b>",
+    "<dom_formosan_john_u_wolff_the_proto_austronesian_pho>",
+    "<dom_formosan_jonathan_kuo_amis_manner_result_verbs>",
+    "<dom_formosan_josiane_cauquelin_aborigines_taiwan_puy>",
+    "<dom_formosan_joy_wu_clausal_modifiers_in_amis>",
+    "<dom_formosan_joy_wu_verb_classification_case_marking>",
     "<dom_formosan_kanakanavu_texts>",
     "<dom_formosan_kavalan_zhang_reference_grammar>",
-    "<dom_formosan_li_peirong_xu_weisheng_introduction_tru>",
-    "<dom_formosan_li_peizhen_wu_xingyu_shouhu_xing>",
+    "<dom_formosan_kumu_tapas_tribal_memory_oral_history_w>",
+    "<dom_formosan_kuo_j_c_argument_alternation_and_argume>",
+    "<dom_formosan_lei_shih_family_system_paiwan_at_su_pai>",
+    "<dom_formosan_li_chen_yeh_developing_a_hla_alua_learn>",
+    "<dom_formosan_li_may_sung_budai_rukai_exclamatives>",
+    "<dom_formosan_li_may_sung_clausal_nominalization_buda>",
+    "<dom_formosan_lillian_huang_atayal_participants_cross>",
+    "<dom_formosan_lillian_huang_ergativity_atayal>",
+    "<dom_formosan_liu_manyi_kaskun_ata_matas_i_ki_quaz>",
+    "<dom_formosan_malcom_d_ross_the_history_and_transitiv>",
+    "<dom_formosan_malcom_ross_proto_austronesian_verbal_m>",
+    "<dom_formosan_man_ni_chu_the_measurement_of_final_i_v>",
+    "<dom_formosan_matsuzawa_kazuko_the_social_and_ritual>",
+    "<dom_formosan_maya_yeh_blaq_uv_construction_atayal>",
+    "<dom_formosan_mayumi_oiwa_adverbial_verb_construction>",
+    "<dom_formosan_mayumi_oiwa_bungard_morphology_and_synt>",
+    "<dom_formosan_mayumi_oiwa_possessor_raising_truku_see>",
+    "<dom_formosan_mei_chin_huang_verb_classification_in_m>",
+    "<dom_formosan_melissa_shih_hui_lin_sakizaya_or_amis_a>",
+    "<dom_formosan_nanang_tadaw_naci_mowna_pgagu>",
     "<dom_formosan_naomi_tsukida_correlative_clauses_seedi>",
+    "<dom_formosan_naomi_tsukida_ditransitive_causative_se>",
     "<dom_formosan_nowbucyang_truku_thesis>",
+    "<dom_formosan_ochiai_izumi_2016_comparative_linguisti>",
+    "<dom_formosan_ochiai_izumi_bu_hwan_vocabulary_recorde>",
+    "<dom_formosan_ochiai_izumi_numerals_paran_seediq_rela>",
+    "<dom_formosan_ochiai_izumi_on_a_seediq_glossary_recor>",
     "<dom_formosan_old_texts>",
-    "<dom_formosan_paiwan_collart_zeitoun_time_reference>",
     "<dom_formosan_paiwan_ho_five_dialects>",
     "<dom_formosan_paiwanstories>",
+    "<dom_formosan_pang_hsin_ting_reconstruction_proto_puy>",
+    "<dom_formosan_patricia_stanley_morphophonemics_of_ver>",
+    "<dom_formosan_paul_jen_kuei_li_comparative_bunun_dial>",
+    "<dom_formosan_paul_jen_kuei_li_conjunction_thao>",
+    "<dom_formosan_paul_jen_kuei_li_internal_relationships>",
+    "<dom_formosan_paul_jen_kuei_li_origins_of_the_east_fo>",
+    "<dom_formosan_paul_jen_kuei_li_review_of_taiwan_abori>",
+    "<dom_formosan_paul_jen_kuei_li_rukai_structure>",
+    "<dom_formosan_paul_jen_kuei_li_some_plant_names_in_fo>",
+    "<dom_formosan_paul_jen_kuei_li_the_preglottalised_sto>",
+    "<dom_formosan_paul_li_atayalic_final_voiced_stops>",
+    "<dom_formosan_paul_li_case_marking_four_formosan_lang>",
+    "<dom_formosan_paul_li_position_atayal_austronesian>",
+    "<dom_formosan_pingdong_county_government_kai_na_sepuc>",
     "<dom_formosan_puyuma_chen_raising_to_object>",
     "<dom_formosan_puyuma_katipol_kopfjagdriten>",
     "<dom_formosan_puyuma_teng_pronominal_systems>",
     "<dom_formosan_puyuma_teng_reference_grammar>",
+    "<dom_formosan_raleigh_ferrell_construction_markers_fo>",
+    "<dom_formosan_raleigh_ferrell_paiwan_phonology_and_pr>",
+    "<dom_formosan_raleigh_j_ferrell_intent_and_volition_i>",
     "<dom_formosan_rik_bunun>",
+    "<dom_formosan_rik_de_busser_the_influence_of_christia>",
+    "<dom_formosan_robert_blust_a_note_on_covert_structure>",
     "<dom_formosan_robert_blust_austronesian_homeland_ling>",
+    "<dom_formosan_robert_blust_some_remarks_linguistic_po>",
+    "<dom_formosan_robert_blust_thao_triplication>",
+    "<dom_formosan_robert_blust_the_austronesian_languages>",
     "<dom_formosan_rukai_mantauran_stories>",
     "<dom_formosan_rukai_texts>",
     "<dom_formosan_rukai_zeitoun_saying>",
-    "<dom_formosan_saaroa_pan_grammar>",
     "<dom_formosan_sakizaya_affixes>",
+    "<dom_formosan_schetelig_mittheilungen_uber_die_sprach>",
     "<dom_formosan_seals>",
     "<dom_formosan_seediq_zhang_reference_grammar>",
+    "<dom_formosan_shigeru_tsuchida_and_isidore_dyen_proto>",
     "<dom_formosan_shigeru_tsuchida_kanakanavu_texts>",
+    "<dom_formosan_shigeru_tsuchida_preliminary_reports_on>",
     "<dom_formosan_shigeru_tsuchida_reconstruction_proto_t>",
+    "<dom_formosan_shih_chi_stella_yeh_stress_and_quantity>",
     "<dom_formosan_shih_rukai_adverbial>",
-    "<dom_formosan_song_limei_introduction_kanakanavu_gram>",
-    "<dom_formosan_song_limei_introduction_seediq_grammar>",
+    "<dom_formosan_sihwei_chen_and_haowen_jiang_ways_of_ta>",
+    "<dom_formosan_sihwei_chen_lexical_aspect_atayal>",
+    "<dom_formosan_soren_egerod_statement_atayal_phonology>",
+    "<dom_formosan_soren_egerod_verb_inflection_atayal>",
+    "<dom_formosan_stacy_teng_malcom_ross_is_puyuma_primar>",
+    "<dom_formosan_stacy_wan_tin_huang_functions_yami_verb>",
     "<dom_formosan_sung_kuo_descriptive_comparative_constr>",
     "<dom_formosan_taoshan_elementary_school_atelier_hui_k>",
+    "<dom_formosan_teresa_m_chen_verbal_construction_and_v>",
+    "<dom_formosan_ting_chi_wei_parallelism_in_amis_sluici>",
+    "<dom_formosan_tingchun_chen_successive_cyclic_case_as>",
+    "<dom_formosan_truku_lowking_demonstratives>",
+    "<dom_formosan_tsai_hsui_liu_complementation_three_lan>",
+    "<dom_formosan_tsai_wei_tien_dylan_conjunctive_reducti>",
+    "<dom_formosan_tsou_descriptive_study>",
     "<dom_formosan_tsukida_naomi_verb_classification_amis>",
+    "<dom_formosan_tu_wen_chiu_a_synchronic_classification>",
+    "<dom_formosan_tung_tsuchida_ting_li_pan_saaroa_texts>",
+    "<dom_formosan_victoria_rau_yi_hsin_wu_and_meng_chien>",
+    "<dom_formosan_wei_huilin_liu_social_structure_yami>",
     "<dom_formosan_wei_matri_clan_lineage_system_ami>",
+    "<dom_formosan_wei_tien_tsai_subjecthood_temporal_adju>",
+    "<dom_formosan_wiedfelt_formosan_aborigines_atayal>",
     "<dom_formosan_wilang_yutas_videos>",
-    "<dom_formosan_wu_jinglan_introduction_amis_grammar>",
+    "<dom_formosan_wu_chunming_adverbials_in_paiwan>",
+    "<dom_formosan_xie_fuhui_introduction_kavalan_grammar>",
     "<dom_formosan_yedda_palemeq_blog>",
     "<dom_formosan_yeddas_blog>",
     "<dom_formosan_yeh_meili_introduction_saisiyat_grammar>",
+    "<dom_formosan_yi_ming_chou_object_control_saisiyat>",
+    "<dom_formosan_yi_ting_chen_a_minimalist_approach_to_a>",
+    "<dom_formosan_yi_ting_chen_amis_left_periphery>",
+    "<dom_formosan_yi_ting_chen_language_use_code_mixing_a>",
+    "<dom_formosan_yi_yang_cheng_kanakanavu_word_level_pro>",
+    "<dom_formosan_yi_yang_cheng_tense_aspect_agent_markin>",
+    "<dom_formosan_yoshiro_nihira_bunun_vocabulary>",
+    "<dom_formosan_yuehchien_chien_the_lexical_system_of_y>",
+    "<dom_formosan_yukihiro_yamada_ying_chu_liao_phonology>",
+    "<dom_formosan_zeng_shifen_reduplication_affixation_pa>",
+    "<dom_formosan_zhan_sujuan_taiwan_indigenous_peoples_h>",
+    "<dom_formosan_zhang_xiujuan_introduction_paiwan_gramm>",
     "<dom_formosan_zhang_yongli_pan_jiarong_introduction_t>",
+    "<dom_formosan_zhao_shanhe_li_taiyuan_zhu_qingyi_abori>",
     "<dom_formosan_zheng_acl_2024>",
     "<dom_formosan_zheng_data>",
+    "<dom_formosan_zheng_yumei_zhu_qingyi_bo_hongming_abor>",
     "<dom_learning_vocab>",
     "<dom_nine_level>",
     "<dom_ntu>",
@@ -407,8 +533,8 @@
   },
   "mask_token": {
     "content": "<mask>",
-    "lstrip": false,
-    "normalized": false,
+    "lstrip": true,
+    "normalized": true,
     "rstrip": false,
     "single_word": false
   },

tokenizer_config.json

@@ -1650,13 +1650,13 @@
     },
     "256203": {
       "content": "<mask>",
-      "lstrip": false,
-      "normalized": false,
+      "lstrip": true,
+      "normalized": true,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259624": {
+    "261357": {
       "content": "ami_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1664,7 +1664,7 @@
       "single_word": false,
       "special": true
     },
-    "259625": {
+    "261358": {
       "content": "bnn_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1672,7 +1672,7 @@
       "single_word": false,
       "special": true
     },
-    "259626": {
+    "261359": {
       "content": "ckv_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1680,7 +1680,7 @@
       "single_word": false,
       "special": true
     },
-    "259627": {
+    "261360": {
       "content": "dru_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1688,7 +1688,7 @@
       "single_word": false,
       "special": true
     },
-    "259628": {
+    "261361": {
       "content": "pwn_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1696,7 +1696,7 @@
       "single_word": false,
       "special": true
     },
-    "259629": {
+    "261362": {
       "content": "pyu_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1704,7 +1704,7 @@
       "single_word": false,
       "special": true
     },
-    "259630": {
+    "261363": {
       "content": "ssf_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1712,7 +1712,7 @@
       "single_word": false,
       "special": true
     },
-    "259631": {
+    "261364": {
       "content": "sxr_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1720,7 +1720,7 @@
       "single_word": false,
       "special": true
     },
-    "259632": {
+    "261365": {
       "content": "szy_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1728,7 +1728,7 @@
       "single_word": false,
       "special": true
     },
-    "259633": {
+    "261366": {
       "content": "tao_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1736,7 +1736,7 @@
       "single_word": false,
       "special": true
     },
-    "259634": {
+    "261367": {
       "content": "tay_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1744,7 +1744,7 @@
       "single_word": false,
       "special": true
     },
-    "259635": {
+    "261368": {
       "content": "trv_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1752,7 +1752,7 @@
       "single_word": false,
       "special": true
     },
-    "259636": {
+    "261369": {
       "content": "tsu_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1760,7 +1760,7 @@
       "single_word": false,
       "special": true
     },
-    "259637": {
+    "261370": {
       "content": "xnb_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1768,7 +1768,7 @@
       "single_word": false,
       "special": true
     },
-    "259638": {
+    "261371": {
       "content": "xsy_Latn",
       "lstrip": false,
       "normalized": false,
@@ -1776,15 +1776,7 @@
       "single_word": false,
       "special": true
     },
-    "259639": {
-      "content": "<dae>",
-      "lstrip": false,
-      "normalized": false,
-      "rstrip": false,
-      "single_word": false,
-      "special": true
-    },
-    "259640": {
+    "261372": {
       "content": "<dialect_central>",
       "lstrip": false,
       "normalized": false,
@@ -1792,7 +1784,7 @@
       "single_word": false,
       "special": true
     },
-    "259641": {
+    "261373": {
       "content": "<dialect_coastal>",
       "lstrip": false,
       "normalized": false,
@@ -1800,7 +1792,7 @@
       "single_word": false,
       "special": true
     },
-    "259642": {
+    "261374": {
       "content": "<dialect_dawu>",
       "lstrip": false,
       "normalized": false,
@@ -1808,7 +1800,7 @@
       "single_word": false,
       "special": true
     },
-    "259643": {
+    "261375": {
       "content": "<dialect_default>",
       "lstrip": false,
       "normalized": false,
@@ -1816,7 +1808,7 @@
       "single_word": false,
       "special": true
     },
-    "259644": {
+    "261376": {
       "content": "<dialect_deluvalley>",
       "lstrip": false,
       "normalized": false,
@@ -1824,7 +1816,7 @@
       "single_word": false,
       "special": true
     },
-    "259645": {
+    "261377": {
       "content": "<dialect_dona>",
       "lstrip": false,
       "normalized": false,
@@ -1832,7 +1824,7 @@
       "single_word": false,
       "special": true
     },
-    "259646": {
+    "261378": {
       "content": "<dialect_duda>",
       "lstrip": false,
       "normalized": false,
@@ -1840,7 +1832,7 @@
       "single_word": false,
       "special": true
     },
-    "259647": {
+    "261379": {
       "content": "<dialect_eastern>",
       "lstrip": false,
       "normalized": false,
@@ -1848,7 +1840,7 @@
       "single_word": false,
       "special": true
     },
-    "259648": {
+    "261380": {
       "content": "<dialect_fourseasons>",
       "lstrip": false,
       "normalized": false,
@@ -1856,7 +1848,7 @@
       "single_word": false,
       "special": true
     },
-    "259649": {
+    "261381": {
       "content": "<dialect_hengchun>",
       "lstrip": false,
       "normalized": false,
@@ -1864,7 +1856,7 @@
       "single_word": false,
       "special": true
     },
-    "259650": {
+    "261382": {
       "content": "<dialect_jianhe>",
       "lstrip": false,
       "normalized": false,
@@ -1872,7 +1864,7 @@
       "single_word": false,
       "special": true
     },
-    "259651": {
+    "261383": {
       "content": "<dialect_junqun>",
       "lstrip": false,
       "normalized": false,
@@ -1880,7 +1872,7 @@
       "single_word": false,
       "special": true
     },
-    "259652": {
+    "261384": {
       "content": "<dialect_kanakanavu>",
       "lstrip": false,
       "normalized": false,
@@ -1888,7 +1880,7 @@
       "single_word": false,
       "special": true
     },
-    "259653": {
+    "261385": {
       "content": "<dialect_kaqun>",
       "lstrip": false,
       "normalized": false,
@@ -1896,7 +1888,7 @@
       "single_word": false,
       "special": true
     },
-    "259654": {
+    "261386": {
       "content": "<dialect_kavalan>",
       "lstrip": false,
       "normalized": false,
@@ -1904,7 +1896,7 @@
       "single_word": false,
       "special": true
     },
-    "259655": {
+    "261387": {
       "content": "<dialect_luanqun>",
       "lstrip": false,
       "normalized": false,
@@ -1912,7 +1904,7 @@
       "single_word": false,
       "special": true
     },
-    "259656": {
+    "261388": {
       "content": "<dialect_malan>",
       "lstrip": false,
       "normalized": false,
@@ -1920,7 +1912,7 @@
       "single_word": false,
       "special": true
     },
-    "259657": {
+    "261389": {
       "content": "<dialect_maolin>",
       "lstrip": false,
       "normalized": false,
@@ -1928,7 +1920,7 @@
       "single_word": false,
       "special": true
     },
-    "259658": {
+    "261390": {
       "content": "<dialect_nanwang>",
       "lstrip": false,
       "normalized": false,
@@ -1936,7 +1928,7 @@
       "single_word": false,
       "special": true
     },
-    "259659": {
+    "261391": {
       "content": "<dialect_northern>",
       "lstrip": false,
       "normalized": false,
@@ -1944,7 +1936,7 @@
       "single_word": false,
       "special": true
     },
-    "259660": {
+    "261392": {
       "content": "<dialect_saaroa>",
       "lstrip": false,
       "normalized": false,
@@ -1952,7 +1944,7 @@
       "single_word": false,
       "special": true
     },
-    "259661": {
+    "261393": {
       "content": "<dialect_saisiyat>",
       "lstrip": false,
       "normalized": false,
@@ -1960,7 +1952,7 @@
       "single_word": false,
       "special": true
     },
-    "259662": {
+    "261394": {
       "content": "<dialect_sakizaya>",
       "lstrip": false,
       "normalized": false,
@@ -1968,7 +1960,7 @@
       "single_word": false,
       "special": true
     },
-    "259663": {
+    "261395": {
       "content": "<dialect_sekolik>",
       "lstrip": false,
       "normalized": false,
@@ -1976,7 +1968,7 @@
       "single_word": false,
       "special": true
     },
-    "259664": {
+    "261396": {
       "content": "<dialect_southern>",
       "lstrip": false,
       "normalized": false,
@@ -1984,7 +1976,7 @@
       "single_word": false,
       "special": true
     },
-    "259665": {
+    "261397": {
       "content": "<dialect_tanqun>",
       "lstrip": false,
       "normalized": false,
@@ -1992,7 +1984,7 @@
       "single_word": false,
       "special": true
     },
-    "259666": {
+    "261398": {
       "content": "<dialect_tegudaya>",
       "lstrip": false,
       "normalized": false,
@@ -2000,7 +1992,7 @@
       "single_word": false,
       "special": true
     },
-    "259667": {
+    "261399": {
       "content": "<dialect_thao>",
       "lstrip": false,
       "normalized": false,
@@ -2008,7 +2000,7 @@
       "single_word": false,
       "special": true
     },
-    "259668": {
+    "261400": {
       "content": "<dialect_truku>",
       "lstrip": false,
       "normalized": false,
@@ -2016,7 +2008,7 @@
       "single_word": false,
       "special": true
     },
-    "259669": {
+    "261401": {
       "content": "<dialect_tsou>",
       "lstrip": false,
       "normalized": false,
@@ -2024,7 +2016,7 @@
       "single_word": false,
       "special": true
     },
-    "259670": {
+    "261402": {
       "content": "<dialect_unknown>",
       "lstrip": false,
       "normalized": false,
@@ -2032,7 +2024,7 @@
       "single_word": false,
       "special": true
     },
-    "259671": {
+    "261403": {
       "content": "<dialect_wanda>",
       "lstrip": false,
       "normalized": false,
@@ -2040,7 +2032,7 @@
       "single_word": false,
       "special": true
     },
-    "259672": {
+    "261404": {
       "content": "<dialect_wanshan>",
       "lstrip": false,
       "normalized": false,
@@ -2048,7 +2040,7 @@
       "single_word": false,
       "special": true
     },
-    "259673": {
+    "261405": {
       "content": "<dialect_wenshui>",
       "lstrip": false,
       "normalized": false,
@@ -2056,7 +2048,7 @@
       "single_word": false,
       "special": true
     },
-    "259674": {
+    "261406": {
       "content": "<dialect_wutai>",
       "lstrip": false,
       "normalized": false,
@@ -2064,7 +2056,7 @@
       "single_word": false,
       "special": true
     },
-    "259675": {
+    "261407": {
       "content": "<dialect_xiqun>",
       "lstrip": false,
       "normalized": false,
@@ -2072,7 +2064,7 @@
       "single_word": false,
       "special": true
     },
-    "259676": {
+    "261408": {
       "content": "<dialect_xiuguluan>",
       "lstrip": false,
       "normalized": false,
@@ -2080,7 +2072,7 @@
       "single_word": false,
       "special": true
     },
-    "259677": {
+    "261409": {
       "content": "<dialect_yami>",
       "lstrip": false,
       "normalized": false,
@@ -2088,7 +2080,7 @@
       "single_word": false,
       "special": true
     },
-    "259678": {
+    "261410": {
       "content": "<dialect_yilanzeaol>",
       "lstrip": false,
       "normalized": false,
@@ -2096,7 +2088,7 @@
       "single_word": false,
       "special": true
     },
-    "259679": {
+    "261411": {
       "content": "<dialect_zeaol>",
       "lstrip": false,
       "normalized": false,
@@ -2104,7 +2096,7 @@
       "single_word": false,
       "special": true
     },
-    "259680": {
+    "261412": {
       "content": "<dialect_zhiben>",
       "lstrip": false,
       "normalized": false,
@@ -2112,7 +2104,7 @@
       "single_word": false,
       "special": true
     },
-    "259681": {
+    "261413": {
       "content": "<dialect_zhuoqun>",
       "lstrip": false,
       "normalized": false,
@@ -2120,7 +2112,7 @@
       "single_word": false,
       "special": true
     },
-    "259682": {
+    "261414": {
       "content": "<dom_classroom_context>",
       "lstrip": false,
       "normalized": false,
@@ -2128,7 +2120,7 @@
       "single_word": false,
       "special": true
     },
-    "259683": {
+    "261415": {
       "content": "<dom_culture>",
       "lstrip": false,
       "normalized": false,
@@ -2136,7 +2128,7 @@
       "single_word": false,
       "special": true
     },
-    "259684": {
+    "261416": {
       "content": "<dom_dictionary>",
       "lstrip": false,
       "normalized": false,
@@ -2144,7 +2136,7 @@
       "single_word": false,
       "special": true
     },
-    "259685": {
+    "261417": {
       "content": "<dom_essays>",
       "lstrip": false,
       "normalized": false,
@@ -2152,31 +2144,39 @@
       "single_word": false,
       "special": true
     },
-    "259686": {
-      "content": "<dom_ethnography_journal_ritual_song>",
+    "261418": {
+      "content": "<dom_formosan_100_paiwan_texts>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259687": {
-      "content": "<dom_formosan_100_paiwan_texts>",
+    "261419": {
+      "content": "<dom_formosan_academia_sinica_oral_legends>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259688": {
-      "content": "<dom_formosan_academia_sinica_oral_legends>",
+    "261420": {
+      "content": "<dom_formosan_adlay_kun_long_liu_syntactic_interactio>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261421": {
+      "content": "<dom_formosan_adlay_liu_lfg_relativisation_jianshi_sq>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259689": {
+    "261422": {
       "content": "<dom_formosan_amis_adversative_constructions>",
       "lstrip": false,
       "normalized": false,
@@ -2184,31 +2184,31 @@
       "single_word": false,
       "special": true
     },
-    "259690": {
-      "content": "<dom_formosan_amis_kavalan_lin_interrogative_verbs>",
+    "261423": {
+      "content": "<dom_formosan_amis_huang_english_phonology_grammar>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259691": {
-      "content": "<dom_formosan_amis_kuo_sung_comparative_constructions>",
+    "261424": {
+      "content": "<dom_formosan_amis_kavalan_lin_interrogative_verbs>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259692": {
-      "content": "<dom_formosan_amis_lifok_dongi_saopo_kimakimad>",
+    "261425": {
+      "content": "<dom_formosan_amis_kuo_sung_comparative_constructions>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259693": {
+    "261426": {
       "content": "<dom_formosan_amis_myths_and_customs>",
       "lstrip": false,
       "normalized": false,
@@ -2216,7 +2216,7 @@
       "single_word": false,
       "special": true
     },
-    "259694": {
+    "261427": {
       "content": "<dom_formosan_amis_pa_verbs>",
       "lstrip": false,
       "normalized": false,
@@ -2224,7 +2224,7 @@
       "single_word": false,
       "special": true
     },
-    "259695": {
+    "261428": {
       "content": "<dom_formosan_amis_serial_verb_constructions>",
       "lstrip": false,
       "normalized": false,
@@ -2232,7 +2232,7 @@
       "single_word": false,
       "special": true
     },
-    "259696": {
+    "261429": {
       "content": "<dom_formosan_amis_tung_chiou>",
       "lstrip": false,
       "normalized": false,
@@ -2240,7 +2240,39 @@
       "single_word": false,
       "special": true
     },
-    "259697": {
+    "261430": {
+      "content": "<dom_formosan_amy_lee_body_part_nomenclature_seediq>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261431": {
+      "content": "<dom_formosan_anna_hsiou_chuan_chang_reference_gramma>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261432": {
+      "content": "<dom_formosan_anton_quack_puyuma_pulingaw>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261433": {
+      "content": "<dom_formosan_apay_tang_anger_happiness_truku_seediq>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261434": {
       "content": "<dom_formosan_asai_sedik_language>",
       "lstrip": false,
       "normalized": false,
@@ -2248,23 +2280,39 @@
       "single_word": false,
       "special": true
     },
-    "259698": {
-      "content": "<dom_formosan_bunun_debusser_dissertation>",
+    "261435": {
+      "content": "<dom_formosan_atayal_you_nao_savak_raxayal>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261436": {
+      "content": "<dom_formosan_bien_daniel_chiang_house_social_hierarc>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261437": {
+      "content": "<dom_formosan_boyizhenu_narrative_oral_literature_tap>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259699": {
-      "content": "<dom_formosan_bunun_moriguchi_northern_texts>",
+    "261438": {
+      "content": "<dom_formosan_bunun_debusser_dissertation>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259700": {
+    "261439": {
       "content": "<dom_formosan_bunun_topic_focus>",
       "lstrip": false,
       "normalized": false,
@@ -2272,7 +2320,39 @@
       "single_word": false,
       "special": true
     },
-    "259701": {
+    "261440": {
+      "content": "<dom_formosan_c_douglas_chretien_oceanic_material_in>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261441": {
+      "content": "<dom_formosan_catherine_tseng_seediq_atayal_foods_the>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261442": {
+      "content": "<dom_formosan_chang_kwo_tan_syncretic_objects_materia>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261443": {
+      "content": "<dom_formosan_chaolin_li_motion_verb_grammaticalizati>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261444": {
       "content": "<dom_formosan_chen_fukuda_relabeling_ergative>",
       "lstrip": false,
       "normalized": false,
@@ -2280,7 +2360,7 @@
       "single_word": false,
       "special": true
     },
-    "259702": {
+    "261445": {
       "content": "<dom_formosan_chen_fukuda_three_ways_steal>",
       "lstrip": false,
       "normalized": false,
@@ -2288,7 +2368,47 @@
       "single_word": false,
       "special": true
     },
-    "259703": {
+    "261446": {
+      "content": "<dom_formosan_cheng_sung_modality_kanakanavu>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261447": {
+      "content": "<dom_formosan_chi_lu_chen_and_micheal_d_coe_an_invest>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261448": {
+      "content": "<dom_formosan_chuming_wu_linking_constructions_mayrin>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261449": {
+      "content": "<dom_formosan_chun_mei_chen_comparative_phonology_pai>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261450": {
+      "content": "<dom_formosan_chunming_wu_two_types_of_noun_incorpora>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261451": {
       "content": "<dom_formosan_cip_atayal_grammar_overview>",
       "lstrip": false,
       "normalized": false,
@@ -2296,7 +2416,23 @@
       "single_word": false,
       "special": true
     },
-    "259704": {
+    "261452": {
+      "content": "<dom_formosan_claire_mcgill_a_brief_tayal_vocabulary>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261453": {
+      "content": "<dom_formosan_david_blundell_ed_austronesian_taiwan>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261454": {
       "content": "<dom_formosan_dean_johnson_year_clouds_smangus>",
       "lstrip": false,
       "normalized": false,
@@ -2304,7 +2440,47 @@
       "single_word": false,
       "special": true
     },
-    "259705": {
+    "261455": {
+      "content": "<dom_formosan_der_hwa_victoria_rau_grammar_atayal>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261456": {
+      "content": "<dom_formosan_der_hwa_victoria_rau_lexical_similarity>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261457": {
+      "content": "<dom_formosan_dong_manv_yami_ode_to_taro>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261458": {
+      "content": "<dom_formosan_dong_yi_lin_universal_quantifier_kavala>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261459": {
+      "content": "<dom_formosan_dong_yijia_tjuwabar_paiwan_ritual_langu>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261460": {
       "content": "<dom_formosan_dorinda_tsai_hsiu_liu_neutral_imperfect>",
       "lstrip": false,
       "normalized": false,
@@ -2312,7 +2488,15 @@
       "single_word": false,
       "special": true
     },
-    "259706": {
+    "261461": {
+      "content": "<dom_formosan_ed_torjesen_amis_taiwan_word_list>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261462": {
       "content": "<dom_formosan_egerod_origin_headhunting_atayal_text_v>",
       "lstrip": false,
       "normalized": false,
@@ -2320,7 +2504,7 @@
       "single_word": false,
       "special": true
     },
-    "259707": {
+    "261463": {
       "content": "<dom_formosan_egerod_soren_word_order_word_classes_at>",
       "lstrip": false,
       "normalized": false,
@@ -2328,7 +2512,7 @@
       "single_word": false,
       "special": true
     },
-    "259708": {
+    "261464": {
       "content": "<dom_formosan_elizabeth_zeitoun_chen_huei_wu_overview>",
       "lstrip": false,
       "normalized": false,
@@ -2336,7 +2520,39 @@
       "single_word": false,
       "special": true
     },
-    "259709": {
+    "261465": {
+      "content": "<dom_formosan_elizabeth_zeitoun_lillian_huang_anna_ch>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261466": {
+      "content": "<dom_formosan_elizabeth_zeitoun_lillian_huang_marie_y>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261467": {
+      "content": "<dom_formosan_elizabeth_zeitoun_pronominal_system_man>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261468": {
+      "content": "<dom_formosan_elizabeth_zeitoun_squib_dynamic_vs_stat>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261469": {
       "content": "<dom_formosan_epark>",
       "lstrip": false,
       "normalized": false,
@@ -2344,7 +2560,7 @@
       "single_word": false,
       "special": true
     },
-    "259710": {
+    "261470": {
       "content": "<dom_formosan_ferrell_taiwan_aboriginal_groups_proble>",
       "lstrip": false,
       "normalized": false,
@@ -2352,7 +2568,15 @@
       "single_word": false,
       "special": true
     },
-    "259711": {
+    "261471": {
+      "content": "<dom_formosan_george_leroy_shelley_vudai_rukai>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261472": {
       "content": "<dom_formosan_gitbook_translations>",
       "lstrip": false,
       "normalized": false,
@@ -2360,7 +2584,7 @@
       "single_word": false,
       "special": true
     },
-    "259712": {
+    "261473": {
       "content": "<dom_formosan_glosbe>",
       "lstrip": false,
       "normalized": false,
@@ -2368,391 +2592,1183 @@
       "single_word": false,
       "special": true
     },
-    "259713": {
-      "content": "<dom_formosan_hala_saku_la_videos>",
+    "261474": {
+      "content": "<dom_formosan_gujing_lin_tsou_serial_verb_constructio>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259714": {
-      "content": "<dom_formosan_holmer_seediq>",
+    "261475": {
+      "content": "<dom_formosan_henry_chang_nominal_aspect_tsou>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259715": {
-      "content": "<dom_formosan_huang_grammaticalization_squliq_atayal>",
+    "261476": {
+      "content": "<dom_formosan_henry_chang_tsou_causative_applicative>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259716": {
-      "content": "<dom_formosan_huang_mei_chin_atayal_reference_grammar>",
+    "261477": {
+      "content": "<dom_formosan_henry_y_chang_formosan_speech_act_mood>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259717": {
-      "content": "<dom_formosan_huteson_rukai_survey>",
+    "261478": {
+      "content": "<dom_formosan_ho_dan_phonological_system_butonglu_pai>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259718": {
-      "content": "<dom_formosan_ilrdf_42_language_practice_word_lists>",
+    "261479": {
+      "content": "<dom_formosan_holmer_seediq>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259719": {
-      "content": "<dom_formosan_ilrdf_tousvusvutu_kita>",
+    "261480": {
+      "content": "<dom_formosan_hsiu_chuan_liao_transitivity_ergativity>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259720": {
-      "content": "<dom_formosan_indigenous_language_literary_awards>",
+    "261481": {
+      "content": "<dom_formosan_huang_dongqiu_atayal_bunun_amis_languag>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259721": {
-      "content": "<dom_formosan_kanakanavu_texts>",
+    "261482": {
+      "content": "<dom_formosan_huang_grammaticalization_squliq_atayal>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259722": {
-      "content": "<dom_formosan_kavalan_zhang_reference_grammar>",
+    "261483": {
+      "content": "<dom_formosan_huang_hui_chuan_glide_formation_takitud>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259723": {
-      "content": "<dom_formosan_li_peirong_xu_weisheng_introduction_tru>",
+    "261484": {
+      "content": "<dom_formosan_huang_mei_chin_atayal_reference_grammar>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259724": {
-      "content": "<dom_formosan_li_peizhen_wu_xingyu_shouhu_xing>",
+    "261485": {
+      "content": "<dom_formosan_hui_chuan_j_huang_syllable_types_in_bun>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259725": {
-      "content": "<dom_formosan_naomi_tsukida_correlative_clauses_seedi>",
+    "261486": {
+      "content": "<dom_formosan_hui_shan_lin_squliq_atayal_reduplicatio>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259726": {
-      "content": "<dom_formosan_nowbucyang_truku_thesis>",
+    "261487": {
+      "content": "<dom_formosan_huteson_rukai_survey>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259727": {
-      "content": "<dom_formosan_old_texts>",
+    "261488": {
+      "content": "<dom_formosan_ian_maddieson_and_richard_wright_the_vo>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259728": {
-      "content": "<dom_formosan_paiwan_collart_zeitoun_time_reference>",
+    "261489": {
+      "content": "<dom_formosan_ilrdf_42_language_practice_word_lists>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259729": {
-      "content": "<dom_formosan_paiwan_ho_five_dialects>",
+    "261490": {
+      "content": "<dom_formosan_ilrdf_tousvusvutu_kita>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259730": {
-      "content": "<dom_formosan_paiwanstories>",
+    "261491": {
+      "content": "<dom_formosan_indigenous_education_resource_center_ke>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259731": {
-      "content": "<dom_formosan_puyuma_chen_raising_to_object>",
+    "261492": {
+      "content": "<dom_formosan_indigenous_language_literary_awards>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259732": {
-      "content": "<dom_formosan_puyuma_katipol_kopfjagdriten>",
+    "261493": {
+      "content": "<dom_formosan_isabelle_bril_roots_and_stems_lexical_a>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259733": {
-      "content": "<dom_formosan_puyuma_teng_pronominal_systems>",
+    "261494": {
+      "content": "<dom_formosan_jian_iwan_guo_qingliu_life_history>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259734": {
-      "content": "<dom_formosan_puyuma_teng_reference_grammar>",
+    "261495": {
+      "content": "<dom_formosan_jian_shilang_introduction_thao_grammar>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259735": {
-      "content": "<dom_formosan_rik_bunun>",
+    "261496": {
+      "content": "<dom_formosan_john_tu_er_wei_notes_on_the_religious_b>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259736": {
-      "content": "<dom_formosan_robert_blust_austronesian_homeland_ling>",
+    "261497": {
+      "content": "<dom_formosan_john_u_wolff_the_proto_austronesian_pho>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259737": {
-      "content": "<dom_formosan_rukai_mantauran_stories>",
+    "261498": {
+      "content": "<dom_formosan_jonathan_kuo_amis_manner_result_verbs>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259738": {
-      "content": "<dom_formosan_rukai_texts>",
+    "261499": {
+      "content": "<dom_formosan_josiane_cauquelin_aborigines_taiwan_puy>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259739": {
-      "content": "<dom_formosan_rukai_zeitoun_saying>",
+    "261500": {
+      "content": "<dom_formosan_joy_wu_clausal_modifiers_in_amis>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259740": {
-      "content": "<dom_formosan_saaroa_pan_grammar>",
+    "261501": {
+      "content": "<dom_formosan_joy_wu_verb_classification_case_marking>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259741": {
-      "content": "<dom_formosan_sakizaya_affixes>",
+    "261502": {
+      "content": "<dom_formosan_kanakanavu_texts>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259742": {
-      "content": "<dom_formosan_seals>",
+    "261503": {
+      "content": "<dom_formosan_kavalan_zhang_reference_grammar>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259743": {
-      "content": "<dom_formosan_seediq_zhang_reference_grammar>",
+    "261504": {
+      "content": "<dom_formosan_kumu_tapas_tribal_memory_oral_history_w>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259744": {
-      "content": "<dom_formosan_shigeru_tsuchida_kanakanavu_texts>",
+    "261505": {
+      "content": "<dom_formosan_kuo_j_c_argument_alternation_and_argume>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259745": {
-      "content": "<dom_formosan_shigeru_tsuchida_reconstruction_proto_t>",
+    "261506": {
+      "content": "<dom_formosan_lei_shih_family_system_paiwan_at_su_pai>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259746": {
-      "content": "<dom_formosan_shih_rukai_adverbial>",
+    "261507": {
+      "content": "<dom_formosan_li_chen_yeh_developing_a_hla_alua_learn>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259747": {
-      "content": "<dom_formosan_song_limei_introduction_kanakanavu_gram>",
+    "261508": {
+      "content": "<dom_formosan_li_may_sung_budai_rukai_exclamatives>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259748": {
-      "content": "<dom_formosan_song_limei_introduction_seediq_grammar>",
+    "261509": {
+      "content": "<dom_formosan_li_may_sung_clausal_nominalization_buda>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259749": {
-      "content": "<dom_formosan_sung_kuo_descriptive_comparative_constr>",
+    "261510": {
+      "content": "<dom_formosan_lillian_huang_atayal_participants_cross>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259750": {
-      "content": "<dom_formosan_taoshan_elementary_school_atelier_hui_k>",
+    "261511": {
+      "content": "<dom_formosan_lillian_huang_ergativity_atayal>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259751": {
-      "content": "<dom_formosan_tsukida_naomi_verb_classification_amis>",
+    "261512": {
+      "content": "<dom_formosan_liu_manyi_kaskun_ata_matas_i_ki_quaz>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259752": {
-      "content": "<dom_formosan_wei_matri_clan_lineage_system_ami>",
+    "261513": {
+      "content": "<dom_formosan_malcom_d_ross_the_history_and_transitiv>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259753": {
-      "content": "<dom_formosan_wilang_yutas_videos>",
+    "261514": {
+      "content": "<dom_formosan_malcom_ross_proto_austronesian_verbal_m>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259754": {
-      "content": "<dom_formosan_wu_jinglan_introduction_amis_grammar>",
+    "261515": {
+      "content": "<dom_formosan_man_ni_chu_the_measurement_of_final_i_v>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259755": {
-      "content": "<dom_formosan_yedda_palemeq_blog>",
+    "261516": {
+      "content": "<dom_formosan_matsuzawa_kazuko_the_social_and_ritual>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259756": {
-      "content": "<dom_formosan_yeddas_blog>",
+    "261517": {
+      "content": "<dom_formosan_maya_yeh_blaq_uv_construction_atayal>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259757": {
-      "content": "<dom_formosan_yeh_meili_introduction_saisiyat_grammar>",
+    "261518": {
+      "content": "<dom_formosan_mayumi_oiwa_adverbial_verb_construction>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259758": {
-      "content": "<dom_formosan_zhang_yongli_pan_jiarong_introduction_t>",
+    "261519": {
+      "content": "<dom_formosan_mayumi_oiwa_bungard_morphology_and_synt>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259759": {
-      "content": "<dom_formosan_zheng_acl_2024>",
+    "261520": {
+      "content": "<dom_formosan_mayumi_oiwa_possessor_raising_truku_see>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259760": {
-      "content": "<dom_formosan_zheng_data>",
+    "261521": {
+      "content": "<dom_formosan_mei_chin_huang_verb_classification_in_m>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261522": {
+      "content": "<dom_formosan_melissa_shih_hui_lin_sakizaya_or_amis_a>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261523": {
+      "content": "<dom_formosan_nanang_tadaw_naci_mowna_pgagu>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261524": {
+      "content": "<dom_formosan_naomi_tsukida_correlative_clauses_seedi>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261525": {
+      "content": "<dom_formosan_naomi_tsukida_ditransitive_causative_se>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261526": {
+      "content": "<dom_formosan_nowbucyang_truku_thesis>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261527": {
+      "content": "<dom_formosan_ochiai_izumi_2016_comparative_linguisti>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261528": {
+      "content": "<dom_formosan_ochiai_izumi_bu_hwan_vocabulary_recorde>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261529": {
+      "content": "<dom_formosan_ochiai_izumi_numerals_paran_seediq_rela>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261530": {
+      "content": "<dom_formosan_ochiai_izumi_on_a_seediq_glossary_recor>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261531": {
+      "content": "<dom_formosan_old_texts>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261532": {
+      "content": "<dom_formosan_paiwan_ho_five_dialects>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261533": {
+      "content": "<dom_formosan_paiwanstories>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261534": {
+      "content": "<dom_formosan_pang_hsin_ting_reconstruction_proto_puy>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261535": {
+      "content": "<dom_formosan_patricia_stanley_morphophonemics_of_ver>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261536": {
+      "content": "<dom_formosan_paul_jen_kuei_li_comparative_bunun_dial>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261537": {
+      "content": "<dom_formosan_paul_jen_kuei_li_conjunction_thao>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261538": {
+      "content": "<dom_formosan_paul_jen_kuei_li_internal_relationships>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261539": {
+      "content": "<dom_formosan_paul_jen_kuei_li_origins_of_the_east_fo>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261540": {
+      "content": "<dom_formosan_paul_jen_kuei_li_review_of_taiwan_abori>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261541": {
+      "content": "<dom_formosan_paul_jen_kuei_li_rukai_structure>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261542": {
+      "content": "<dom_formosan_paul_jen_kuei_li_some_plant_names_in_fo>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261543": {
+      "content": "<dom_formosan_paul_jen_kuei_li_the_preglottalised_sto>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261544": {
+      "content": "<dom_formosan_paul_li_atayalic_final_voiced_stops>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261545": {
+      "content": "<dom_formosan_paul_li_case_marking_four_formosan_lang>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261546": {
+      "content": "<dom_formosan_paul_li_position_atayal_austronesian>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261547": {
+      "content": "<dom_formosan_pingdong_county_government_kai_na_sepuc>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261548": {
+      "content": "<dom_formosan_puyuma_chen_raising_to_object>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261549": {
+      "content": "<dom_formosan_puyuma_katipol_kopfjagdriten>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261550": {
+      "content": "<dom_formosan_puyuma_teng_pronominal_systems>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261551": {
+      "content": "<dom_formosan_puyuma_teng_reference_grammar>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261552": {
+      "content": "<dom_formosan_raleigh_ferrell_construction_markers_fo>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261553": {
+      "content": "<dom_formosan_raleigh_ferrell_paiwan_phonology_and_pr>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261554": {
+      "content": "<dom_formosan_raleigh_j_ferrell_intent_and_volition_i>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261555": {
+      "content": "<dom_formosan_rik_bunun>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261556": {
+      "content": "<dom_formosan_rik_de_busser_the_influence_of_christia>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261557": {
+      "content": "<dom_formosan_robert_blust_a_note_on_covert_structure>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261558": {
+      "content": "<dom_formosan_robert_blust_austronesian_homeland_ling>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261559": {
+      "content": "<dom_formosan_robert_blust_some_remarks_linguistic_po>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261560": {
+      "content": "<dom_formosan_robert_blust_thao_triplication>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261561": {
+      "content": "<dom_formosan_robert_blust_the_austronesian_languages>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261562": {
+      "content": "<dom_formosan_rukai_mantauran_stories>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261563": {
+      "content": "<dom_formosan_rukai_texts>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261564": {
+      "content": "<dom_formosan_rukai_zeitoun_saying>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261565": {
+      "content": "<dom_formosan_sakizaya_affixes>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261566": {
+      "content": "<dom_formosan_schetelig_mittheilungen_uber_die_sprach>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261567": {
+      "content": "<dom_formosan_seals>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261568": {
+      "content": "<dom_formosan_seediq_zhang_reference_grammar>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261569": {
+      "content": "<dom_formosan_shigeru_tsuchida_and_isidore_dyen_proto>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261570": {
+      "content": "<dom_formosan_shigeru_tsuchida_kanakanavu_texts>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261571": {
+      "content": "<dom_formosan_shigeru_tsuchida_preliminary_reports_on>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261572": {
+      "content": "<dom_formosan_shigeru_tsuchida_reconstruction_proto_t>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261573": {
+      "content": "<dom_formosan_shih_chi_stella_yeh_stress_and_quantity>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261574": {
+      "content": "<dom_formosan_shih_rukai_adverbial>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261575": {
+      "content": "<dom_formosan_sihwei_chen_and_haowen_jiang_ways_of_ta>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261576": {
+      "content": "<dom_formosan_sihwei_chen_lexical_aspect_atayal>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261577": {
+      "content": "<dom_formosan_soren_egerod_statement_atayal_phonology>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261578": {
+      "content": "<dom_formosan_soren_egerod_verb_inflection_atayal>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261579": {
+      "content": "<dom_formosan_stacy_teng_malcom_ross_is_puyuma_primar>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261580": {
+      "content": "<dom_formosan_stacy_wan_tin_huang_functions_yami_verb>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261581": {
+      "content": "<dom_formosan_sung_kuo_descriptive_comparative_constr>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261582": {
+      "content": "<dom_formosan_taoshan_elementary_school_atelier_hui_k>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261583": {
+      "content": "<dom_formosan_teresa_m_chen_verbal_construction_and_v>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261584": {
+      "content": "<dom_formosan_ting_chi_wei_parallelism_in_amis_sluici>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261585": {
+      "content": "<dom_formosan_tingchun_chen_successive_cyclic_case_as>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261586": {
+      "content": "<dom_formosan_truku_lowking_demonstratives>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261587": {
+      "content": "<dom_formosan_tsai_hsui_liu_complementation_three_lan>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261588": {
+      "content": "<dom_formosan_tsai_wei_tien_dylan_conjunctive_reducti>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261589": {
+      "content": "<dom_formosan_tsou_descriptive_study>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261590": {
+      "content": "<dom_formosan_tsukida_naomi_verb_classification_amis>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261591": {
+      "content": "<dom_formosan_tu_wen_chiu_a_synchronic_classification>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261592": {
+      "content": "<dom_formosan_tung_tsuchida_ting_li_pan_saaroa_texts>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261593": {
+      "content": "<dom_formosan_victoria_rau_yi_hsin_wu_and_meng_chien>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261594": {
+      "content": "<dom_formosan_wei_huilin_liu_social_structure_yami>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261595": {
+      "content": "<dom_formosan_wei_matri_clan_lineage_system_ami>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261596": {
+      "content": "<dom_formosan_wei_tien_tsai_subjecthood_temporal_adju>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261597": {
+      "content": "<dom_formosan_wiedfelt_formosan_aborigines_atayal>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261598": {
+      "content": "<dom_formosan_wilang_yutas_videos>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261599": {
+      "content": "<dom_formosan_wu_chunming_adverbials_in_paiwan>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261600": {
+      "content": "<dom_formosan_xie_fuhui_introduction_kavalan_grammar>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261601": {
+      "content": "<dom_formosan_yedda_palemeq_blog>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261602": {
+      "content": "<dom_formosan_yeddas_blog>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261603": {
+      "content": "<dom_formosan_yeh_meili_introduction_saisiyat_grammar>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261604": {
+      "content": "<dom_formosan_yi_ming_chou_object_control_saisiyat>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261605": {
+      "content": "<dom_formosan_yi_ting_chen_a_minimalist_approach_to_a>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261606": {
+      "content": "<dom_formosan_yi_ting_chen_amis_left_periphery>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261607": {
+      "content": "<dom_formosan_yi_ting_chen_language_use_code_mixing_a>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261608": {
+      "content": "<dom_formosan_yi_yang_cheng_kanakanavu_word_level_pro>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261609": {
+      "content": "<dom_formosan_yi_yang_cheng_tense_aspect_agent_markin>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261610": {
+      "content": "<dom_formosan_yoshiro_nihira_bunun_vocabulary>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261611": {
+      "content": "<dom_formosan_yuehchien_chien_the_lexical_system_of_y>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261612": {
+      "content": "<dom_formosan_yukihiro_yamada_ying_chu_liao_phonology>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261613": {
+      "content": "<dom_formosan_zeng_shifen_reduplication_affixation_pa>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261614": {
+      "content": "<dom_formosan_zhan_sujuan_taiwan_indigenous_peoples_h>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261615": {
+      "content": "<dom_formosan_zhang_xiujuan_introduction_paiwan_gramm>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261616": {
+      "content": "<dom_formosan_zhang_yongli_pan_jiarong_introduction_t>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261617": {
+      "content": "<dom_formosan_zhao_shanhe_li_taiyuan_zhu_qingyi_abori>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261618": {
+      "content": "<dom_formosan_zheng_acl_2024>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261619": {
+      "content": "<dom_formosan_zheng_data>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261620": {
+      "content": "<dom_formosan_zheng_yumei_zhu_qingyi_bo_hongming_abor>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "259761": {
+    "261621": {
       "content": "<dom_learning_vocab>",
       "lstrip": false,
       "normalized": false,
@@ -2760,7 +3776,7 @@
       "single_word": false,
       "special": true
     },
-    "259762": {
+    "261622": {
       "content": "<dom_nine_level>",
       "lstrip": false,
       "normalized": false,
@@ -2768,7 +3784,7 @@
       "single_word": false,
       "special": true
     },
-    "259763": {
+    "261623": {
       "content": "<dom_ntu>",
       "lstrip": false,
       "normalized": false,
@@ -2776,7 +3792,7 @@
       "single_word": false,
       "special": true
     },
-    "259764": {
+    "261624": {
       "content": "<dom_picture_book>",
       "lstrip": false,
       "normalized": false,
@@ -2784,7 +3800,7 @@
       "single_word": false,
       "special": true
     },
-    "259765": {
+    "261625": {
       "content": "<dom_picture_story>",
       "lstrip": false,
       "normalized": false,
@@ -2792,7 +3808,7 @@
       "single_word": false,
       "special": true
     },
-    "259766": {
+    "261626": {
       "content": "<dom_presidential_apology>",
       "lstrip": false,
       "normalized": false,
@@ -2800,7 +3816,7 @@
       "single_word": false,
       "special": true
     },
-    "259767": {
+    "261627": {
       "content": "<dom_reading_writing>",
       "lstrip": false,
       "normalized": false,
@@ -2808,7 +3824,7 @@
       "single_word": false,
       "special": true
     },
-    "259768": {
+    "261628": {
       "content": "<dom_unknown>",
       "lstrip": false,
       "normalized": false,
@@ -2816,7 +3832,7 @@
       "single_word": false,
       "special": true
     },
-    "259769": {
+    "261629": {
       "content": "<dom_youtube>",
       "lstrip": false,
       "normalized": false,
@@ -2824,7 +3840,7 @@
       "single_word": false,
       "special": true
     },
-    "259770": {
+    "261630": {
       "content": "<src_ami>",
       "lstrip": false,
       "normalized": false,
@@ -2832,7 +3848,7 @@
       "single_word": false,
       "special": true
     },
-    "259771": {
+    "261631": {
       "content": "<src_bnn>",
       "lstrip": false,
       "normalized": false,
@@ -2840,7 +3856,7 @@
       "single_word": false,
       "special": true
     },
-    "259772": {
+    "261632": {
       "content": "<src_ckv>",
       "lstrip": false,
       "normalized": false,
@@ -2848,7 +3864,7 @@
       "single_word": false,
       "special": true
     },
-    "259773": {
+    "261633": {
       "content": "<src_dru>",
       "lstrip": false,
       "normalized": false,
@@ -2856,7 +3872,7 @@
       "single_word": false,
       "special": true
     },
-    "259774": {
+    "261634": {
       "content": "<src_eng>",
       "lstrip": false,
       "normalized": false,
@@ -2864,7 +3880,7 @@
       "single_word": false,
       "special": true
     },
-    "259775": {
+    "261635": {
       "content": "<src_pwn>",
       "lstrip": false,
       "normalized": false,
@@ -2872,7 +3888,7 @@
       "single_word": false,
       "special": true
     },
-    "259776": {
+    "261636": {
       "content": "<src_pyu>",
       "lstrip": false,
       "normalized": false,
@@ -2880,7 +3896,7 @@
       "single_word": false,
       "special": true
     },
-    "259777": {
+    "261637": {
       "content": "<src_ssf>",
       "lstrip": false,
       "normalized": false,
@@ -2888,7 +3904,7 @@
       "single_word": false,
       "special": true
     },
-    "259778": {
+    "261638": {
       "content": "<src_sxr>",
       "lstrip": false,
       "normalized": false,
@@ -2896,7 +3912,7 @@
       "single_word": false,
       "special": true
     },
-    "259779": {
+    "261639": {
       "content": "<src_szy>",
       "lstrip": false,
       "normalized": false,
@@ -2904,7 +3920,7 @@
       "single_word": false,
       "special": true
     },
-    "259780": {
+    "261640": {
       "content": "<src_tao>",
       "lstrip": false,
       "normalized": false,
@@ -2912,7 +3928,7 @@
       "single_word": false,
       "special": true
     },
-    "259781": {
+    "261641": {
       "content": "<src_tay>",
       "lstrip": false,
       "normalized": false,
@@ -2920,7 +3936,7 @@
       "single_word": false,
       "special": true
     },
-    "259782": {
+    "261642": {
       "content": "<src_trv>",
       "lstrip": false,
       "normalized": false,
@@ -2928,7 +3944,7 @@
       "single_word": false,
       "special": true
     },
-    "259783": {
+    "261643": {
       "content": "<src_tsu>",
       "lstrip": false,
       "normalized": false,
@@ -2936,7 +3952,7 @@
       "single_word": false,
       "special": true
     },
-    "259784": {
+    "261644": {
       "content": "<src_xnb>",
       "lstrip": false,
       "normalized": false,
@@ -2944,7 +3960,7 @@
       "single_word": false,
       "special": true
     },
-    "259785": {
+    "261645": {
       "content": "<src_xsy>",
       "lstrip": false,
       "normalized": false,
@@ -2952,7 +3968,7 @@
       "single_word": false,
       "special": true
     },
-    "259786": {
+    "261646": {
       "content": "<src_zh>",
       "lstrip": false,
       "normalized": false,
@@ -2960,7 +3976,7 @@
       "single_word": false,
       "special": true
     },
-    "259787": {
+    "261647": {
       "content": "<to_ami>",
       "lstrip": false,
       "normalized": false,
@@ -2968,7 +3984,7 @@
       "single_word": false,
       "special": true
     },
-    "259788": {
+    "261648": {
       "content": "<to_bnn>",
       "lstrip": false,
       "normalized": false,
@@ -2976,7 +3992,7 @@
       "single_word": false,
       "special": true
     },
-    "259789": {
+    "261649": {
       "content": "<to_ckv>",
       "lstrip": false,
       "normalized": false,
@@ -2984,7 +4000,7 @@
       "single_word": false,
       "special": true
     },
-    "259790": {
+    "261650": {
       "content": "<to_dru>",
       "lstrip": false,
       "normalized": false,
@@ -2992,7 +4008,7 @@
       "single_word": false,
       "special": true
     },
-    "259791": {
+    "261651": {
       "content": "<to_eng>",
       "lstrip": false,
       "normalized": false,
@@ -3000,7 +4016,7 @@
       "single_word": false,
       "special": true
     },
-    "259792": {
+    "261652": {
       "content": "<to_pwn>",
       "lstrip": false,
       "normalized": false,
@@ -3008,7 +4024,7 @@
       "single_word": false,
       "special": true
     },
-    "259793": {
+    "261653": {
       "content": "<to_pyu>",
       "lstrip": false,
       "normalized": false,
@@ -3016,7 +4032,7 @@
       "single_word": false,
       "special": true
     },
-    "259794": {
+    "261654": {
       "content": "<to_ssf>",
       "lstrip": false,
       "normalized": false,
@@ -3024,7 +4040,7 @@
       "single_word": false,
       "special": true
     },
-    "259795": {
+    "261655": {
       "content": "<to_sxr>",
       "lstrip": false,
       "normalized": false,
@@ -3032,7 +4048,7 @@
       "single_word": false,
       "special": true
     },
-    "259796": {
+    "261656": {
       "content": "<to_szy>",
       "lstrip": false,
       "normalized": false,
@@ -3040,7 +4056,7 @@
       "single_word": false,
       "special": true
     },
-    "259797": {
+    "261657": {
       "content": "<to_tao>",
       "lstrip": false,
       "normalized": false,
@@ -3048,7 +4064,7 @@
       "single_word": false,
       "special": true
     },
-    "259798": {
+    "261658": {
       "content": "<to_tay>",
       "lstrip": false,
       "normalized": false,
@@ -3056,7 +4072,7 @@
       "single_word": false,
       "special": true
     },
-    "259799": {
+    "261659": {
       "content": "<to_trv>",
       "lstrip": false,
       "normalized": false,
@@ -3064,7 +4080,7 @@
       "single_word": false,
       "special": true
     },
-    "259800": {
+    "261660": {
       "content": "<to_tsu>",
       "lstrip": false,
       "normalized": false,
@@ -3072,7 +4088,7 @@
       "single_word": false,
       "special": true
     },
-    "259801": {
+    "261661": {
       "content": "<to_xnb>",
       "lstrip": false,
       "normalized": false,
@@ -3080,7 +4096,7 @@
       "single_word": false,
       "special": true
     },
-    "259802": {
+    "261662": {
       "content": "<to_xsy>",
       "lstrip": false,
       "normalized": false,
@@ -3088,7 +4104,7 @@
       "single_word": false,
       "special": true
     },
-    "259803": {
+    "261663": {
       "content": "<to_zh>",
       "lstrip": false,
       "normalized": false,
@@ -3315,8 +4331,6 @@
     "tsu_Latn",
     "xnb_Latn",
     "xsy_Latn",
-    "<mask>",
-    "<dae>",
     "<dialect_central>",
     "<dialect_coastal>",
     "<dialect_dawu>",
@@ -3363,81 +4377,209 @@
     "<dom_culture>",
     "<dom_dictionary>",
     "<dom_essays>",
-    "<dom_ethnography_journal_ritual_song>",
     "<dom_formosan_100_paiwan_texts>",
     "<dom_formosan_academia_sinica_oral_legends>",
+    "<dom_formosan_adlay_kun_long_liu_syntactic_interactio>",
+    "<dom_formosan_adlay_liu_lfg_relativisation_jianshi_sq>",
     "<dom_formosan_amis_adversative_constructions>",
+    "<dom_formosan_amis_huang_english_phonology_grammar>",
     "<dom_formosan_amis_kavalan_lin_interrogative_verbs>",
     "<dom_formosan_amis_kuo_sung_comparative_constructions>",
-    "<dom_formosan_amis_lifok_dongi_saopo_kimakimad>",
     "<dom_formosan_amis_myths_and_customs>",
     "<dom_formosan_amis_pa_verbs>",
     "<dom_formosan_amis_serial_verb_constructions>",
     "<dom_formosan_amis_tung_chiou>",
+    "<dom_formosan_amy_lee_body_part_nomenclature_seediq>",
+    "<dom_formosan_anna_hsiou_chuan_chang_reference_gramma>",
+    "<dom_formosan_anton_quack_puyuma_pulingaw>",
+    "<dom_formosan_apay_tang_anger_happiness_truku_seediq>",
     "<dom_formosan_asai_sedik_language>",
+    "<dom_formosan_atayal_you_nao_savak_raxayal>",
+    "<dom_formosan_bien_daniel_chiang_house_social_hierarc>",
+    "<dom_formosan_boyizhenu_narrative_oral_literature_tap>",
     "<dom_formosan_bunun_debusser_dissertation>",
-    "<dom_formosan_bunun_moriguchi_northern_texts>",
     "<dom_formosan_bunun_topic_focus>",
+    "<dom_formosan_c_douglas_chretien_oceanic_material_in>",
+    "<dom_formosan_catherine_tseng_seediq_atayal_foods_the>",
+    "<dom_formosan_chang_kwo_tan_syncretic_objects_materia>",
+    "<dom_formosan_chaolin_li_motion_verb_grammaticalizati>",
     "<dom_formosan_chen_fukuda_relabeling_ergative>",
     "<dom_formosan_chen_fukuda_three_ways_steal>",
+    "<dom_formosan_cheng_sung_modality_kanakanavu>",
+    "<dom_formosan_chi_lu_chen_and_micheal_d_coe_an_invest>",
+    "<dom_formosan_chuming_wu_linking_constructions_mayrin>",
+    "<dom_formosan_chun_mei_chen_comparative_phonology_pai>",
+    "<dom_formosan_chunming_wu_two_types_of_noun_incorpora>",
     "<dom_formosan_cip_atayal_grammar_overview>",
+    "<dom_formosan_claire_mcgill_a_brief_tayal_vocabulary>",
+    "<dom_formosan_david_blundell_ed_austronesian_taiwan>",
     "<dom_formosan_dean_johnson_year_clouds_smangus>",
+    "<dom_formosan_der_hwa_victoria_rau_grammar_atayal>",
+    "<dom_formosan_der_hwa_victoria_rau_lexical_similarity>",
+    "<dom_formosan_dong_manv_yami_ode_to_taro>",
+    "<dom_formosan_dong_yi_lin_universal_quantifier_kavala>",
+    "<dom_formosan_dong_yijia_tjuwabar_paiwan_ritual_langu>",
     "<dom_formosan_dorinda_tsai_hsiu_liu_neutral_imperfect>",
+    "<dom_formosan_ed_torjesen_amis_taiwan_word_list>",
     "<dom_formosan_egerod_origin_headhunting_atayal_text_v>",
     "<dom_formosan_egerod_soren_word_order_word_classes_at>",
     "<dom_formosan_elizabeth_zeitoun_chen_huei_wu_overview>",
+    "<dom_formosan_elizabeth_zeitoun_lillian_huang_anna_ch>",
+    "<dom_formosan_elizabeth_zeitoun_lillian_huang_marie_y>",
+    "<dom_formosan_elizabeth_zeitoun_pronominal_system_man>",
+    "<dom_formosan_elizabeth_zeitoun_squib_dynamic_vs_stat>",
     "<dom_formosan_epark>",
     "<dom_formosan_ferrell_taiwan_aboriginal_groups_proble>",
+    "<dom_formosan_george_leroy_shelley_vudai_rukai>",
     "<dom_formosan_gitbook_translations>",
     "<dom_formosan_glosbe>",
-    "<dom_formosan_hala_saku_la_videos>",
+    "<dom_formosan_gujing_lin_tsou_serial_verb_constructio>",
+    "<dom_formosan_henry_chang_nominal_aspect_tsou>",
+    "<dom_formosan_henry_chang_tsou_causative_applicative>",
+    "<dom_formosan_henry_y_chang_formosan_speech_act_mood>",
+    "<dom_formosan_ho_dan_phonological_system_butonglu_pai>",
     "<dom_formosan_holmer_seediq>",
+    "<dom_formosan_hsiu_chuan_liao_transitivity_ergativity>",
+    "<dom_formosan_huang_dongqiu_atayal_bunun_amis_languag>",
     "<dom_formosan_huang_grammaticalization_squliq_atayal>",
+    "<dom_formosan_huang_hui_chuan_glide_formation_takitud>",
     "<dom_formosan_huang_mei_chin_atayal_reference_grammar>",
+    "<dom_formosan_hui_chuan_j_huang_syllable_types_in_bun>",
+    "<dom_formosan_hui_shan_lin_squliq_atayal_reduplicatio>",
     "<dom_formosan_huteson_rukai_survey>",
+    "<dom_formosan_ian_maddieson_and_richard_wright_the_vo>",
     "<dom_formosan_ilrdf_42_language_practice_word_lists>",
     "<dom_formosan_ilrdf_tousvusvutu_kita>",
+    "<dom_formosan_indigenous_education_resource_center_ke>",
     "<dom_formosan_indigenous_language_literary_awards>",
+    "<dom_formosan_isabelle_bril_roots_and_stems_lexical_a>",
+    "<dom_formosan_jian_iwan_guo_qingliu_life_history>",
+    "<dom_formosan_jian_shilang_introduction_thao_grammar>",
+    "<dom_formosan_john_tu_er_wei_notes_on_the_religious_b>",
+    "<dom_formosan_john_u_wolff_the_proto_austronesian_pho>",
+    "<dom_formosan_jonathan_kuo_amis_manner_result_verbs>",
+    "<dom_formosan_josiane_cauquelin_aborigines_taiwan_puy>",
+    "<dom_formosan_joy_wu_clausal_modifiers_in_amis>",
+    "<dom_formosan_joy_wu_verb_classification_case_marking>",
     "<dom_formosan_kanakanavu_texts>",
     "<dom_formosan_kavalan_zhang_reference_grammar>",
-    "<dom_formosan_li_peirong_xu_weisheng_introduction_tru>",
-    "<dom_formosan_li_peizhen_wu_xingyu_shouhu_xing>",
+    "<dom_formosan_kumu_tapas_tribal_memory_oral_history_w>",
+    "<dom_formosan_kuo_j_c_argument_alternation_and_argume>",
+    "<dom_formosan_lei_shih_family_system_paiwan_at_su_pai>",
+    "<dom_formosan_li_chen_yeh_developing_a_hla_alua_learn>",
+    "<dom_formosan_li_may_sung_budai_rukai_exclamatives>",
+    "<dom_formosan_li_may_sung_clausal_nominalization_buda>",
+    "<dom_formosan_lillian_huang_atayal_participants_cross>",
+    "<dom_formosan_lillian_huang_ergativity_atayal>",
+    "<dom_formosan_liu_manyi_kaskun_ata_matas_i_ki_quaz>",
+    "<dom_formosan_malcom_d_ross_the_history_and_transitiv>",
+    "<dom_formosan_malcom_ross_proto_austronesian_verbal_m>",
+    "<dom_formosan_man_ni_chu_the_measurement_of_final_i_v>",
+    "<dom_formosan_matsuzawa_kazuko_the_social_and_ritual>",
+    "<dom_formosan_maya_yeh_blaq_uv_construction_atayal>",
+    "<dom_formosan_mayumi_oiwa_adverbial_verb_construction>",
+    "<dom_formosan_mayumi_oiwa_bungard_morphology_and_synt>",
+    "<dom_formosan_mayumi_oiwa_possessor_raising_truku_see>",
+    "<dom_formosan_mei_chin_huang_verb_classification_in_m>",
+    "<dom_formosan_melissa_shih_hui_lin_sakizaya_or_amis_a>",
+    "<dom_formosan_nanang_tadaw_naci_mowna_pgagu>",
     "<dom_formosan_naomi_tsukida_correlative_clauses_seedi>",
+    "<dom_formosan_naomi_tsukida_ditransitive_causative_se>",
     "<dom_formosan_nowbucyang_truku_thesis>",
+    "<dom_formosan_ochiai_izumi_2016_comparative_linguisti>",
+    "<dom_formosan_ochiai_izumi_bu_hwan_vocabulary_recorde>",
+    "<dom_formosan_ochiai_izumi_numerals_paran_seediq_rela>",
+    "<dom_formosan_ochiai_izumi_on_a_seediq_glossary_recor>",
     "<dom_formosan_old_texts>",
-    "<dom_formosan_paiwan_collart_zeitoun_time_reference>",
     "<dom_formosan_paiwan_ho_five_dialects>",
     "<dom_formosan_paiwanstories>",
+    "<dom_formosan_pang_hsin_ting_reconstruction_proto_puy>",
+    "<dom_formosan_patricia_stanley_morphophonemics_of_ver>",
+    "<dom_formosan_paul_jen_kuei_li_comparative_bunun_dial>",
+    "<dom_formosan_paul_jen_kuei_li_conjunction_thao>",
+    "<dom_formosan_paul_jen_kuei_li_internal_relationships>",
+    "<dom_formosan_paul_jen_kuei_li_origins_of_the_east_fo>",
+    "<dom_formosan_paul_jen_kuei_li_review_of_taiwan_abori>",
+    "<dom_formosan_paul_jen_kuei_li_rukai_structure>",
+    "<dom_formosan_paul_jen_kuei_li_some_plant_names_in_fo>",
+    "<dom_formosan_paul_jen_kuei_li_the_preglottalised_sto>",
+    "<dom_formosan_paul_li_atayalic_final_voiced_stops>",
+    "<dom_formosan_paul_li_case_marking_four_formosan_lang>",
+    "<dom_formosan_paul_li_position_atayal_austronesian>",
+    "<dom_formosan_pingdong_county_government_kai_na_sepuc>",
     "<dom_formosan_puyuma_chen_raising_to_object>",
     "<dom_formosan_puyuma_katipol_kopfjagdriten>",
     "<dom_formosan_puyuma_teng_pronominal_systems>",
     "<dom_formosan_puyuma_teng_reference_grammar>",
+    "<dom_formosan_raleigh_ferrell_construction_markers_fo>",
+    "<dom_formosan_raleigh_ferrell_paiwan_phonology_and_pr>",
+    "<dom_formosan_raleigh_j_ferrell_intent_and_volition_i>",
     "<dom_formosan_rik_bunun>",
+    "<dom_formosan_rik_de_busser_the_influence_of_christia>",
+    "<dom_formosan_robert_blust_a_note_on_covert_structure>",
     "<dom_formosan_robert_blust_austronesian_homeland_ling>",
+    "<dom_formosan_robert_blust_some_remarks_linguistic_po>",
+    "<dom_formosan_robert_blust_thao_triplication>",
+    "<dom_formosan_robert_blust_the_austronesian_languages>",
     "<dom_formosan_rukai_mantauran_stories>",
     "<dom_formosan_rukai_texts>",
     "<dom_formosan_rukai_zeitoun_saying>",
-    "<dom_formosan_saaroa_pan_grammar>",
     "<dom_formosan_sakizaya_affixes>",
+    "<dom_formosan_schetelig_mittheilungen_uber_die_sprach>",
     "<dom_formosan_seals>",
     "<dom_formosan_seediq_zhang_reference_grammar>",
+    "<dom_formosan_shigeru_tsuchida_and_isidore_dyen_proto>",
     "<dom_formosan_shigeru_tsuchida_kanakanavu_texts>",
+    "<dom_formosan_shigeru_tsuchida_preliminary_reports_on>",
     "<dom_formosan_shigeru_tsuchida_reconstruction_proto_t>",
+    "<dom_formosan_shih_chi_stella_yeh_stress_and_quantity>",
     "<dom_formosan_shih_rukai_adverbial>",
-    "<dom_formosan_song_limei_introduction_kanakanavu_gram>",
-    "<dom_formosan_song_limei_introduction_seediq_grammar>",
+    "<dom_formosan_sihwei_chen_and_haowen_jiang_ways_of_ta>",
+    "<dom_formosan_sihwei_chen_lexical_aspect_atayal>",
+    "<dom_formosan_soren_egerod_statement_atayal_phonology>",
+    "<dom_formosan_soren_egerod_verb_inflection_atayal>",
+    "<dom_formosan_stacy_teng_malcom_ross_is_puyuma_primar>",
+    "<dom_formosan_stacy_wan_tin_huang_functions_yami_verb>",
     "<dom_formosan_sung_kuo_descriptive_comparative_constr>",
     "<dom_formosan_taoshan_elementary_school_atelier_hui_k>",
+    "<dom_formosan_teresa_m_chen_verbal_construction_and_v>",
+    "<dom_formosan_ting_chi_wei_parallelism_in_amis_sluici>",
+    "<dom_formosan_tingchun_chen_successive_cyclic_case_as>",
+    "<dom_formosan_truku_lowking_demonstratives>",
+    "<dom_formosan_tsai_hsui_liu_complementation_three_lan>",
+    "<dom_formosan_tsai_wei_tien_dylan_conjunctive_reducti>",
+    "<dom_formosan_tsou_descriptive_study>",
     "<dom_formosan_tsukida_naomi_verb_classification_amis>",
+    "<dom_formosan_tu_wen_chiu_a_synchronic_classification>",
+    "<dom_formosan_tung_tsuchida_ting_li_pan_saaroa_texts>",
+    "<dom_formosan_victoria_rau_yi_hsin_wu_and_meng_chien>",
+    "<dom_formosan_wei_huilin_liu_social_structure_yami>",
     "<dom_formosan_wei_matri_clan_lineage_system_ami>",
+    "<dom_formosan_wei_tien_tsai_subjecthood_temporal_adju>",
+    "<dom_formosan_wiedfelt_formosan_aborigines_atayal>",
     "<dom_formosan_wilang_yutas_videos>",
-    "<dom_formosan_wu_jinglan_introduction_amis_grammar>",
+    "<dom_formosan_wu_chunming_adverbials_in_paiwan>",
+    "<dom_formosan_xie_fuhui_introduction_kavalan_grammar>",
     "<dom_formosan_yedda_palemeq_blog>",
     "<dom_formosan_yeddas_blog>",
     "<dom_formosan_yeh_meili_introduction_saisiyat_grammar>",
+    "<dom_formosan_yi_ming_chou_object_control_saisiyat>",
+    "<dom_formosan_yi_ting_chen_a_minimalist_approach_to_a>",
+    "<dom_formosan_yi_ting_chen_amis_left_periphery>",
+    "<dom_formosan_yi_ting_chen_language_use_code_mixing_a>",
+    "<dom_formosan_yi_yang_cheng_kanakanavu_word_level_pro>",
+    "<dom_formosan_yi_yang_cheng_tense_aspect_agent_markin>",
+    "<dom_formosan_yoshiro_nihira_bunun_vocabulary>",
+    "<dom_formosan_yuehchien_chien_the_lexical_system_of_y>",
+    "<dom_formosan_yukihiro_yamada_ying_chu_liao_phonology>",
+    "<dom_formosan_zeng_shifen_reduplication_affixation_pa>",
+    "<dom_formosan_zhan_sujuan_taiwan_indigenous_peoples_h>",
+    "<dom_formosan_zhang_xiujuan_introduction_paiwan_gramm>",
     "<dom_formosan_zhang_yongli_pan_jiarong_introduction_t>",
+    "<dom_formosan_zhao_shanhe_li_taiyuan_zhu_qingyi_abori>",
     "<dom_formosan_zheng_acl_2024>",
     "<dom_formosan_zheng_data>",
+    "<dom_formosan_zheng_yumei_zhu_qingyi_bo_hongming_abor>",
     "<dom_learning_vocab>",
     "<dom_nine_level>",
     "<dom_ntu>",

training_profile.json

@@ -0,0 +1,94 @@
+{
+  "schema_version": 3,
+  "recipe_id": "nllb200-spm8k-directional-v3",
+  "model_family": "nllb",
+  "corpus_pipeline_version": "formosan-mt-corpus-v3",
+  "mt_standardization": {
+    "id": "formosan-mt-standard-v3",
+    "sha256": "4bbded87eb4833b2d1cd5a88a1f2b059560416e0e69ecfa55433a77f418c903e",
+    "namespace": "formosan-mt"
+  },
+  "project_root": ".",
+  "corpus_variants": ["public_no_bible", "private_no_bible"],
+  "base_model": {
+    "name": "facebook/nllb-200-distilled-600M",
+    "revision": "f8d333a098d19b4fd9a8b18f94170487ad3f821d"
+  },
+  "primary_input_pattern": "corpus_builds/{corpus_name}/pivot_corpora_final/big_corpus_{target}_in_domain_hard.csv",
+  "targets": {
+    "english": {
+      "input_pattern": "corpus_builds/{corpus_name}/pivot_corpora_final/big_corpus_en_in_domain_hard.csv",
+      "target_col": "english_sentence",
+      "directions": ["f2en", "en2f"],
+      "selected_recipe": "NLLB-200 SPM8k directional"
+    },
+    "chinese": {
+      "input_pattern": "corpus_builds/{corpus_name}/pivot_corpora_final/big_corpus_zh_in_domain_hard.csv",
+      "target_col": "chinese_sentence",
+      "directions": ["f2zh", "zh2f"],
+      "selected_recipe": "NLLB-200 SPM8k directional"
+    }
+  },
+  "splits": {
+    "english_output_dir": "formosan_mt_experiments/data/splits_en_v1",
+    "chinese_output_dir": "formosan_mt_experiments/data/splits_zh_v1",
+    "train_ratio": 0.9,
+    "validate_ratio": 0.025,
+    "test_ratio": 0.075,
+    "min_test_rows": 500,
+    "min_validate_rows": 150,
+    "character_ngram_jaccard_threshold": 0.82,
+    "tiers": ["in_domain_hard"],
+    "headline_tier": "in_domain_hard"
+  },
+  "tokenizer": {
+    "mode": "spm",
+    "spm_vocab_sweep": [8192],
+    "default_spm_vocab": 8192,
+    "chinese_default_spm_vocab": 8192,
+    "min_char_frequency": 3,
+    "max_dialect_tags": 200,
+    "min_dialect_frequency": 3,
+    "training_columns": ["formosan_sentence"],
+    "setup_splits": ["train"]
+  },
+  "training_defaults": {
+    "steps": 300000,
+    "batch_size": 16,
+    "grad_accum_steps": 4,
+    "effective_batch_size": 64,
+    "max_length": 384,
+    "learning_rate": 0.00002,
+    "warmup_steps": 4000,
+    "weight_decay": 0.001,
+    "alpha": 0.5,
+    "precision": "bf16",
+    "label_smoothing": 0.1,
+    "save_interval": 0,
+    "log_interval": 500,
+    "generation_eval_interval": 10000,
+    "generation_eval_samples_per_language": 128,
+    "generation_eval_batch_size": 16,
+    "validation_beam": 2,
+    "best_metric": "chrF2",
+    "early_stopping_patience": 5,
+    "early_stopping_min_delta": 0.05,
+    "early_stopping_start_step": 30000,
+    "f2en_easy_source_weight": 0.05,
+    "en2f_easy_source_weight": 0.15,
+    "f2zh_easy_source_weight": 0.05,
+    "zh2f_easy_source_weight": 0.15
+  },
+  "generation_defaults": {
+    "beam": 4,
+    "max_length": 384,
+    "max_new_tokens": 256,
+    "min_new_tokens": 1,
+    "no_repeat_ngram_size": 0,
+    "repetition_penalty": 1.0,
+    "length_penalty": 1.0,
+    "metadata_modes": ["default", "oracle"],
+    "bootstrap_samples": 200,
+    "bootstrap_seed": 42
+  }
+}