--- pretty_name: MorphBench (English) language: - en license: cc-by-sa-4.0 tags: - morphology - tokenization - evaluation - english dataset_info: - config_name: task1_inflection features: - name: lemma dtype: string - name: feats dtype: string - name: target dtype: string - name: infl_type dtype: string - name: status dtype: string splits: - name: train num_bytes: 674115 num_examples: 9307 - name: dev num_bytes: 35771 num_examples: 495 - name: test num_bytes: 76239 num_examples: 1049 - name: test_rare num_bytes: 39673 num_examples: 499 - name: test_memorization num_bytes: 30661 num_examples: 436 download_size: 240727 dataset_size: 856459 - config_name: task2_segmentation features: - name: word dtype: string - name: segmentation dtype: string - name: deriv_type dtype: string - name: affix_position dtype: string - name: status dtype: string splits: - name: train num_bytes: 730103 num_examples: 9094 - name: dev num_bytes: 74367 num_examples: 918 - name: test_main num_bytes: 134223 num_examples: 1735 - name: test_rare num_bytes: 61086 num_examples: 734 - name: test_memorization num_bytes: 10014 num_examples: 137 download_size: 336942 dataset_size: 1009793 - config_name: task3a_derivation features: - name: base dtype: string - name: affix dtype: string - name: target dtype: string - name: deriv_type dtype: string - name: affix_position dtype: string - name: status dtype: string splits: - name: train num_bytes: 757459 num_examples: 9095 - name: dev num_bytes: 77121 num_examples: 918 - name: test_main num_bytes: 139428 num_examples: 1735 - name: test_rare num_bytes: 63288 num_examples: 734 - name: test_memorization num_bytes: 10425 num_examples: 137 download_size: 329387 dataset_size: 1047721 - config_name: task3b_derivation_mcq features: - name: query_base dtype: string - name: query_derived dtype: string - name: query_affix dtype: string - name: query_pos dtype: string - name: affix_subset dtype: string - name: pretrain_cell dtype: string - name: options struct: - name: A dtype: string - name: B dtype: string - name: C dtype: string - name: D dtype: string - name: correct dtype: string - name: demos list: - name: base dtype: string - name: derived dtype: string splits: - name: dev num_bytes: 142712 num_examples: 495 - name: test num_bytes: 549737 num_examples: 1905 download_size: 399699 dataset_size: 692449 - config_name: task4a_affix_function features: - name: word dtype: string - name: function dtype: string - name: freq dtype: int64 splits: - name: train num_bytes: 290974 num_examples: 7800 - name: dev num_bytes: 35388 num_examples: 946 - name: test num_bytes: 66505 num_examples: 1766 download_size: 136209 dataset_size: 392867 - config_name: task4b_affix_function_paraphrase features: - name: meaning dtype: string - name: function dtype: string - name: freq dtype: int64 splits: - name: train num_bytes: 10766 num_examples: 200 - name: dev num_bytes: 11672 num_examples: 227 - name: test num_bytes: 21099 num_examples: 407 download_size: 22942 dataset_size: 43537 - config_name: task5a_definition features: - name: word dtype: string - name: gloss dtype: string - name: base dtype: string - name: affix dtype: string - name: function dtype: string - name: derived_freq dtype: int64 - name: base_exposure dtype: int64 - name: status dtype: string splits: - name: train num_bytes: 824645 num_examples: 6651 - name: dev num_bytes: 117667 num_examples: 973 - name: test_main num_bytes: 49381 num_examples: 428 - name: test_main_oov num_bytes: 99786 num_examples: 770 - name: test_rare num_bytes: 47530 num_examples: 387 - name: test_memorization num_bytes: 16352 num_examples: 118 download_size: 554488 dataset_size: 1155361 - config_name: task5b_definition_mcq features: - name: word dtype: string - name: base dtype: string - name: affix dtype: string - name: function dtype: string - name: gold_gloss dtype: string - name: distractors sequence: string - name: pretrain_status dtype: string - name: derived_freq dtype: int64 - name: base_exposure dtype: int64 splits: - name: dev num_bytes: 357700 num_examples: 972 - name: main num_bytes: 155783 num_examples: 428 - name: main_oov num_bytes: 296225 num_examples: 769 - name: rare num_bytes: 143704 num_examples: 387 - name: memorization num_bytes: 44107 num_examples: 118 download_size: 461392 dataset_size: 997519 configs: - config_name: task1_inflection data_files: - split: train path: task1_inflection/train-* - split: dev path: task1_inflection/dev-* - split: test path: task1_inflection/test-* - split: test_rare path: task1_inflection/test_rare-* - split: test_memorization path: task1_inflection/test_memorization-* - config_name: task2_segmentation data_files: - split: train path: task2_segmentation/train-* - split: dev path: task2_segmentation/dev-* - split: test_main path: task2_segmentation/test_main-* - split: test_rare path: task2_segmentation/test_rare-* - split: test_memorization path: task2_segmentation/test_memorization-* - config_name: task3a_derivation data_files: - split: train path: task3a_derivation/train-* - split: dev path: task3a_derivation/dev-* - split: test_main path: task3a_derivation/test_main-* - split: test_rare path: task3a_derivation/test_rare-* - split: test_memorization path: task3a_derivation/test_memorization-* - config_name: task3b_derivation_mcq data_files: - split: dev path: task3b_derivation_mcq/dev-* - split: test path: task3b_derivation_mcq/test-* - config_name: task4a_affix_function data_files: - split: train path: task4a_affix_function/train-* - split: dev path: task4a_affix_function/dev-* - split: test path: task4a_affix_function/test-* - config_name: task4b_affix_function_paraphrase data_files: - split: train path: task4b_affix_function_paraphrase/train-* - split: dev path: task4b_affix_function_paraphrase/dev-* - split: test path: task4b_affix_function_paraphrase/test-* - config_name: task5a_definition data_files: - split: train path: task5a_definition/train-* - split: dev path: task5a_definition/dev-* - split: test_main path: task5a_definition/test_main-* - split: test_main_oov path: task5a_definition/test_main_oov-* - split: test_rare path: task5a_definition/test_rare-* - split: test_memorization path: task5a_definition/test_memorization-* - config_name: task5b_definition_mcq data_files: - split: dev path: task5b_definition_mcq/dev-* - split: main path: task5b_definition_mcq/main-* - split: main_oov path: task5b_definition_mcq/main_oov-* - split: rare path: task5b_definition_mcq/rare-* - split: memorization path: task5b_definition_mcq/memorization-* - config_name: verb_cloze data_files: - split: train path: verb_cloze/train.jsonl - split: validation path: verb_cloze/validation.jsonl - split: test path: verb_cloze/test.jsonl --- # MorphBench — English Morphology Evaluation Suite A benchmark for **morphology-aware tokenizers and language models** in English. Eight numbered tasks probe inflection, segmentation, derivation, affix semantics, and whole-word meaning. Each task is a **config**; difficulty/generalization tiers are **splits**. Related tasks are paired (`a` = generation / from word, `b` = recognition / from meaning). ```python from datasets import load_dataset ds = load_dataset("yuanxin112/morphbench-en", "task3a_derivation", split="test_main") ``` ## Tasks ### `task1_inflection` Produce the inflected form from a lemma + UniMorph features. **input** `goggle` + `V;V.PTCP;PST` → **output** `goggled` columns: `lemma, feats, target, infl_type, status` ### `task2_segmentation` Split a word into its morphemes. **input** `creekline` → **output** `creek line` (space-separated morphemes) columns: `word, segmentation, deriv_type, affix_position, status` ### `task3a_derivation` — *generation* Build the derived word from a base + affix. **input** `pledge` + `in-` → **output** `impledge` columns: `base, affix, target, deriv_type, affix_position, status` ### `task3b_derivation_mcq` — *recognition* Multiple-choice version of 3a: pick the correct derived form from 4 options (with demo pairs of the same affix). Easier recognition counterpart to 3a's free generation. columns: `query_base, query_derived, query_affix, query_pos, affix_subset, pretrain_cell, options, correct, demos` ### `task4a_affix_function` — *from the word* Given a derived word, classify the **semantic function of its affix** (13 classes: `negation, without, agent_person, make_become, repetition_again, …`). **input** `knackless` → **output** `without` columns: `word, function, freq` ### `task4b_affix_function_paraphrase` — *from the meaning* The same 13-way classification, but the input is a **paraphrased meaning** instead of the word, so the surface affix string is removed. This is the **control** for 4a — it tests whether the model understands the affix's *meaning* rather than reading its spelling. **input** `"lacking practical skill"` → **output** `without` columns: `meaning, function, freq` > Models score ~98% on 4a but collapse to ~22% on 4b — i.e. 4a is solved almost entirely by the surface affix cue. ### `task5a_definition` — *generation* Generate the dictionary gloss of a derived word. **input** `steadfastness` → **output** `Loyalty in the face of trouble and difficulty` columns: `word, gloss, base, affix, function, derived_freq, base_exposure, status` ### `task5b_definition_mcq` — *recognition* Multiple-choice version of 5a: pick the correct gloss among hard distractors. columns: `word, base, affix, function, gold_gloss, distractors, pretrain_status, derived_freq, base_exposure` ## Task pairs | Pair | `a` | `b` | |------|-----|-----| | 3 | derivation — **generate** the form | derivation — **recognize** (4-choice) | | 4 | affix function — **from the word** | affix function — **from meaning only** (surface-cue control) | | 5 | definition — **generate** the gloss | definition — **recognize** (4-choice) | ## Splits — how items are bucketed Every test split is defined by the **pretraining exposure** of the item's parts (frequency in the BabyLM training corpus): *seen* = frequency above a threshold, *unseen* = frequency 0. The headline idea is the same everywhere — **the target is UNSEEN while its base/lemma IS seen** (compositional generalization); `memorization` and `rare` are the seen / hardest controls. `train` and `dev` are always training / validation, and splits are lemma/base-disjoint from `train`. **`task1_inflection`** — by *(lemma seen?)* × *(this inflected form seen?)*: | split | condition | what it tests | |-------|-----------|---------------| | `test` (main) | lemma seen, form **unseen** | inflect a known lemma into an unseen form | | `test_memorization` | lemma seen, form seen | recall a seen form | | `test_rare` | lemma **unseen**, form unseen | hardest — neither seen | **`task2_segmentation`, `task3a_derivation`** — by *(base seen?)* × *(derived word seen?)*, where *seen* = corpus freq ≥ 10: | split | condition | what it tests | |-------|-----------|---------------| | `test_main` | base seen, derived **unseen** | compositional generalization | | `test_memorization` | base seen, derived seen | memorization | | `test_rare` | base **unseen**, derived unseen | rare / OOV | (The `base_unseen + derived_seen` cell is dropped as too rare.) **`task5a_definition`, `task5b_definition_mcq`** — by the derived word's corpus frequency `derived_freq`, splitting the *unseen* case further by `base_exposure` (how attested the base is): | split | condition | what it tests | |-------|-----------|---------------| | `memorization` | `derived_freq ≥ 50` | recall the gloss of a frequent word | | `rare` | `derived_freq` 1–5 | word seen only a few times | | `main` | `derived_freq = 0` **and** `base_exposure ≥ 500` | define an **unseen** derived word from a **familiar** base | | `main_oov` | `derived_freq = 0` **and** `base_exposure < 500` | hardest — derived word unseen **and** base barely seen | **`task3b_derivation_mcq`** uses `dev` / `test`; **`task4a_affix_function`** and **`task4b_affix_function_paraphrase`** use `train` / `dev` / `test` (no difficulty buckets — each row still carries an exposure `status` where applicable). The `status` / `pretrain_cell` column on each row records the exposure cell directly, e.g. `base_seen+derived_unseen`, `lem_seen+form_unseen`. ## Provenance Built from English Wiktionary (via [kaikki.org](https://kaikki.org) / wiktextract) and UniMorph. Companion lexical resource: [`yuanxin112/wiktionary-morph`](https://huggingface.co/datasets/yuanxin112/wiktionary-morph). German counterpart: [`yuanxin112/morphbench-de`](https://huggingface.co/datasets/yuanxin112/morphbench-de). ## License Derived from Wiktionary — [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/); attribute Wiktionary and its contributors. ## `verb_cloze` — Contextual Verb-Inflection Cloze (added config) Leave-one-out cloze: given a verb `lemma`, partial `feats` (one morphological dimension withheld), and a natural sentence with the target verb replaced by a blank marker, generate the inflected form. Splits are **lemma-disjoint**. Sentences come from Universal Dependencies treebanks (a *derivative*; CC BY-SA 4.0 — verify per-treebank licenses, e.g. English ParTUT is CC BY-NC-SA). ```python from datasets import load_dataset ds = load_dataset("yuanxin112/morphbench-en", "verb_cloze") ```