Omballa's picture
Upload README.md with huggingface_hub
ae10e58 verified
|
Raw
History Blame Contribute Delete
1.45 kB
metadata
language: luo
license: apache-2.0
tags:
  - pos-tagging
  - token-classification
  - dholuo
  - afroxlmr
  - masakhane
datasets:
  - kencorpus
model-index:
  - name: Dholuo POS Tagger (AfroXLM-R)
    results:
      - task:
          type: token-classification
          name: POS Tagging
        dataset:
          name: KenCorpus (Dholuo)
          type: kencorpus
          split: test
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.7167
          - name: Macro F1
            type: f1
            value: 0.5338
          - name: Weighted F1
            type: f1
            value: 0.7248

Dholuo POS Tagger (AfroXLM-R)

This model is a fine-tuned version of masakhane/luo-pos-tagger-afroxlmr trained on KenCorpus POS-tagged Dholuo data.

Evaluation Results

The model was evaluated on the KenCorpus Dholuo POS test set.

Metric Value
Accuracy 0.7167
Macro F1 0.5338
Weighted F1 0.7248

Per-tag Performance (Test Set)

  • Strong performance on NOUN (F1 ≈ 0.86) and VERB (F1 ≈ 0.84)
  • Moderate performance on function words (ADP, PRON, DET)
  • Rare tags (INTJ, PART, PUNCT) have near-zero scores due to very low support

Usage

from transformers import AutoTokenizer, AutoModelForTokenClassification

model = AutoModelForTokenClassification.from_pretrained(
    "Omballa/dholuo-pos-afroxlmr-finetuned"
)
tokenizer = AutoTokenizer.from_pretrained(
    "Omballa/dholuo-pos-afroxlmr-finetuned"
)