lmompelat/mbert-martinican-pos-tagger

CreoleEval: Martinican Creole POS Tagger

Model type: AutoModelForTokenClassification
Task: Universal Part-Of-Speech Tagging (UPOS)
Language: Martinican Creole (Kréyol Matnik)
License: CC BY 4.0

Model Description

This model is an mBert bert-base-multilingual-cased checkpoint fine-tuned on the Martinican Creole treebank for UPOS tagging. It takes tokenized text and predicts one of the 17 universal part-of-speech tags for each token.

Accuracy: 0.88

First Version: 4/22/2025 - Last Updated on 4/25/2025

---

Intended Use

• Primary use: Automatic UPOS tagging of Martinican Creole text.
• Not for: Named-entity recognition, dependency parsing, or languages other than Martinican Creole without further fine-tuning.

from transformers import pipeline

tagger = pipeline(
    "token-classification",
    model="lmompelat/mbert-martinican-pos-tagger",
    tokenizer="lmompelat/mbert-martinican-pos-tagger"
)

example = "Mwen té ka alé an lékol."
predictions = tagger(example)
print(predictions)

---

Citation

Author: Ludovic Mompelat

Email: [email protected]