Romansh Idiom Classifier (XLM-RoBERTa Large)
Fine-tuned xlm-roberta-large for 6-class Romansh idiom classification on newspaper-article chunks.
Dataset format
JSONL, one sample per line:
{"text":"Quegl e par igl mument betga pussevel an nossas scolas.","label":"sutsilvan","doc_id":"sutsilvan_doc002505_chunk0004"}
- Downloads last month
- 29
Model tree for nomichaelno/xlm-roberta-large-romansh-idiom-classifier
Base model
FacebookAI/xlm-roberta-largeEvaluation results
- Accuracy on Newspaper corpus (JSONL, doc-level split)Transformers Trainer (doc-level split)0.978
- Macro F1 on Newspaper corpus (JSONL, doc-level split)Transformers Trainer (doc-level split)0.959