Hugging Face's logo
language: yo datasets:
xlm-roberta-base-finetuned-yoruba
Model description
xlm-roberta-base-finetuned-yoruba is a Yoruba RoBERTa model obtained by fine-tuning xlm-roberta-base model on Yorùbá language texts. It provides better performance than the XLM-RoBERTa on text classification and named entity recognition datasets.
Specifically, this model is a xlm-roberta-base model that was fine-tuned on Yorùbá corpus.
Intended uses & limitations
How to use
You can use this model with Transformers pipeline for masked token prediction.
>>> from transformers import pipeline
>>> unmasker = pipeline('fill-mask', model='Davlan/xlm-roberta-base-finetuned-yoruba')
>>> unmasker("Arẹmọ Phillip to jẹ ọkọ <mask> Elizabeth to ti wa lori aisan ti dagbere faye lẹni ọdun mọkandilọgọrun")
[{'sequence': '<s> Arẹmọ Phillip to jẹ ọkọ Queen Elizabeth to ti wa lori aisan ti dagbere faye lẹni ọdun mọkandilọgọrun</s>', 'score': 0.24844281375408173,
'token': 44109,
'token_str': '▁Queen'},
{'sequence': '<s> Arẹmọ Phillip to jẹ ọkọ ile Elizabeth to ti wa lori aisan ti dagbere faye lẹni ọdun mọkandilọgọrun</s>', 'score': 0.1665010154247284,
'token': 1350,
'token_str': '▁ile'},
{'sequence': '<s> Arẹmọ Phillip to jẹ ọkọ ti Elizabeth to ti wa lori aisan ti dagbere faye lẹni ọdun mọkandilọgọrun</s>', 'score': 0.07604238390922546,
'token': 1053,
'token_str': '▁ti'},
{'sequence': '<s> Arẹmọ Phillip to jẹ ọkọ baba Elizabeth to ti wa lori aisan ti dagbere faye lẹni ọdun mọkandilọgọrun</s>', 'score': 0.06353845447301865,
'token': 12878,
'token_str': '▁baba'},
{'sequence': '<s> Arẹmọ Phillip to jẹ ọkọ Oba Elizabeth to ti wa lori aisan ti dagbere faye lẹni ọdun mọkandilọgọrun</s>', 'score': 0.03836742788553238,
'token': 82879,
'token_str': '▁Oba'}]
Limitations and bias
This model is limited by its training dataset of entity-annotated news articles from a specific span of time. This may not generalize well for all use cases in different domains.
Training data
This model was fine-tuned on Bible, JW300, Menyo-20k, Yoruba Embedding corpus and CC-Aligned, Wikipedia, news corpora (BBC Yoruba, VON Yoruba, Asejere, Alaroye), and other small datasets curated from friends.
Training procedure
This model was trained on a single NVIDIA V100 GPU
Eval results on Test set (F-score, average over 5 runs)
Dataset | XLM-R F1 | yo_roberta F1 |
---|---|---|
MasakhaNER | 77.58 | 83.66 |
BBC Yorùbá Textclass |
BibTeX entry and citation info
By David Adelani
- Downloads last month
- 4