|
|
--- |
|
|
language: multilingual |
|
|
tags: |
|
|
- pos-tagging |
|
|
- afro-xlmr |
|
|
- token-classification |
|
|
license: mit |
|
|
--- |
|
|
|
|
|
|
|
|
|
|
|
This model was fine-tuned on part-of-speech tagging using the `Davlan/afro-xlmr-mini` model. |
|
|
It supports token classification for low-resource African languages. |
|
|
|
|
|
|
|
|
|
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForTokenClassification |
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained("your-username/afroxlmr-pos") |
|
|
model = AutoModelForTokenClassification.from_pretrained("your-username/afroxlmr-pos") |
|
|
|
|
|
|
|
|
We have fine tune the pretrained model with another architechture and we had obtained a higher precision , recall and F1 score. For the training and dev data we have split our own data into train and dev and we have use their test set to get the final performance. |
|
|
|
|
|
|
|
|
|
|
|
Our own Data : https://github.com/hausanlp/HERDPhobia |
|
|
Their Data : https://github.com/masakhane-io/masakhane-pos |
