|
|
--- |
|
|
language: |
|
|
- en |
|
|
- bn |
|
|
tags: |
|
|
- sentiment-analysis |
|
|
- cross-lingual |
|
|
- xlm-roberta |
|
|
- text-classification |
|
|
datasets: |
|
|
- glue |
|
|
- sepidmnorozy/Bengali_sentiment |
|
|
metrics: |
|
|
- accuracy |
|
|
- f1 |
|
|
library_name: transformers |
|
|
pipeline_tag: text-classification |
|
|
--- |
|
|
|
|
|
# arif481/crosslingual-sentiment-model |
|
|
|
|
|
A cross-lingual sentiment analysis model fine-tuned on XLM-RoBERTa for binary sentiment classification (positive/negative) across en, bn. |
|
|
|
|
|
## Model Description |
|
|
|
|
|
This model performs sentiment classification across multiple languages using transfer learning. It was trained using the **combined** strategy. |
|
|
|
|
|
### Supported Languages |
|
|
- English (en) |
|
|
- Bengali (bn) |
|
|
|
|
|
### Training Mode: combined |
|
|
Trained on combined English and Bengali data for multilingual learning. |
|
|
|
|
|
## Usage |
|
|
|
|
|
```python |
|
|
from transformers import pipeline |
|
|
|
|
|
classifier = pipeline("sentiment-analysis", model="arif481/crosslingual-sentiment-model") |
|
|
|
|
|
# English |
|
|
result = classifier("This movie is absolutely fantastic!") |
|
|
print(result) # [{'label': 'positive', 'score': 0.99}] |
|
|
|
|
|
# Bengali |
|
|
result = classifier("এই সিনেমাটি অসাধারণ ছিল!") |
|
|
print(result) # [{'label': 'positive', 'score': 0.95}] |
|
|
``` |
|
|
|
|
|
## Training |
|
|
|
|
|
```python |
|
|
from transformers import AutoModelForSequenceClassification, AutoTokenizer |
|
|
|
|
|
model = AutoModelForSequenceClassification.from_pretrained("arif481/crosslingual-sentiment-model") |
|
|
tokenizer = AutoTokenizer.from_pretrained("arif481/crosslingual-sentiment-model") |
|
|
``` |
|
|
|
|
|
|
|
|
## Metrics |
|
|
|
|
|
| Metric | Value | |
|
|
|--------|-------| |
|
|
| Accuracy | N/A | |
|
|
| Macro F1 | N/A | |
|
|
| Precision | N/A | |
|
|
| Recall | N/A | |
|
|
|
|
|
|
|
|
## Limitations |
|
|
|
|
|
- Binary classification only (positive/negative) |
|
|
- May not perform well on neutral sentiment |
|
|
- Bengali performance may be lower than English due to limited training data |
|
|
|
|
|
## Citation |
|
|
|
|
|
If you use this model, please cite: |
|
|
|
|
|
```bibtex |
|
|
@misc{crosslingual-sentiment, |
|
|
author = {Cross-Lingual Sentiment Team}, |
|
|
title = {Cross-Lingual Sentiment Analysis Model}, |
|
|
year = {2024}, |
|
|
publisher = {Hugging Face}, |
|
|
url = {https://huggingface.co/arif481/crosslingual-sentiment-model} |
|
|
} |
|
|
``` |
|
|
|
|
|
## License |
|
|
|
|
|
This model is released under the MIT License. |
|
|
|