FinSentinel — French Financial Sentiment Analysis
CamemBERT fine-tuned for financial sentiment classification.
Model Description
This model classifies financial text into 3 categories:
- Bearish (negative sentiment)
- Bullish (positive sentiment)
- Neutral
Fine-tuned from camembert-base on 12,816 financial sentences.
Training Results
| Epoch | Accuracy | F1 Score |
|---|---|---|
| 1 | 70.67% | 0.641 |
| 2 | 76.83% | 0.756 |
| 3 | 78.00% | 0.772 |
Test Results
| Metric | Score |
|---|---|
| Accuracy | 78.08% |
| F1 Score | 0.7759 |
Per-class Results
| Class | Precision | Recall | F1 |
|---|---|---|---|
| Bearish | 0.60 | 0.52 | 0.55 |
| Bullish | 0.69 | 0.64 | 0.66 |
| Neutral | 0.84 | 0.89 | 0.87 |
Training Details
- Base model: camembert-base
- Epochs: 3
- Batch size: 32
- Learning rate: 2e-5
- Max length: 128
- Train samples: 10,252
- Val samples: 1,282
- Test samples: 1,282
Usage
from transformers import pipeline
classifier = pipeline(
"text-classification",
model="Walid692/finsentinel-camembert"
)
result = classifier("La hausse des taux d'interet inquiete les investisseurs institutionnels.")
print(result)
Author
Ahmed Walid BOUANZOUL — AI Engineer GitHub: https://github.com/AhmedWalidbou/FinSentinel
- Downloads last month
- 51