Vietnamese Sentiment Analysis with PhoBERT
Fine-tuned VinAI/PhoBERT-base for 3-class sentiment classification on Vietnamese product reviews.
Model Description
This model classifies Vietnamese text into three sentiment categories:
- Positive (tích cực)
- Neutral (trung lập)
- Negative (tiêu cực)
Training Data
- 10,000+ Vietnamese product reviews
- Balanced across 3 sentiment classes
- Preprocessed with Vietnamese word segmentation
Performance
| Metric | Score |
|---|---|
| Accuracy | 85% |
| F1-Score (weighted) | 0.84 |
Usage
from transformers import pipeline
classifier = pipeline("text-classification", model="sanvo/vietnamese-sentiment-phobert")
result = classifier("Sản phẩm rất tốt, tôi rất hài lòng")
print(result)
Training Details
- Base model: vinai/phobert-base
- Epochs: 5
- Batch size: 16
- Learning rate: 2e-5
- Max length: 256 tokens
- Optimizer: AdamW with linear warmup