visolex / bilstm-absa-restaurant

bilstm-absa-restaurant: Aspect-based Sentiment Analysis for Vietnamese Reviews

This model is a fine-tuned version of bilstm on the VLSP2018-ABSA-Restaurant dataset for aspect-based sentiment analysis in Vietnamese reviews.

Model Details

• Base Model : bilstm
• Description : BiLSTM for Vietnamese ABSA
• Dataset : VLSP2018-ABSA-Restaurant
• Fine-tuning Framework : HuggingFace Transformers
• Task : Aspect-based Sentiment Classification (3 classes)

Hyperparameters

• Batch size: 32
• Learning rate: 3e-5
• Epochs: 100
• Max sequence length: 256
• Weight decay: 0.01
• Warmup steps: 500
• Optimizer: AdamW

Dataset

Model was trained on VLSP2018 ABSA Restaurant dataset for aspect-based sentiment analysis.

Sentiment Labels:

• 0 - Negative (Tiêu cực): Negative opinions
• 1 - Neutral (Trung lập): Neutral, objective opinions
• 2 - Positive (Tích cực): Positive opinions

Aspect Categories:

Model được train để phân tích sentiment cho các aspects sau:

• AMBIENCE#GENERAL
• DRINKS#PRICES
• DRINKS#QUALITY
• DRINKS#STYLE&OPTIONS
• FOOD#PRICES
• FOOD#QUALITY
• FOOD#STYLE&OPTIONS
• LOCATION#GENERAL
• RESTAURANT#GENERAL
• RESTAURANT#MISCELLANEOUS
• RESTAURANT#PRICES
• SERVICE#GENERAL

Evaluation Results

The model was evaluated on test set with the following metrics:

• Accuracy : 0.8277
• Macro-F1 : 0.2890
• Weighted-F1 : 0.5580
• Macro-Precision : 0.4477
• Macro-Recall : 0.2342

Usage Example

from transformers import AutoTokenizer, AutoModelForSequenceClassification

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("visolex/bilstm-absa-restaurant")
model = AutoModelForSequenceClassification.from_pretrained("visolex/bilstm-absa-restaurant")

# Example review
text = "Khách sạn rất sạch sẽ, dịch vụ tuyệt vời nhưng giá hơi cao."

# Tokenize
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=256)

# Predict
outputs = model(**inputs)
predicted_sentiment = outputs.logits.argmax(dim=-1).item()

# Map to sentiment name
sentiment_map = {
    0: "Negative",
    1: "Neutral",
    2: "Positive"
}

predicted_name = sentiment_map[predicted_sentiment]
print(f"Text: {text}")
print(f"Predicted sentiment: {predicted_name}")

Citation

If you use this model, please cite:

@misc{visolex_absa_bilstm_absa_restaurant,
  title={BiLSTM for Vietnamese ABSA for Vietnamese Aspect-based Sentiment Analysis},
  author={ViSoLex Team},
  year={2025},
  url={https://huggingface.co/visolex/bilstm-absa-restaurant}
}

License

This model is released under the Apache-2.0 license.

Acknowledgments

• Base model: bilstm
• Dataset: VLSP2018-ABSA-Restaurant
• ViSoLex Toolkit