import torch
from transformers import AutoTokenizer, AutoModel

# Load model and tokenizer
repo = "visolex/bilstm-absa-smartphone"
tokenizer = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
model = AutoModel.from_pretrained(repo, trust_remote_code=True)
model.eval()

# Aspect labels for ViSFD
aspect_labels = [
    "BATTERY",
    "CAMERA",
    "DESIGN",
    "FEATURES",
    "GENERAL",
    "PERFORMANCE",
    "PRICE",
    "SCREEN",
    "SER&ACC",
    "STORAGE"
]

# Sentiment labels
sentiment_labels = ["POSITIVE", "NEGATIVE", "NEUTRAL"]

# Example review text
text = "Pin rất tốt, camera đẹp nhưng giá hơi cao."

# Tokenize
inputs = tokenizer(
    text,
    return_tensors="pt",
    padding=True,
    truncation=True,
    max_length=256
)
inputs.pop("token_type_ids", None)

# Predict
with torch.no_grad():
    outputs = model(**inputs)

# Get logits: shape [1, num_aspects, num_sentiments + 1]
logits = outputs.logits.squeeze(0)  # [num_aspects, num_sentiments + 1]
probs = torch.softmax(logits, dim=-1)

# Predict for each aspect
none_id = probs.size(-1) - 1  # Index of "none" class
results = []

for i, aspect in enumerate(aspect_labels):
    prob_i = probs[i]
    pred_id = int(prob_i.argmax().item())
    
    if pred_id != none_id and pred_id < len(sentiment_labels):
        score = prob_i[pred_id].item()
        if score >= 0.5:  # threshold
            results.append((aspect, sentiment_labels[pred_id].lower()))

print(f"Text: {text}")
print(f"Predicted aspects: {results}")
# Output example: [('aspects', 'positive'), ('aspects', 'positive'), ('aspects', 'negative')]

Citation

If you use this model, please cite:

bibtex

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

License

This model is released under the Apache-2.0 license.

Acknowledgments

Base model: bilstm
Dataset: ViSFD
ViSoLex Toolkit