from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("Khoa/vietnamese-sentiment-analysis-with-entity")
model = AutoModelForSequenceClassification.from_pretrained("Khoa/vietnamese-sentiment-analysis-with-entity")

# Function to predict sentiment
def predict_sentiment(text, entity, model, tokenizer):
    combined_text = f"Đối với {entity}, {text}"
    inputs = tokenizer(combined_text, return_tensors="pt", truncation=True, padding=True)

    with torch.no_grad():
        outputs = model(**inputs)
        predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
        predicted_class = torch.argmax(predictions, dim=-1).item()

    sentiment_labels = {0: "NEGATIVE", 1: "NEUTRAL", 2: "POSITIVE"}
    return sentiment_labels[predicted_class], predictions[0].tolist()

# Example usage
text = "Món ăn rất ngon nhưng giá hơi đắt"
entity = "Nhà hàng ABC"
sentiment, confidence = predict_sentiment(text, entity, model, tokenizer)
print(f"Sentiment: {sentiment}")

Training procedure

The model was fine-tuned on a custom Vietnamese dataset with entity-specific sentiment annotations.

This model was fine-tuned on a custom Vietnamese sentiment analysis dataset. It achieves the following metrics on the test set:

Accuracy: 0.0000
F1 Score: 0.0000
Precision: 0.0000
Recall: 0.0000