AventIQ-AI/bert-movie-review-sentiment-analysis

🎬 Movie Review Sentiment Analysis - Fine-Tuned BERT Model

This repository hosts a fine-tuned BERT-based model optimized for sentiment analysis on movie reviews using the IMDb dataset. The model classifies movie reviews as either Positive or Negative with high accuracy.

📌 Model Details

• Model Architecture: BERT
• Task: Sentiment Analysis
• Dataset: [IMDb Movie Reviews]
• Fine-tuning Framework: HF中国镜像站 Transformers
• Quantization: Float16

🚀 Usage

Installation

pip install transformers torch

Loading the Model

from transformers import BertTokenizer, BertForSequenceClassification
import torch

device = "cuda" if torch.cuda.is_available() else "cpu"

model_name = "AventIQ-AI/bert-movie-review-sentiment-analysis"
model = BertForSequenceClassification.from_pretrained(model_name).to(device)
tokenizer = BertTokenizer.from_pretrained(model_name)

Sentiment Prediction

import torch
import torch.nn.functional as F

def predict_sentiment(review_text):
    model.eval()  # Set model to evaluation mode
    inputs = tokenizer(review_text, padding=True, truncation=True, max_length=512, return_tensors="pt")

    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        probs = F.softmax(logits, dim=1)  # Convert logits to probabilities
        confidence, prediction = torch.max(probs, dim=1)  # Get class with highest probability

    sentiment = "Positive 😊" if prediction.item() == 1 else "Negative 😞"

    # Print probabilities for debugging
    print(f"Softmax Probabilities: {probs.tolist()}")

    # **Force correction for low confidence negative reviews**
    if confidence.item() < 0.7 and "not good" in review_text.lower():
        sentiment = "Negative 😞"

    return sentiment

# 🔹 **Test with Your Review**
review = "The movie was filled with boring dailogues and unrealistic action."
result = predict_sentiment(review)

print(f"Review: {review}")
print(f"Predicted Sentiment: {result}")

📊 Evaluation Results

After fine-tuning, the model was evaluated on the IMDb dataset, achieving the following performance:

Metric	Score	Meaning
Accuracy	92.5%	Percentage of correctly classified reviews
F1 Score	91.8%	Balance between precision and recall

🔧 Fine-Tuning Details

Dataset

The IMDb Movie Reviews dataset was used for training and evaluation. The dataset consists of 25,000 labeled movie reviews (positive/negative).

Training Configuration

• Number of epochs: 10
• Batch size: 32
• Optimizer: AdamW
• Learning rate: 3e-5
• Evaluation strategy: Epoch-based

Quantization

The model was quantized using float16 for inference, reducing latency and memory usage while maintaining accuracy.

📂 Repository Structure

.
├── model/               # Contains the fine-tuned model files
├── tokenizer_config/    # Tokenizer configuration and vocabulary files
├── model.safetensors/   # Quantized Model
├── README.md            # Model documentation

⚠️ Limitations

• The model may struggle with sarcasm and nuanced sentiments.
• Performance may vary across different writing styles and review lengths.
• Quantization may slightly affect accuracy compared to the full-precision model.

🤝 Contributing

Contributions are welcome! Feel free to open an issue or submit a pull request if you have suggestions or improvements.

---