Sentiment Analysis Using BERT

Overview

This project implements a sentiment analysis model using BERT (Bidirectional Encoder Representations from Transformers). The model is fine-tuned on the SST-2 dataset from the GLUE benchmark to classify text as either positive or negative.

Features

• Preprocessing & Tokenization: Uses Hugging Face's AutoTokenizer to preprocess text data.
• Model Architecture: Implements BertForSequenceClassification from the transformers library.
• Training & Fine-Tuning: Utilizes TrainingArguments for optimized training.
• Evaluation & Metrics: Uses the evaluate library for performance analysis.
• Model Saving: Saves trained models for inference and reuse.

Dataset

The project uses the SST-2 dataset from the GLUE benchmark, which consists of sentences labeled as either positive or negative.

Installation

Ensure you have Python installed, then install the necessary dependencies:

pip install torch transformers datasets evaluate

Usage

1. Clone the Repository

git clone https://github.com/your-repo/sentiment-analysis-bert.git
cd sentiment-analysis-bert

2. Run the Training Script

python train.py

3. Evaluate the Model

python evaluate.py

4. Use the Model for Inference

from transformers import AutoTokenizer, BertForSequenceClassification
import torch

tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
model = BertForSequenceClassification.from_pretrained("path_to_saved_model")

text = "I love this movie!"
inputs = tokenizer(text, return_tensors="pt")
outputs = model(**inputs)
predictions = torch.argmax(outputs.logits, dim=-1)
print("Sentiment:", "Positive" if predictions.item() == 1 else "Negative")

Results

The model is evaluated using accuracy and F1-score. Fine-tuning BERT achieves high performance on SST-2, demonstrating its effectiveness in sentiment classification.

Future Improvements

• Experiment with different transformer architectures (e.g., RoBERTa, DistilBERT).
• Optimize hyperparameters for better accuracy.
• Extend the model to multi-class sentiment classification.

License

This project is licensed under the MIT License.

Acknowledgments

• Hugging Face Transformers
• SST-2 Dataset
• PyTorch