IMDB Sentiment Analysis with BERT

This project implements a sentiment analysis model using a BERT-based transformer from the Hugging Face library. The model is trained on the IMDB dataset and can predict sentiment (positive or negative) for given text inputs.

• Uses BERT (bert-base-uncased) for sentiment classification
• Trains on the IMDB dataset
• Provides an interactive mode for real-time sentiment analysis
• Utilizes PyTorch and the Transformers library

• Python 3.7+
• PyTorch
• Transformers
• Datasets

You can install the required packages using pip:

pip install torch transformers datasets

1. Run the script:

python sentiment_analysis.py

2. The script will perform the following steps:

   • Load the IMDB dataset
   • Initialize the BERT model and tokenizer
   • Train the model on the training set
   • Evaluate the model on the test set
   • Enter interactive mode

3. In interactive mode, you can input sentences to analyze their sentiment:

   • Type a sentence and press Enter
   • The model will output the predicted sentiment (Positive or Negative) and a score
   • Type 'quit' to exit the interactive mode

Enter text: This movie was absolutely fantastic!
Sentiment: Positive (Score: 0.98)

Enter text: I was disappointed by the poor acting and weak plot.
Sentiment: Negative (Score: 0.03)

Enter text: quit
Thank you for using the sentiment analyzer!

You can modify the following parameters in the script to customize the model:

• num_epochs: Number of training epochs
• BATCH_SIZE: Batch size for training
• max_length in IMDBDataset: Maximum sequence length for tokenization

• The model uses the GPU if available, otherwise it falls back to CPU.
• Training may take some time depending on your hardware.
• The interactive mode allows for quick testing of the model's performance on custom inputs.

• Implement early stopping to prevent overfitting
• Add support for saving and loading trained models
• Experiment with different pre-trained models or fine-tuning strategies

Feel free to contribute to this project or report any issues you encounter!