The sentiment classifier uses a Long Short-Term Memory (LSTM) network to process sequences of word indices and determine the sentiment of a review. It includes modules for data processing, model training, and evaluation.
Reviews from IMDB dataset
- Sentiment classification: Classifies reviews into positive or negative sentiment
- Data handling: Processes and loads movie review data
- Vocabulary building: Constructs and saves a vocabulary for mapping words to indices
- Model training: Trains an LSTM model for sentiment analysis
- Evaluation: Evaluates the model and plots training and validation losses and accuracies
config.py: Contains configuration settings for the model.classifier.py: Defines theSentimentClassifiermodel.dataset.py: Contains theMyDatasetclass for handling data.utils.py: Utility functions for data processing, vocabulary handling, and accuracy calculation.train.py: Script to train the model.predict.py: Script to make predictions on new reviews.saved/: Directory for saving models and vocabulary.
- Clone the repository:
git clone https://github.com/yourusername/sentiment-classifier.git
- Navigate into the project directory:
cd sentiment-classifier - Install the required dependencies:
pip install torch torchvision torchtext matplotlib
To train the model, run the following command:
python train.pyThis will:
- Load training and testing data.
- Build or load the vocabulary.
- Train the model and save the best-performing model.
- Plot and save training and validation losses and accuracies.
To classify a review, use the predict.py script. Provide the review text as a command-line argument:
python predict.py "This movie was absolutely fantastic!"This will output the classification of the review as either Positive or Negative.
Configuration settings are located in config.py. You can adjust parameters such as vocabulary size, embedding dimension, hidden size, and learning rate.
- Python 3.7+
- PyTorch
- TorchText
- Matplotlib