This repository implements several Recurrent Neural Network (RNN) architectures, including Simple RNN, LSTM, GRU, and LSTM with Attention, to perform Sentiment Analysis on the IMDB movie reviews dataset. The models are trained and evaluated on this dataset to determine the effectiveness of each architecture for sentiment prediction.
The dataset used for this project is the IMDB movie reviews dataset. It consists of 50,000 reviews, equally divided between positive and negative reviews. Each review is encoded as a sequence of integers, with each integer representing a unique word in the vocabulary.
- Training Data: 25,000 samples
- Test Data: 25,000 samples
- Vocabulary Size: 10,000 most frequent words
- Maximum Review Length: 500 words (truncated or padded)
The Simple RNN model is a basic recurrent network that processes sequences of words in the reviews. It consists of the following layers:
- Embedding Layer: Converts word indices into dense vectors.
- RNN Layer: Processes the sequence data.
- Dense Layer: Outputs a binary classification (positive or negative sentiment).
The LSTM model is a more advanced type of RNN designed to overcome the vanishing gradient problem in Simple RNNs. It captures long-range dependencies in the sequences, making it more effective for tasks involving long text sequences.
The GRU model is another type of RNN similar to LSTM but with a simpler architecture. GRUs are computationally more efficient and have fewer parameters than LSTMs.
The LSTM with Attention model enhances the LSTM by using an Attention mechanism. This allows the model to focus on specific parts of the input sequence, improving its performance in tasks such as sentiment analysis where certain words or phrases hold more significance.
All models are trained using the following parameters:
- Optimizer: Adam
- Loss Function: Binary Cross-Entropy
- Metrics: Accuracy
- Epochs: 10
- Batch Size: 64