LSTM-Next-Word

This repository contains a Streamlit app that predicts the next word in a given sentence fragment using a Long Short-Term Memory (LSTM) neural network trained on Shakespeare's Hamlet from the Gutenberg corpus.

Features

• Interactive Web App: A user-friendly Streamlit app to input a sentence and predict the next word.
• Trained LSTM Model: A deep learning model trained on the Hamlet text to generate contextually relevant predictions.
• Data Processing Notebook: A Jupyter notebook for data preprocessing, tokenization, model training, and evaluation.

Demo

Explore the live app here: LSTM Next Word Predictor

File Structure

LSTM-Next-Word/
│
├── app.py               # The Streamlit app script
├── processing.ipynb     # Jupyter notebook for data processing and model training
├── requirements.txt     # List of dependencies
├── next_word_lstm.h5    # Trained LSTM model
├── tokenizer.pkl        # Tokenizer object for word-index mapping

Installation

1. Clone the repository:

   git clone https://github.com/VanshajR/LSTM-Next-Word.git
   cd LSTM-Next-Word

1. Create and activate a virtual environment (optional but recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

1. Install dependencies:

   pip install -r requirements.txt

1. Run the Streamlit app locally:

   streamlit run app.py

Dataset

The app uses the text of Hamlet by William Shakespeare, sourced from the Gutenberg corpus (nltk.corpus.gutenberg).

How It Works

1. Data Preprocessing:

   • The Hamlet text is tokenized, and sequences of words are generated.
   • Input sequences are padded to uniform lengths, and labels are one-hot encoded.

2. Model Architecture:

   • An embedding layer maps words to dense vectors.
   • Two LSTM layers capture temporal dependencies in the text.
   • A dense softmax layer outputs the probabilities of the next word.

3. Streamlit App:

   • Users input a sentence fragment.
   • The app tokenizes the input, pads it, and uses the LSTM model to predict the next word.

Example

Input: "To be or not to be"
Output: "buried"

Training

Model training is detailed in processing.ipynb. It includes:

• Splitting data into training and test sets.
• Training the LSTM model for 80 epochs.
• Saving the trained model and tokenizer.

Dependencies

The project uses the following Python libraries:

• tensorflow
• numpy
• pandas
• scikit-learn
• nltk
• streamlit

All dependencies are listed in requirements.txt.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Built With

• jupyter-notebook
• python