Poetic-Text-Generation-with-LSTM

This project implements a text generation model using an LSTM neural network in TensorFlow/Keras. It generates poetic text based on a training dataset, such as Shakespeare's plays, and allows you to experiment with different "temperature" values to control the randomness of the generated text.

Features

• Preprocessing of a text dataset.
• Training an LSTM-based model for sequence prediction.
• Generating text with adjustable creativity (temperature).
• Saving and reusing the trained model.

---

Dataset

The dataset used is Shakespeare's text, downloaded directly from TensorFlow's storage: Shakespeare Dataset

---

Installation and Setup

1. Clone this repository:

   git clone https://github.com/yourusername/poetic-text-generation.git
   cd poetic-text-generation

2. Install required libraries:

   pip install tensorflow numpy

3. Run the script to train the model:

   python Poetic_text_generation.py

---

File Structure

• Poetic_text_generation.py: Main script for preprocessing, training, and generating text.
• textgenerator.keras: Trained model saved in Keras format.
• README.md: Project documentation.

---

Usage

1. Training the Model

The script preprocesses the dataset, trains an LSTM model, and saves the trained model as textgenerator.keras.

2. Generating Text

The script provides a function to generate text:

generate_text(length, temperature)

• length: Number of characters to generate.
• temperature: Controls randomness in text generation. A lower temperature generates more predictable text, while a higher temperature generates more random text.

Example outputs for different temperatures:

• Temperature 0.2:

  thou art a villain, a coward, a slave, and a
  

• Temperature 1.0:

  thou hrtnpotie - deliebly ur' heawife, thou art thou!
  

---

Code Highlights

Model Architecture

• A simple sequential LSTM model:

  model = Sequential([
      LSTM(128, input_shape=(SEQ_LENGTH, len(characters))),
      Dense(len(characters)),
      Activation('softmax')
  ])

Sampling Function

• Adjusts randomness in text generation:

  def sample(preds, temperature=1.0):
      preds = np.asarray(preds).astype('float64')
      preds = np.log(preds) / temperature
      exp_preds = np.exp(preds)
      preds = exp_preds / np.sum(exp_preds)
      return np.argmax(np.random.multinomial(1, preds, 1))

---

Results

After training, the model can generate coherent poetic text similar to Shakespeare's style. Experiment with the temperature value to observe changes in creativity.

---

License

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

---

Let me know if you’d like me to create or tweak any part of this!