Predictive Text Model

An Innovative Solution for Next-Word Prediction

Samuel Owusu Manu

2024-07-27

How the Model Works

• Data Collection: The en_US text data was collected, cleaned and preprocessed it all into a one text.
• Tokenizer: The text data was converted into sequences of integers using a tokenizer.
• Model Architecture: We used a Sequential LSTM network for prediction.
• Training: The model was trained using the Keras library in TensorFlow.

Predictive Performance

• Evaluation Metrics:
  • Accuracy: 85%
  • Perplexity: 20.5
• Model Comparison:
  • Our model outperforms baseline n-gram models.

Showcasing the App

• User Interface:
  • Simple and intuitive input box for n-grams.
  • Predict button to generate the next word.
• Visualization:
  • Bar chart of top predicted words.

App Screenshot

Why Our Product is Awesome

• High Accuracy:
  • Significantly improves next-word prediction accuracy.
• User-Friendly:
  • Intuitive and easy-to-use interface.
• Scalable:
  • Can be integrated into various applications like messaging apps, search engines, etc.
• Visual Insights:
  • Provides visual feedback to users with bar charts and word clouds.