LexiPredict is a lightweight, responsive web application designed to assist users in text entry by predicting the most likely next word in a sequence.

• Speed: Optimized N-gram tables for instant results.
• Accuracy: Leverages a 5% sample of the SwiftKey dataset.
• Accessibility: Hosted on shinyapps.io for universal access.

1. It searches for a matching Trigram (3-word sequence).
2. If no match is found, it “backs off” to a Bigram (2-word sequence).
3. If still no match, it provides the most frequent word in the corpus (“the”).

This ensures a prediction is always returned without the computational heaviness of complex Markov chains.

To maintain a balance between accuracy and app speed:

• Model Size: Compressed RDS files (< 5MB).
• Latency: Average prediction time < 0.1 seconds.
• Coverage: Successfully predicts common English idioms and sentence structures from Twitter and News datasets.

Using the app is simple and intuitive:

1. Navigate to the LexiPredict URL.
2. Enter your text into the input field.
3. View the predicted word instantly in the dashboard.

It’s like having a digital mind-reader at your fingertips!

LexiPredict demonstrates a scalable architecture for Natural Language Processing (NLP).

• Efficient memory management.
• Robust handling of “Out of Vocabulary” (OOV) words.
• Ready for integration into mobile keyboards or messaging apps.

Thank you for your attention