Text-Autocomplete

python, n-gram language models, wikitext-103 dataset

A next-word prediction interface built around an N-gram language model trained on WikiText-103.

view project

Challenge

The goal was to learn how autocomplete system works.

Approach

I trained an N-gram pipeline, tuned the prediction flow for practical suggestions, and wrapped it in a lightweight interface that reacts as the user types.

Outcome

The result is a fast predictive typing demo that makes the underlying model legible, interactive, and easy to evaluate.

Highlights

  • trained on WikiText-103
  • real-time next-word suggestions
  • lightweight interface for instant feedback
NextArsenal-Forecast