November 21, 2026

1. Project Overview

SwiftKey Predictive Text Application

Objective: Develop an NLP-based application that predicts the next word based on user input.

  • Developed using R and Shiny.
  • Implements Unigram, Bigram and Trigram models.
  • Uses text from Blogs, News and Twitter.
  • Provides intelligent next-word suggestions.

2. Dataset & Methodology

Training Data

  • Total training records: 60,000
  • Data sources: Blogs, News and Twitter.
  • Unigrams: 50,000
  • Bigrams: 50,000
  • Trigrams: 50,000

Methodology

  1. Text data collection.
  2. Text cleaning and preprocessing.
  3. N-gram model generation.
  4. Next-word prediction.
  5. Alternative suggestion generation.

3. Predictive Text Application

Application Features

  • Interactive Shiny interface.
  • Sentence and phrase input.
  • Main next-word prediction.
  • Four alternative suggestions.
  • Model information display.

Live Application

Open SwiftKey Application

4. Implementation & Deployment

Technical Implementation

  • Programming Language: R
  • Framework: Shiny
  • Language Models: Unigram, Bigram, Trigram
  • Model Storage: RDS file
  • Deployment Platform: ShinyApps.io

Deployment Status

Successfully deployed and publicly accessible.

Application URL:

https://dhruvtyagi.shinyapps.io/swiftkey_capstone/

5. Conclusion

Project Outcomes

  • Developed a functional predictive text application.
  • Implemented three N-gram language models.
  • Integrated the trained model with Shiny.
  • Created an interactive prediction interface.
  • Successfully deployed the application online.

Presented by: Dhruv Tyagi

Data Science Capstone Project

Thank You!