Text Prediction Application

Alok Yadav
August 23, 2020

What Is This About ?!

It's tough to predict the future but not words!

Applications derived from word based statistical model has helped improve mankind's quality of life.

For examples,

  1. SwiftKey Keyboard predicts the next word (as user types) that makes composition task on mobile a breeze!

  2. Text-To-Speech Synthesize assists in speech construction when challenged by disability and breaks down physical communication boundaries. (E.g. Stephen Hawking)

How It Works ?

The High Level Architecture as below,

High Level Architecture

Markov Chain Model represents each word as transitional state with probability. An intuitive example, word “B” has 30% chance to transit to word “A”.

Add-1 Smoothing is applied to distribute probability mass to produce a less-biased prediction model.

Too Easy !

Type some text and hit the SUBMIT button! Results come back less than seconds!

User Interface Performance Reading

What's Next ?

The prediction model is somewhat skewed. Possible improvements are,

  1. Increase Corpus size and/or type.
  2. Increase machine resources - create more complex model.
  3. Increase Shiny io resources - host more complex text prediction application.

What are you waiting for ? Give it a try here or here!