Next word predictive model

Shiny based online application as a data science final project product

  • by Slavik Koval
  • 26 April 2015

Goal of the project

Use the machine learning techniques to build an algorithm of the next word prediction during typing.

  • Self learning, collecting natural text data from different sources.
  • Use Markov chain approach with variable memory length
  • Achieve high accuracy of prediction by Statistical Learning

Product

Alt Text

Algorithm

Chains of words have different probabilities. By statistical analysis of many natural texts we can teach software to choose the most common continuation of phrase. The algorithm suggests the high probability occurences.

Alt Text

Performance

By choosing the optimal words chain length we achieve high prediction accuracy. When encountering a new chain which is not in database we suggest a random word from very common words and add the chain for future prediction.

Alt Text

Great value for your money!!!