Word Predictor

Alexander Klonov
23.11.2020

Web App

Web app can be found here: https://zennoma.shinyapps.io/Word_Predictor/.

App has several abilities:

  • It can be used to predict based on a single word of input;
  • However it's capable of analyze sentences and make predictions based on couple of words;
  • It's fast and reactive, meaning you get an instant answer;
  • Pretty straightforward and easy to use

Algorithm explanation

The main idea of the algorithm is to use provided data, to separate them by chunks and make prediction based of similarity of those chunks to the user's input. This method called nGramTokenization

After some data cleaning and refactoring we can get table which looks like this

trigrams <- readRDS("./Word_Predictor/data/trigrams")
head(trigrams)

       Predictor Freq Result
5732       a lot  153     of
201717    one of  151    the
281296     the u   80      s
150920    it was   73      a
248013   some of   72    the
293469     to be   69      a

Quick start guide

After you opened this web application you get to home or main page. It's there you use this program, you simple write something in special input box and get prediction. Also, user can set the number of different predicted words, they are formed in frequincy based list, meaning the most common combination will be first etc.

How does the page looks like

Sources

All work was done by Alexander Klonov for Coursera's Data Science Specialization course.

Source files including shiny app can be found in git repository https://github.com/Shureks-den/Capstone_Project

Thank you for your attention.