Auto Complete App

Shounak Shastri
29/09/2020

Introduction

How does the app work?

The auto complete app predicts the next word the user is about to type.

The app refers to accompanying files which contain ngram phrases and their frequency of occurance in the corpus.

The user enters the words and the app cleans it and checks it against its files.

The word is predicted based on the frequency of the phrases in its ngram tables

The phrases which occur more frequently are given more preference.

Examples:

Working Examples

In the first image, the app is trying to figure out what the user is about to type and in the second image, once the app has the complete word, it predicts the next word.

Known bugs and future plans

Known bugs and issues

  1. The app works for pentagrams if the sentence is copied and pasted in the textbox. But if it is being typed in real time, then the predictions are from the bigram training list.

  2. The “…thinking…” status looks very similar to the predicted words

Plans for improvements

  1. Modify the code so that everytime there is a change in the textbox, the code should start in the pentagrams list.

  2. Change the colour of “…thinking…” so that it is not confused with the predicted words.

  3. Show multiple predictions. So instead of predicting 1 word, the app would give 5 most frequent words.

  4. Handling the error when there is nothing in the textbox.

Link to the app

https://shouankshastri.shinyapps.io/word_prediction/