Predict next word app

Freddy F. Tapia C.
November 25th, 2020

Description

With this app you will be able to predict the next word of a sentence or word that you consider. This app is the final activity of the Capstone of the specialized program of Data Science.

This specialization consists of seven weeks duration, where in each of them an important step is carried out to understand the data with which you are working.

In the first two weeks you work with the data. In the third week, the first model is created that helps predict the next word given in a sentence or word.

On the other hand, in the fourth and fifth week, the performance of the model is studied in order to obtain a quick response. Finally, in the last two weeks, the app is created as well as its presentation

Sections

The sections that you will find in the app are,

  • Welcome: contains a description of the app, the sections and how to use it.

  • Predict next word: contains the main tool that predict the next word, given a sentence or a single word.

  • About: contains information about the developer of this app.

How to use it

In the section Predict next word you will find a box where you have to write a word or a sentence in order to predict the next word, you need to follow this steps,

  • Enter to the section Predict next word and write a word or a sentence, then click in the Predict button.

  • A table with the main ten (or less) words will be displayed, this table will show the probability of each word, the first word represent the most probable next word according to the word or sentence that you wrote.

How it works

This app use data provided by SwiftKey, for this app three main files will be used,

  • Blogs: contains data colected from differents blogs.
  • News: contains data colected from differents news sources.
  • Twitter: contains data colected from differents users of Twitter.

In this app we will use the files that are in english. In order to obtain the final data that we will use for the prediction, the files where pre-processed.