Word Prediction Application

The Coursera Data Science Specialization Capstone project by Johns Hopkins University(JHU) teaches students to create a public data product that can show their skills to their employers. For their capstone project JHU partnered with SwiftKey(http://swiftkey.com/en/) to apply data science in the field of Natural Language Processing(NLP).

The objective of this project was to build a word prediction model that predicts the next word after the input text. The data used in this model came from Coursera Dataset(https://d396qusza40orc.cloudfront.net/dsscapstone/dataset/Coursera-SwiftKey.zip).

N-gram model[1] was the base of the algorithm developed to predict the next word. As the data set was very large, a small portion of it was used to compute the n-grams. The n-grams like unigram, bigram and trigram are used.

To improve the accuracy, smoothing was applied to the algorithm, combining unigram, bigram and trigram probabilities. The default predictions were employed using the part of speech tagging(POST)[2].

1-https://en.wikipedia.org/wiki/N-gram

2-https://en.wikipedia.org/wiki/Part-of-speech_tagging

After the algorithm was developed a Shiny application was developed using that algorithm that accepts a phrase as input and predicts the most likely next word based on the n-grams. Here is the link to the Shiny Application (https://riyasutaria.shinyapps.io/app-10/).