Swiftkey : Datascience Capstone Project for Coursera

For more details on my Shiny App please visit https://sananand007.shinyapps.io/swiftkey/.

• Goal of the project is to predict the next word
• Requirements of this project were met based on the Final capstone guidelines
• Most of the concepts were based on Natural Language processing
• The predictive text model built for this project is using the data provided from a corpus called HC Corpora

• This project uses the classical n-gram modelling to develop the final algorithm
• Words were tokenized using the tidytext and the ngrams packages respectively to see if we can speed up the process
• Different Sample sizes were taken to form the Training set also to speed up the process
• unigram , Bigram and Trigrams are used for the KBO model
• Predicting for the quizzes were done with Morkov chain rule using Bayes theorem
• For the Final App , Katz backoff was used , where we discount the bigrams and the trigrams based on observed and unobsereved cases
• Katz backoff understanding and implementation was extremely timetaking and a real challenge

• App is extremely simple and ordinary
• Please wait for at least 1 minute with the sentence input to get the prediction

• Milestone report : https://rpubs.com/crafter707/297022
• Github : https://github.com/sananand007/project_nlp