This is my final project using NLP for Data science capstone where I have made a data product that predicts the next word of the input string.

Overview of how I created the data product

• Cleaning & exploration of data from twitter,blogs & news
• Tokenization & creation of n-grams data (uni,bi,tri,quad-gram)
• Came up with a prediction model that predicts the next word of a given string
• Wrote a shiny application that uses the model to predict and deployed it into shiny server

• After receiving the text input, it first cleans the text(to lowercase etc)
• Then it counts the words of input & assigns the input to it’s respective n-gram and a prediction data
• Used a backoff model for inputs that didnt have a match so, it can get a match in a lower gram (N-1)
• Finally if no match found it will just print the words with most frequency in unigram(single word)
• otherwise it will print the top 3 matches with most frequency

The Application is very much easy to use & is accessible by anyone.

Firstly, put your text input (string of words) into the text box given on the left side and click on the predict button

kindly wait for the model to generate the output as it might take some time (usually 4-7 secs).

& voila!! you will have the predicted word generated right below the action button along with a boxplot showcasing top 3 words matched with highest frequency.

The application is deployed out here:

shiny application

The codes for the project is available out here:

Github