Auto Word Predictor

• The aim of this project to present an shiny web app that can help a user in predicting the next word based on user input.

• User can predict upto three most likely words using a drop down.

• The application uses an n-gram model to predict the next word.

• Click here to access the app.

Data Ingestion:

• The textual data used to train the predictive model is extracted from a sample of 800,000 lines extracted from the large corpus of blogs, news and twitter data provided by Swiftkey.

Data Cleaning and Preparation:

• Removing web and email addresses
• Removing hashtags and twitter handle names from the tweets
• Removing ordinal numbers.
• Removing profane words.
• Removing punctuations and extra whitespaces.
• The cleaned text corpus is then converted to N-grams upto frequency 4.

• When user enters text, the algorithm iterates from longest n-gram (4-gram) to shortest (2-gram) to detect a match.
• Unigrams are left to trade-off with performance.
• The predicted next word is considered using the longest, most frequent matching n-gram.
• The algorithm makes use of a simple back-off strategy.

• User can start typing in the text box given, the text box with heading 'Input' will show the text you are entering.
• Select the number of most likely words to predict using the drop down (default value is 1).
• The text box/boxes with heading 'Prediction will show the most likely next word predictions using the algorithm.