Tech stack

Frontend: RMarkdown

Backend: R, Shiny

Features

• Simple UI
• Sensible prediction

What I did

Everything

Implementation

Suggestions are based from a corpus of tweets, news and blogs, which are then filtered to remove profanities, punctuations and converted to lowercase. Further analysis shows that when users type a word, it is more likely that they are using a collocation of two words, therefore suggestions mainly focus on bigrams, with trigrams less considered. For example typing go will give 3 suggestions of to, on, and home, all of which form phrases that are ubiquitous in daily life.