Smart Keyboard

by Kamal Dobriyal
09-01-2022

Quick Overview

Smart Keyboard   is   a   web   app   working   on   NLP   n_grams   models.

It's   interface   is   very   simple,   there   a   sidebar   which   have   all   instructions   for   any   new   user.

The   user   just   have   to   enter   text   on   given   text   field   and   as   soon   as   space   is   entered   a   call   goes   to   backend   with   the   entered   text   as   input.

Eventually   user   sees   atmost   three   words   in   green   color   which   are   predicted   by   the   app.

Data Usage

We   have   use   US_News,   Us_Blogs,   and   US_Twitter   comments   for   generating   corpus.   Dataset   Link:   Data
Then   we   have   preprocessed   all   the   data   and   combined   it   to   form   a   single   corpus.
Library   and   techniques   used:

  • tokenizers
  • tm

We   removed   all   the   punctuations   except   apostrophe   sign   as   it   was   very   prevalent,   also   removed   the   hash-tags   because   they   were   without   space.

Backend & Predictive Model

In   the   backend,   we   have   already   loaded   three   n_grams   i.e.,   biGrams,   triGrams   and   quadGrams.

When   input   comes   in   the   backend   it   is   first   pre-processed,   like   removal   of   numerics,   punctuation,   hash-tags.

Then   based   on   the   tokens   generated   from   the   input   text,   subsequent   n_grams   dataset   is   being   looked up.

And   top   three   n_gram   with   highest   probability   is spit   out   by   the   repective   n_gram   function   and   passed to the UI.

App Link

Hope   you   like   the   app   overview   if   you   are   really   interested   please   check   out   the   app.

As   the   app   is   fully   based   on   shiny   framework   eventually   it   is   being   deployed   on   shinyapp.io   servers   for   free.

App   link:   Smart Keyboard