by Kamal Dobriyal
09-01-2022
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.
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:
We removed all the punctuations except apostrophe sign as it was very prevalent, also removed the hash-tags because they were without space.
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.
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