Presentation Coursera Castone

• This is assignment of coursera datascience capstone project to find next predictive word.

• The Capstone is a cooperation between Coursera and SwiftKey company.

• Solution approach: Soultion is done in 3 parts

  • a) Create ngram dictionary (1 to 5) of words from twitter,news and blogs taking 50k each. Stopwords are removed.
  • b) Create prediction function which search the nth dictionary if user input n-1 words if thats not found in n-1 dictionary search in n-2 and so on and show that in the app
  • c) Put the code and dictionary in shiny server and user can access app from website
  • d) Once user enter something stopwords are removed and use predict function and reply
  • e) In the input box user type part of phrases and select no of prediction to be made(1 to 10) and the output is shown below

• Code: Code is available at : https://github.com/suman12345678/datasciencecapstone

• App : Application can be access at https://suman123456.shinyapps.io/shiny/

• Presentation : http://rpubs.com/suman12345678/capstoneproj

setwd("C:\\Users\\suman\\Desktop\\datasciencecoursera\\capstone\\shiny")
#nextword<-predictnextword("hello good",2)

• This is just a sample of .15 mil raw data so there might be very less phrases

• There is capacity for 8GB ram to hold and process this much data and make balance between speed and size

• The application might be bit slow so please keep patience

• Stopwords are removed so user shouldnot expect any artical/preposition/aux verbs as predicting word

• The Capstone is a cooperation between Coursera and SwiftKey company.

• Maybe stemword can be used to ignore non english and short word character

• It can be integrated with spark to process huge data

• Probablly news and blogs data are more gramaticall correct so should take more sample from those