Text-Mining Capstone for Data Science Specialization from John Hopkin via Coursera

N-grams model with markov assumptions. A model is trained on the given data after cleaning the corpus, a frequency matrix is created and N-gram model with markov assumptions is used to predict last work in my app.

All the code for the project is in the RR.R file, below is brief explanation for the fuction that generates the last word in Shiny app.

Here you will find the Shiny app that runs on the trained model.

And here are couple slides in Rpubs.

plw <- function(words) {
  
    tok <- tokens(removePunctuation(tolower(removeNumbers(replace_contraction(words)))))[[1]]

#strip the given sequence of words to most basic words and use the last 3 words/token

            len <- length(tok)
            wz <- tok[len]
            wy <- tok[len-1]
            wx <- tok[len-2]
       
    ifelse(len >= 3 & (!identical(dfm_4[w1 == wx & w2 == wy & w3 == wz][, w4], character(0))),
        return(dfm_4[w1 == wx & w2 == wy & w3 == wz][, w4]),
        
        ifelse(len >=2 & (!identical(dfm_3[w1 == wy & w2 == wz][, w3], character(0))),
               return(dfm_3[w1 == wy & w2 == wz][, w3]),
               
               ifelse(len >= 1 & (!identical( dfm_2[w1 == wz][, w2], character(0) )),
                      return(dfm_2[w1 == wz][, w2]), print("No Match Found"))))
                      
#choose the final words after matching last 3, 2, or 1 word, if no match return "No Match Found".
}