Importing the data set.

The list of libraries used, tokenizers, tm, stringi, stringr, ngramr, NLP, RWeka, SnowballC.

# Importing sample data
word_sample <- readLines(
  paste(getwd(),
        "Coursera-SwiftKey/final/en_US/en_US.corpus.txt",
        sep = "/")
)

Building corpus using custom function.

#Building corpus using custom function

build_corpus <- function(dataset){
  docs <- VCorpus(VectorSource(dataset))
  to_space <- content_transformer(
    function(x,pattern)
      gsub(pattern, " ",x)
  )
  
  # Removing URL, Twitter handles, email pattern.
  docs <- tm_map(docs, to_space,"(f|ht)tp(s?)://(.*)[.][a-z]+")
  docs <- tm_map(docs, to_space,"@[^\\s]+")
  docs <- tm_map(docs, to_space, "\\b[A-Z a-z 0-9._ - ]*[@](.*?)[.]{1,3} \\b")
  
  # Removing Profane words from the sample data. 
  
  con <- file(
    paste(getwd(),'Coursera-SwiftKey/final/en_US/en_US.profanity.txt',sep = "/"),
    "r"
  )
  profanity <-readLines(
    con, warn = FALSE, encoding = "UTF-8", skipNul = TRUE
  )
  close(con)
  
  profanity <- iconv(profanity, "latin1", "ASCII", sub = "")
  docs <- tm_map(docs, removeWords,profanity)
  
  # Converting the words to lower case.
  docs <- tm_map(docs, tolower)
  # Removing stopwords.
  docs <- tm_map(docs, removeWords, stopwords("english"))
  # Removing punctuation marks.
  docs <- tm_map(docs, removePunctuation)
  # Removing white spaces.
  docs <- tm_map(docs, stripWhitespace)
  # Removing Plain text document
  docs <- tm_map(docs, PlainTextDocument)
  # Removing numbers.
  docs <- tm_map(docs, removeNumbers)
  
  return(docs)
}

pf <- read.csv("profanity_en.csv")

head(pf)

library(readr)

profanity_words <- pf$text
text_file_path <- paste(
  getwd(),"Coursera-SwiftKey/final/en_US/en_US.profanity.txt",sep = "/"
)

con <- file(text_file_path,open = "w")
writeLines(profanity_words,con)
close(con)

rm(profanity_words)

Building a corpus based on the word sample.

word_sample_corpus <- build_corpus(word_sample)

corpus_text <- data.frame(text = unlist(
  sapply(word_sample_corpus, '[', "content")), stringsAsFactors = FALSE
)

corpus_filename <- paste(getwd(),"Coursera-SwiftKey/final/en_US/en_US.Mycorpus.txt",sep = "/")
con <- file(corpus_filename, open = "w")
writeLines(corpus_text$text,con)
close(con)


corpus_path <- paste(
  getwd(),"Coursera-SwiftKey/final/en_US/en_US.Mycorpus.txt",
  sep = "/"
)

corpus_text <- tolower(
  readLines(corpus_path, warn = FALSE)
)

corpus_words <- unlist(tokenize_words(
  corpus_text
))

Function to generate N-gram tokens.

# list of n-grams

generate_ngrams <- function(text, n) {
  ngram_list <- list()
  for (i in 1:(length(text) - n + 1)) {
    ngram <- paste(text[i:(i + n - 1)], collapse = " ")
    ngram_list <- append(ngram_list, ngram)
  }
  return(ngram_list)
}

# generating tokens

unigram_list <- generate_ngrams(corpus_words, 1)
bigram_list <- generate_ngrams(corpus_words, 2)
trigram_list <- generate_ngrams(corpus_words, 3)

Function for text prediction.

• This type of word prediction is often referred to as “n-gram language modeling.”

predict_next_word <- function(input_word, ngram_list) {
  input_word <- tolower(input_word)
  matching_ngrams <- grep(paste("^", input_word, sep = ""), ngram_list, value = TRUE)
  if (length(matching_ngrams) > 0) {
    next_ngram <- matching_ngrams[1]
    next_word <- strsplit(next_ngram, " ")[[1]][length(input_word) + 1]
    return(next_word)
  } else {
    return(NULL)
  }
}

Word prediction experiment

• The input word to test the prediction function is lime, the expected output is juice.

input_word <- "lime"

predicted_word <- predict_next_word(input_word, bigram_list)

if (!is.null(predicted_word)) {
  cat("Predicted next word:", predicted_word, "\n")
} else {
  cat("No prediction available\n")
}

## Predicted next word: juice

Summary

To put it simply this algorithm looks for groups of words in a list that begin with the provided input word and guesses the word by assuming that the first matching group is a continuation. This is a method of predicting words based on how certain groups of words appear in the data we have. However sophisticated models, for natural language processing like language models such, as LSTM or Transformer based models are frequently utilized to make more precise predictions that take into account the context.