Data cleaning on user reviews have been done. And word cloud has also been created.

rm(list = ls())

# installing packages

install.packages("hunspell")

install.packages("NLP")

install.packages("wordcloud2")

install.packages("tm")

install.packages("wordcloud2")

install.packages("SnowballC")

library(hunspell)

library(NLP)

library(wordcloud2)

library(tm)

library(wordcloud2)

library(SnowballC)

#loading of ncessary dataset

load(file = "D:/Hohenheim/SEM 3/ADS/introads_ass2_team18/01_data/raw/gamereviews.RData")

load(file = "D:/Hohenheim/SEM 3/ADS/introads_ass2_team18/01_data/raw")

# Cleaning of the user reviews---------------------------------------------

game.rev <- read.csv("gamereviews.RData")

# New dataframe for data cleaning

reviews.clean <- game.rev

clean_rev <- function(review) {

  cleaned_review <- gsub("[^\\x00-\\x7F]+", "", review, perl = TRUE)

  cleaned_review <- gsub("http[s]?://\\S+", "", cleaned_review)

  cleaned_review <- gsub("(.)\\1{2,}", "\\1 ", cleaned_review)

  cleaned_review <- gsub("n't", "not ", cleaned_review)

  cleaned_review <- gsub("\\d", " ", cleaned_review)

  cleaned_review <- gsub("[[:punct:]]", "", cleaned_review)

  cleaned_review <- gsub("\\n", " ", cleaned_review)

  cleaned_review <- gsub("^\\s+", " ", cleaned_review)

  cleaned_review <- gsub("\\s+", " ", cleaned_review)

  cleaned_review <- ifelse(nchar(cleaned_review) == 1, "", cleaned_review)

  return(cleaned_review)

}

reviews.clean$reviews <- sapply(reviews.clean$review, clean_rev)

#reviews.clean$reviews <- gsub("[^\\x00-\\x7F]+", "", reviews.clean$reviews, perl = TRUE) # remove non-ASCII character

#reviews.clean$reviews <- gsub("http[s]?://\\S+", "", reviews.clean$reviews) # remove external URL

#reviews.clean$reviews <- gsub("(.)\\1{2,}", "\\1 ", reviews.clean$reviews) # remove repeating character

#reviews.clean$reviews <- gsub("n't", "not ", reviews.clean$reviews)  # replacing n't with not with an empty space

#reviews.clean$reviews <- gsub("\\d", " ", reviews.clean$reviews) # remove numerical character

#reviews.clean$reviews <- gsub("[[:punct:]]", "", reviews.clean$reviews) # remove punctuation

#reviews.clean$reviews <- gsub("\\n", " ", reviews.clean$reviews) # replace new lines with empty space

#reviews.clean$reviews <- gsub("^\\s+", " ", reviews.clean$reviews)  # remove unnecessary empty space at the beginning

#reviews.clean$reviews <- gsub("\\s+", " ", reviews.clean$reviews) # remove unncessary empty space at the end

#reviews.clean$reviews <- ifelse(nchar(reviews.clean$reviews) == 1, "", reviews.clean$reviews)

reviews.clean$review <- reviews.clean$reviews # replacing clean reviews column with raw review column as examination has been done

reviews.clean <- reviews.clean[, -ncol(reviews.clean)] # remove clean review column as it has been replaced

#saving the dataset on a local computer

setwd("D:/Hohenheim/SEM 3/ADS/introads_ass2_team18/01_data")

write.csv(reviews.clean, file = "reviews.clean.csv")

# another dataframe for spelling correction

reviews.checked <- reviews.clean

# checking on 1st review

#review1 <- reviews.checked$review[1]

# 5.2 separating 1st review into words

#words <- unlist(strsplit(review1, " "))

# 5.3 checking if the words are spelled corectly based on US english dictionary

#check.spell <- hunspell_check(words, dict="en_US")

#print(check.spell)

# 5.4 another vector which has only correct words from review 1

#words_correct <- words[!words %in% check.spell]

#print(words_correct)

# 5.5 transforming into one string

#review1 <- paste(words_correct, collapse = " ")

# 5.6 Replacing new review with original review 1 in the dataframe

#reviews.checked$review[1] <- review1

#print(reviews.checked$review[1])

# Spell checking of all reviews---------------------------------------------

# Loop through all reviews to check spelling based on US english dictionary

for (i in 1:nrow(reviews.checked)){

  reviews <- reviews.checked$review[i]

  #extract words

  words <- unlist(strsplit(reviews, " "))

  #Spell check

  spell_check <- hunspell_check(words, dict = "en_US")

  # Consider only correct words

  correct_words <- words[!words %in% spell_check]

  # Replacing correctly spelled review with original review

  correct_review <- paste(correct_words, collapse = " ")

  # update the dataframe

  reviews.checked$review[i] <- correct_review

}

#reviews.checked$review <- reviews.clean$reviews # replacing clean reviews column with raw review column as examination has been done

#reviews.clean <- reviews.clean[, -ncol(reviews.clean)] # remove clean review column as it has been replaced

# turn all upper case characters into lower case character

reviews.checked$review <- tolower(reviews.checked$review)

# delete reviews with ""

reviews.final <- reviews.checked[reviews.checked$review != "",]

write.csv(reviews.final, file = "reviews.final.csv")

# word cloud--------------------------------------------------------------------

# combine all reviews into a single text

all_reviews <- paste(reviews.final$review, collapse = "")

# a vectore of all words

all.words <- unlist(strsplit(all_reviews,""))

all.words <- character()

# loop through all reviews

for (i in 1:nrow(reviews.final)) {

  #split the review into words

  tmp <- unlist(strsplit(reviews.final$review[i], " "))

  # Concatenate the words to the all.words vector

  all.words <- c(all.words, tmp)

}

# Deleting words with less than 3 letters

all.words <- all.words[nchar(all.words)>=4]

# Counting the frequency of the words

word.freq <- data.frame(table(all.words))

# creating word count

word.cloud1 <- wordcloud2(data = word.freq, size=1.2,

                          color = "random-light", backgroundColor = "black")

# remove the filling words

all.words.wof <- all.words[!(all.words %in% stopwords("english")) | all.words == "not"]

# dateframe after removing filling words

words.freq <- data.frame(table(all.words.wof))

# word clod 2

wordcloud2(data = words.freq, size=1.2,

           color = "random-light", backgroundColor = "black")

# Obaining word stems

word.stems <- wordStem(all.words.wof)

# Creating a word cloud using word stems

wordcloud2(table(word.stems), size = 1.2, color = "random-light", 

           backgroundColor = "black")