Introduction

The goal of the capstone project is to create a predictive text model using a large text corpus of documents as training data. Natural language processing techniques will be used to perform the analysis.

Loading (Blogs, News, Twitter)

path <- getwd()
# download.file(url = "https://d396qusza40orc.cloudfront.net/dsscapstone/dataset/Coursera-SwiftKey.zip"
              # , destfile = paste(path, "dataset.zip", sep = "/"))
# since the file is too large I have commented it
# unzip("dataset.zip")
blogs <- readLines("final/en_US/en_US.blogs.txt", warn = FALSE, encoding = "UTF-8")
news <- readLines("final/en_US/en_US.news.txt", warn = FALSE, encoding = "UTF-8")
twitter <- readLines("final/en_US/en_US.twitter.txt", warn = FALSE, encoding = "UTF-8")

Sampling

library(stringi) # stats files
# Size of Files file.size
size_blogs <- file.info("final/en_US/en_US.blogs.txt")$size / 1024^2 # 200 Megabytes
size_news <- file.info("final/en_US/en_US.news.txt")$size  / 1024^2 # 196 Megabytes
size_twitter <- file.info("final/en_US/en_US.twitter.txt")$size / 1024^2 # 159 Megabytes
# Number of Lines num.lines
len_blogs <- length(blogs) # 899,288 lines
len_news <- length(news)  # 1,010,242 lines
len_twitter <- length(twitter) # 2,360,148

# Number of characters
nchar_blogs <- sum(nchar(blogs))
nchar_news <- sum(nchar(news))
nchar_twitter <- sum(nchar(twitter))

# Counting the Words (num.words)
nword_blogs <- sum(stri_count_words(blogs)) # words at blogs = 37,546,246
nword_news <- sum(stri_count_words(news))  # words at news =  34,762,395
nword_twitter <-sum(stri_count_words(twitter)) # words at twitter = 30,093,410

# create table 
data.frame(file.name = c("blogs", "news", "twitter"),
files.size.MB = c(size_blogs,size_news,size_twitter),
num.lines = c(len_blogs,len_news,len_twitter),
num.character = c(nchar_blogs,nchar_news,nchar_twitter),
num.words = c(nword_blogs,nword_news,nword_twitter))
##   file.name files.size.MB num.lines num.character num.words
## 1     blogs      200.4242    899288     206824505  37546806
## 2      news      196.2775     77259      15639408   2674561
## 3   twitter      159.3641   2360148     162096031  30096649

Sampling

Remove all non-English characters and then compile a sample dataset that is composed of 1% of each of the 3 original datasets.

set.seed(123)
blogs1 <-iconv(blogs,"latin1","ASCII",sub="")
news1 <-iconv(news,"latin1","ASCII",sub="")
twitter1 <-iconv(twitter,"latin1","ASCII",sub="")

# sample data set only 1% of each file
sample_data <-c(sample(blogs1,length(blogs1)*0.01),
               sample(news1,length(news1)*0.01),
               sample(twitter1,length(twitter1)*0.01))

Since Data sets is too big for processing, so using sample() function, I sample 1% of each file.

Clean and Build Corpus

library(tm) # Text mining
## Loading required package: NLP
library(NLP)

corpus <- VCorpus(VectorSource(sample_data))
corpus1 <- tm_map(corpus,removePunctuation)
corpus2 <- tm_map(corpus1,stripWhitespace)
corpus3 <- tm_map(corpus2,tolower) # Convert to lowercase
corpus4 <- tm_map(corpus3,removeNumbers)
corpus5 <- tm_map(corpus4,PlainTextDocument)
#removing stop words in English (a, as, at, so, etc.)
corpus6 <- tm_map(corpus5,removeWords,stopwords("english"))

Build N-Grams

In Natural Language Processing (NLP), n-gram is a contiguous sequence of n items from a given sequence of text or speech. Unigrams are single words. Bigrams are two words combinations. Trigrams are three-word combinations.

The following function is used to extract 1-grams, 2-grams, 3-grams from the text Corpus using RWeka.

library(RWeka)# tokenizer - create unigrams, bigrams, trigrams

#I use the RWeka package to construct functions that tokenize the sample and construct matrices of uniqrams, bigrams, and trigrams.

one<-function(x) NGramTokenizer(x,Weka_control(min=1,max=1))
two<-function(x) NGramTokenizer(x,Weka_control(min=2,max=2))
thr<-function(x) NGramTokenizer(x,Weka_control(min=3,max=3))

one_table<-TermDocumentMatrix(corpus6,control=list(tokenize=one))
two_table<-TermDocumentMatrix(corpus6,control=list(tokenize=two))
thr_table<-TermDocumentMatrix(corpus6,control=list(tokenize=thr))

#Then I find the frequency of terms in each of these 3 matrices and construct dataframes of these frequencies.

one_corpus<-findFreqTerms(one_table,lowfreq=1000)
two_corpus<-findFreqTerms(two_table,lowfreq=80)
thr_corpus<-findFreqTerms(thr_table,lowfreq=10)

one_corpus_num<-rowSums(as.matrix(one_table[one_corpus,]))
one_corpus_table<-data.frame(Word=names(one_corpus_num),frequency=one_corpus_num)
one_corpus_sort<-one_corpus_table[order(-one_corpus_table$frequency),]
head(one_corpus_sort)
##      Word frequency
## just just      2507
## will will      2224
## like like      2214
## one   one      2128
## get   get      1916
## can   can      1863
two_corpus_num<-rowSums(as.matrix(two_table[two_corpus,]))
two_corpus_table<-data.frame(Word=names(two_corpus_num),frequency=two_corpus_num)
two_corpus_sort<-two_corpus_table[order(-two_corpus_table$frequency),]
head(two_corpus_sort)
##                            Word frequency
## right now             right now       249
## cant wait             cant wait       188
## dont know             dont know       167
## last night           last night       145
## im going               im going       129
## looking forward looking forward       123
thr_corpus_num<-rowSums(as.matrix(thr_table[thr_corpus,]))
thr_corpus_table<-data.frame(Word=names(thr_corpus_num),frequency=thr_corpus_num)
thr_corpus_sort<-thr_corpus_table[order(-thr_corpus_table$frequency),]
head(thr_corpus_sort)
##                                Word frequency
## cant wait see         cant wait see        43
## happy mothers day happy mothers day        28
## damn damn damn       damn damn damn        24
## let us know             let us know        20
## happy new year       happy new year        17
## dont even know       dont even know        16

Exploratory Analysis (Graphs & Visualizations)

The frequency distribution of each n-grams category were visualized into 3 different bar plots.

library(ggplot2) #visualization
## 
## Attaching package: 'ggplot2'
## The following object is masked from 'package:NLP':
## 
##     annotate
one_g<-ggplot(one_corpus_sort[1:10,],aes(x=reorder(Word,-frequency),y=frequency,fill=frequency))
one_g<-one_g+geom_bar(stat="identity")
one_g<-one_g+labs(title="Unigrams",x="Words",y="Frequency")
one_g<-one_g+theme(axis.text.x=element_text(angle=90))
one_g

two_g<-ggplot(two_corpus_sort[1:10,],aes(x=reorder(Word,-frequency),y=frequency,fill=frequency))
two_g<-two_g+geom_bar(stat="identity")
two_g<-two_g+labs(title="Bigrams",x="Words",y="Frequency")
two_g<-two_g+theme(axis.text.x=element_text(angle=90))
two_g

thr_g<-ggplot(thr_corpus_sort[1:10,],aes(x=reorder(Word,-frequency),y=frequency,fill=frequency))
thr_g<-thr_g+geom_bar(stat="identity")
thr_g<-thr_g+labs(title="Trigrams",x="Words",y="Frequency")
thr_g<-thr_g+theme(axis.text.x=element_text(angle=90))
thr_g

Conclusion & Next Steps

In this milestone report, I have successfully processed and analyzed a large text corpus consisting of blogs, news, and Twitter data. Through sampling, cleaning, and tokenization, I built unigrams, bigrams, and trigrams from the text data. Exploratory analysis was performed to identify the most frequent word combinations, which will serve as a foundation for the predictive model.

The next steps include:

  1. Building a predictive algorithm – I will develop a model capable of predicting the most likely next word based on the input text.
  2. Shiny App Development – I will create a user-friendly Shiny app that suggests the next word after a phrase is typed.
  3. Final Presentation – I will prepare and present the final product, including a pitch for the app, and deploy it to the “shinyapps.io” platform.