Spend less time in entry!
- By predicting the next word, this application helps user to save
time in text entry if integrated in text editing softwares.
- This application was developed and is running in R shinyapp, a free
server accessible to everyone.
To start with, we load the required packages and make connection to
the dat. R codes are shown for your information but not run
- Loading packages
- Connecting to data
# Set working directory and load data (done ahead to save time)
fileurl <- "https://d396qusza40orc.cloudfront.net/dsscapstone/dataset/Coursera-SwiftKey.zip"
download.file(fileurl, destfile ="Coursera-SwiftKey.zip")
unzip("Coursera-SwiftKey.zip")
setwd("Week 2/final/en_US/")
blogs <- read_lines("en_US.blogs.txt", skip_empty_rows = TRUE)
news <- read_lines("en_US.news.txt", skip_empty_rows = TRUE)
twitter <- read_lines("en_US.twitter.txt", skip_empty_rows = TRUE)
all_text <- c(blogs, news, twitter)
rm(blogs, news, twitter)
Tidying data
- Some basic transformations like removing internet characters and
symbols, converting all words into lower case, etc. are processed using
a randomly selected sample
set.seed("12345")
trainingData <- linesInFile[rbinom(fileNLine, 1, 0.01)==1]
corpusFeeds <- VCorpus(VectorSource(trainingData))
corpusFeeds <- tm_map(corpusFeeds, removePunctuation) # remove punctuation
corpusFeeds <- tm_map(corpusFeeds, content_transformer(tolower)) # put in lower char
corpusFeeds <- tm_map(corpusFeeds, content_transformer(remove_chars)) # remove internet chars
corpusFeeds <- tm_map(corpusFeeds, removeWords, stopwords("english")) # remove English stop words
corpusFeeds <- tm_map(corpusFeeds, content_transformer(remove_symbols)) #remove symbols
corpusFeeds <- tm_map(corpusFeeds, stripWhitespace) # remove extra spaces
filerul2 <- "http://www.bannedwordlist.com/lists/swearWords.txt"
download.file(filerul2, destfile = "badwords.txt")
badwords <- readLines("badwords.txt")
profanity <- VectorSource(badwords)
corpusFeeds <- tm_map(corpusFeeds, removeWords, profanity)
Descriptive analysis
I chose two ways to present interesting findings. The first one is
with word to cloud to see which words are the most frequently used 
The second is with NGrams histogram to understand which combination
of words are frequently used. 
