sentimentanalysis.R

library(twitteR)
library(curl)
library(tm)
library(ggplot2)

#Read in the Sentiment Dictionary which we will use later.
dict <- read.csv("sentimentdictionary.csv",stringsAsFactors = F)

#The following keys are unique to each user, you can get yours by signing up for an app "https://apps.twitter.com/".
consumer_key <- "EJfvJef43wxK5SytXH9ugB1MP"
consumer_secret <- "2KxcMOsl6LEWD9H4tXSN3csWL9ndKKK39ibJiYol3BCOKodUWm"
access_token <- "807488357806678016-9gIqjHwDSsoNTwqDG9WN8UvoUTdgtaP"
access_secret <- "E3PcW7hmyvlNmRwk3ZaC4rE2R09YLaM1kw0AVZs9oP6R2"

#Create a connection with the Twitter API. 
setup_twitter_oauth(consumer_key,consumer_secret,access_token,access_secret)

sentiment <- function(word,tweets_num=500) {
  
  #Fetches tweets from twitter and stores them in a data frame.
  string <- paste0(word,"-filter:retweets")
  tweets <- searchTwitter(string,n=tweets_num, lang="en",resultType = "recent")
  tweetsdf <- twListToDF(tweets)
  tweetsdf <- data.frame(text=tweetsdf$text)
  
  #Text processing and clean up. Remove unnecessary punctuation and characters. 
  tweetsdf$text <- gsub("@[[:alnum:]]+ *","",tweetsdf$text)
  tweetsdf$text <- gsub("http[[:alnum:][:punct:]]+ *","",tweetsdf$text)
  tweetsdf$text <- gsub("[^[:alpha:][:space:]]*","",tweetsdf$text)
  tweetsdf$text <- gsub("[[:punct:]]","",tweetsdf$text)
  tweetsdf$text <- gsub("\n","",tweetsdf$text)
  tweetsdf$text <- tolower(tweetsdf$text)
  
  #Matches the tweets with words in the Sentiment Dictionary and assigns the appropriate score to each tweet.
  sentiscores <- unlist(lapply(tweetsdf$text,function(x){
    sum(as.numeric(dict$Score[match(unlist(strsplit(x," ")),dict$Word)]),na.rm=T)
  }))
 
  sentiscores[is.na(sentiscores)] <- as.numeric(0)
  sentiscoresdf <- data.frame(Scores=sentiscores)
  sentiscoresdf$color <- ifelse(sentiscoresdf$Scores > 0,"positive",ifelse(sentiscoresdf$Scores == 0,"neutral","negative"))
  
  #Writes the resulting data frame to a file in the current working irectory and displays a bar plot of the Sentiment Score. 
  list(write.csv(cbind(tweetsdf,Scores=sentiscores),paste0(word,"scores.csv"),row.names = F),
            ggplot(data=sentiscoresdf,aes(x=Scores))+geom_bar(aes(fill=color))+
         ylab("No. Of Tweets")+xlab("Sentiment Score")+theme_minimal()+geom_text(stat="count",aes(label=..count..),vjust=-.5)
          +scale_x_continuous(breaks=c(-7:7))+scale_fill_manual(values = c(positive="steelblue",negative="firebrick1",neutral="yellowgreen")))
  
}


#After loading the script call the function "sentiment" which has two arguments, "word" which is the search string nd "tweets_num" which is the number of tweets to be fetched.