Sentiment Analysis

Sentiment analysis of paper

# load in the libraries we'll need
library(tidyverse)
library(tidytext)
library(glue)
library(stringr)
library("corpus")
library(syuzhet)
library(sentimentr)

# get a list of the files in the input directory
files <- readLines("/Users/john/Downloads/Commentary.txt")
fileText <- glue(read_file(files))
# remove any dollar signs (they're special characters in R)
fileText <- gsub("\\$", "", fileText) 

# tokenize
tokens <- data_frame(text = fileText) %>% unnest_tokens(word, text)

Calculate overall sentiment: (positive words are assigned +1, negative words are assigned -1, the overall tone is neutral; a score of +4)

tokens %>%
  inner_join(get_sentiments("bing")) %>% # pull out only sentiment words
  count(sentiment) %>% # count the # of positive & negative words
  spread(sentiment, n, fill = 0) %>% # made data wide rather than narrow
  mutate(sentiment = positive - negative) # # of positive words - # of negative words

Joining, by = "word"

s_v <- syuzhet::get_sentences(fileText)
s_v_sentiment <- as.data.frame(syuzhet::get_sentiment(s_v,method = "bing"))
colnames(s_v_sentiment) <- "sentiment"
s_v_sentiment <- tibble::rowid_to_column(s_v_sentiment, "sentence")
s_v_sentiment$neg <- s_v_sentiment$sentiment<0
s_v_sentiment$pos <- s_v_sentiment$sentiment>0


DF <- data.frame(x1 = head(s_v_sentiment$sentence, -1), x2 = tail(s_v_sentiment$sentence, -1) , 
                 y1 = head(s_v_sentiment$sentiment, -1), y2 = tail(s_v_sentiment$sentiment, -1))
DF$col <- DF$y1>=0


ggplot(DF, aes(x=x1, y=y1, xend = x2, yend = y2, colour=col))+geom_segment()+geom_hline(yintercept = 0)+
  scale_x_continuous(labels=as.character(s_v_sentiment$sentence),breaks=s_v_sentiment$sentence)+
  theme_minimal()+xlab("sentence")+ylab("sentiment")+ theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

now to pull out the negative sentences identified above…

s_v[s_v_sentiment$neg==TRUE]

[1] "In this commentary, we address limitations of the conclusions of the target article that stem from the relatively superficial operationalization of bilingualism."
[2] "Instead of reporting the presence of the variable, we recorded the absence of such information."                                                                  
[3] "An \"upset plot\" (Figure 1) shows that 28% of the studies reported none of the 14 variables related to bilingualism."                                            

# get the sentiment from the first text: 
tokens %>%
  inner_join(get_sentiments("bing")) %>% # pull out only sentiment words
  count(sentiment) %>% # count the # of positive & negative words
  spread(sentiment, n, fill = 0) %>% # made data wide rather than narrow
  mutate(sentiment = positive - negative) # # of positive words - # of negative owrds

Joining, by = "word"

# get words already get_sentiments("bing")

(sentiment <- tokens %>%
inner_join(get_sentiments("bing")) %>%
  count(word, sentiment, sort = TRUE))

Joining, by = "word"

Plot top 20 influential positive and negative words

library(ggplot2)

sentiment %>%
  group_by(sentiment) %>%
  top_n(20) %>%
  ungroup() %>%
  mutate(word = reorder(word, n)) %>%
  ggplot(aes(n, word, fill = sentiment)) +
  geom_col(show.legend = FALSE) +
  facet_wrap(~sentiment, scales = "free_y") +
  labs(x = "Contribution to sentiment",
       y = NULL)

Selecting by n