Predict Sentiment Using Bag-of-Words (Bing) Dictionary

Example Qbit to predict the sentiment of sentences based on the bag-of-words methodology and the Bing sentiment dictionary.

library(dplyr)
library(tidyr)
library(tidytext)
library(SnowballC)
library(stopwords)

# See also https://smltar.com/stemming.html#how-to-stem-text-in-r
get_sentiments_stem <- function(dict) {
  get_sentiments(dict) %>%
    mutate(stem = wordStem(word)) %>%
    distinct(stem, sentiment)
}

sentiment_list <- function(sentence) {
    tibble(text = sentence) %>%
    unnest_tokens(word, text) %>%
    anti_join(get_stopwords()) %>%
    mutate(stem = wordStem(word)) %>%
    left_join(get_sentiments_stem("bing"), by = "stem")
}

sentiment_score <- function(sentence) {
    sentiment_list(sentence) %>%
    mutate(score = ifelse(sentiment == "positive", 1, -1))
}

polarity <- function(sentence) {
    df <- sentiment_score(sentence)
    sum(df$score, na.rm = TRUE) / sum(abs(df$score), na.rm = TRUE)
}

sentence_positive <- "I'm so happy that the sun is shining and the birds singing!"
sentence_negative <- "I had the worst day of my live. I'm very sad and depressed."

sentiment_list(sentence_positive)
sentiment_list(sentence_negative)

sentiment_score(sentence_positive)
sentiment_score(sentence_negative)

polarity(sentence_positive)
polarity(sentence_negative)