Sentiment analysis in R

Sentiment analysis in R involves processing textual data to determine its emotional tone (positive, negative, or neutral). Here's a step-by-step guide to performing sentiment analysis in R, including libraries and code examples:

1. Installing Necessary Packages

You'll need libraries for text processing, sentiment scoring, and visualization:

# Install required packages

install.packages(c("tidyverse", "tidytext", "textdata", "ggplot2"))

install.packages("wordcloud")  # Optional, for visualization

install.packages("syuzhet")    # For advanced sentiment analysis

2. Loading the Libraries

library(tidyverse)

library(tidytext)

library(textdata)

library(ggplot2)

library(wordcloud)

library(syuzhet)

3. Preparing the Text Data

Start with a text dataset, such as customer reviews, social media posts, or any textual content.

Example Text Data:

text_data <- data.frame(

  id = 1:5,

  text = c(

    "I love this product! It's amazing.",

    "The service was terrible and disappointing.",

    "This is the best experience I've ever had.",

    "I am not happy with the quality of the item.",

    "Neutral feelings about this purchase."

  )

)

4. Tokenizing the Text

Tokenization splits text into individual words or tokens for analysis.

# Tokenize text

library(tidytext)

tokens <- text_data %>%

  unnest_tokens(word, text)

5. Using a Sentiment Lexicon

Use sentiment lexicons like Bing, AFINN, or NRC to assign sentiment scores to words.

Loading the Bing Lexicon:

# Get the Bing lexicon

bing <- get_sentiments("bing")

# Join tokens with the lexicon

sentiment_analysis <- tokens %>%

  inner_join(bing, by = "word") %>%

  count(id, sentiment, sort = TRUE) %>%

  spread(sentiment, n, fill = 0) %>%

  mutate(sentiment_score = positive - negative)

6. Visualizing Sentiment

Sentiment Bar Chart:

ggplot(sentiment_analysis, aes(x = factor(id), y = sentiment_score, fill = sentiment_score > 0)) +

  geom_bar(stat = "identity") +

  labs(title = "Sentiment Analysis by Text ID",

       x = "Text ID",

       y = "Sentiment Score") +

  scale_fill_manual(values = c("red", "green"), guide = "none")

Word Cloud:

tokens %>%

  inner_join(bing, by = "word") %>%

  count(word, sentiment, sort = TRUE) %>%

  with(wordcloud(word, n, max.words = 100, colors = c("red", "green")))

7. Advanced Sentiment Analysis Using Syuzhet

The syuzhet package provides methods for extracting sentiment and analyzing the emotional arcs of narratives.

Example:

# Get sentiments using Syuzhet

syuzhet_scores <- get_nrc_sentiment(text_data$text)

# Add scores to the original data

text_data <- cbind(text_data, syuzhet_scores)

# Visualize the emotion scores

emotion_data <- colSums(syuzhet_scores[, 1:8])  # First 8 columns are emotions

barplot(emotion_data,

        main = "Emotion Distribution",

        col = rainbow(8),

        las = 2,

        ylab = "Count")

8. Custom Sentiment Analysis

You can create your own sentiment lexicon or rules if the pre-built ones don't fit your data.

Example:

# Custom lexicon

custom_lexicon <- data.frame(

  word = c("love", "amazing", "terrible", "disappointing"),

  sentiment = c("positive", "positive", "negative", "negative")

)

# Join custom lexicon with tokens

custom_sentiment_analysis <- tokens %>%

  inner_join(custom_lexicon, by = "word") %>%

  count(id, sentiment, sort = TRUE) %>%

  spread(sentiment, n, fill = 0) %>%

  mutate(sentiment_score = positive - negative)

9. Exporting Results

Save sentiment analysis results for reporting or further use:

write.csv(sentiment_analysis, "sentiment_results.csv", row.names = FALSE)

10. Next Steps

• Handle Negation: Account for words like "not" or "never," which can flip sentiment.

• Fine-tune Lexicons: Modify lexicons to better match your dataset.

• Incorporate Machine Learning: Use libraries like caret or text2vec to build models for sentiment classification.