tf_idf <- cleaned_austen %>% count(book, word) %>% bind_tf_idf(word, book, n) %>% arrange(desc(tf_idf)) tf_idf %>% group_by(book) %>% slice_max(tf_idf, n = 3) 4.1. N-grams (Pairs of Words) austen_bigrams <- austen_books() %>% unnest_tokens(bigram, text, token = "ngrams", n = 2) Count common bigrams bigram_counts <- austen_bigrams %>% separate(bigram, into = c("word1", "word2"), sep = " ") %>% filter(!word1 %in% stop_words$word) %>% filter(!word2 %in% stop_words$word) %>% count(word1, word2, sort = TRUE) 4.2. Topic Modeling (Latent Dirichlet Allocation) Using tidytext + topicmodels to discover hidden themes.
# Using bing lexicon (positive/negative) bing_sent <- get_sentiments("bing") sentiment_scores <- cleaned_austen %>% inner_join(bing_sent, by = "word") %>% count(book = austen_books()$book, sentiment) %>% # approximate pivot_wider(names_from = sentiment, values_from = n, values_fill = 0) %>% mutate(net_sentiment = positive - negative)
is an exceptional language for text mining. With a rich ecosystem of packages—most notably the tidytext , quanteda , and tm frameworks—R allows analysts to clean, tokenize, analyze sentiment, model topics, and visualize textual patterns efficiently. Text Mining With R
data(stop_words) cleaned_austen <- tidy_austen %>% anti_join(stop_words, by = "word") Count most common words:
graph LR A[Raw Text] --> B[Preprocessing] --> C[Tokenization] --> D[Stop Word Removal] --> E[Analysis] --> F[Visualization] library(tidyverse) library(tidytext) library(janeaustenr) Load sample text (Jane Austen's books) austen_books <- austen_books() head(austen_books) 3.2. Preprocessing & Tokenization Tokenization splits text into meaningful units (words, sentences, n-grams). tidytext uses unnest_tokens() . - austen_books() %>
1. Introduction In the age of big data, most information exists as unstructured text —emails, social media posts, reviews, news articles, and research papers. Unlike numerical data, text cannot be directly fed into a statistical model. Text mining (or text analytics) is the process of transforming this free-form text into structured, quantifiable data for analysis, pattern discovery, and prediction.
tidy_austen <- austen_books() %>% unnest_tokens(word, text) # one word per row tidy_austen Stop words (the, and, to, of) carry little meaning. tidytext provides get_stopwords() . % filter(n >
word_counts %>% filter(n > 500) %>% ggplot(aes(x = reorder(word, n), y = n)) + geom_col(fill = "steelblue") + coord_flip() + labs(title = "Most Frequent Words in Jane Austen's Novels", x = "Word", y = "Count") + theme_minimal() Sentiment lexicons (e.g., AFINN , bing , nrc ) assign emotional valence to words.