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() .
word_counts <- cleaned_austen %>% count(word, sort = TRUE) word_counts %>% head(10)
data(stop_words) cleaned_austen <- tidy_austen %>% anti_join(stop_words, by = "word") Count most common words: Text Mining With R
sentiment_scores library(wordcloud) word_counts %>% with(wordcloud(word, n, max.words = 100, colors = brewer.pal(8, "Dark2"))) 3.7. Term Frequency – Inverse Document Frequency (TF-IDF) TF-IDF identifies words that are important to a document within a corpus.
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. % filter(n >
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.
with a bar chart:
This write-up outlines a reproducible workflow for text mining using R, emphasizing tidy data principles. | Package | Purpose | | :--- | :--- | | tidytext | Converts text to tidy data frames (one token per row). Integrates with dplyr , ggplot2 . | | dplyr | Data manipulation (filter, group, mutate). | | ggplot2 | Visualization of text metrics (word frequencies, sentiment scores). | | janeaustenr | Sample texts for practice. | | tidyverse | Meta-package for data science. | | wordcloud | Generates word clouds. | | quanteda | Advanced text analysis (DFM, keywords-in-context). | | tm | Classic text mining (corpus, term-document matrix). | Installation: install.packages(c("tidytext", "tidyverse", "wordcloud", "quanteda")) 3. The Text Mining Workflow A standard text mining pipeline in R consists of these steps:
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. % ggplot(aes(x = reorder(word
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() .
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.