Package {textclassificationtutorial}

Title: Reproducible Text Classification Workflows
Version: 0.1.2
Description: Dependency-light tools and tutorials for teaching reproducible text classification. The package covers HTML text extraction, sentence segmentation, text preprocessing, document-term matrices, TF-IDF, keyword extraction, cosine similarity, stratified cross-validation, classification metrics, and a multinomial Naive Bayes classifier. It modernizes the code accompanying Kobayashi, V. B., Berkers, H. A., Mol, S. T. Kismihok, G., and Den Hartog, D. N. (2017) <doi:10.1177/1094428117719322> The package replaces the original scripts in the paper.
License: Apache License (≥ 2)
URL: https://github.com/vkobayashi/textclassificationtutorial
BugReports: https://github.com/vkobayashi/textclassificationtutorial/issues
Depends: R (≥ 4.1.0)
Suggests: knitr, rmarkdown, testthat (≥ 3.0.0), xml2
VignetteBuilder: knitr
Config/testthat/edition: 3
Encoding: UTF-8
Config/roxygen2/version: 8.0.0
NeedsCompilation: no
Packaged: 2026-08-20 21:22:17 UTC; vladimerkobayashi
Author: Vladimer Kobayashi [aut, cre], Stefan Mol [aut], Gabor Kismihok [aut]
Maintainer: Vladimer Kobayashi <vladimer.kobayashi@gmail.com>
Repository: CRAN
Date/Publication: 2026-08-26 19:30:02 UTC

Calculate binary classification metrics

Description

Calculate binary classification metrics

Usage

classification_metrics(truth, estimate, positive)

Arguments

truth

Vector of true classes.

estimate

Vector of predicted classes.

positive

Value identifying the positive class.

Value

A one-row data frame containing confusion counts, accuracy, balanced accuracy, precision, recall, specificity, and F1.

Examples

classification_metrics(c("task", "task", "other"), c("task", "other", "other"),
                       positive = "task")

Compute cosine similarity

Description

Compute cosine similarity

Usage

cosine_similarity(x, y = NULL)

Arguments

x

Numeric matrix whose rows are observations.

y

Optional numeric matrix with the same columns as x. When omitted, computes all pairwise similarities among rows of x.

Value

A numeric similarity matrix. Similarities involving zero vectors are returned as NA.

Examples

x <- rbind(a = c(1, 1, 0), b = c(1, 0, 0), c = c(0, 0, 1))
cosine_similarity(x)

Construct a document-term matrix

Description

Construct a document-term matrix

Usage

document_term_matrix(
  text,
  document_ids = NULL,
  binary = FALSE,
  min_doc_freq = 1L,
  max_doc_prop = 1
)

Arguments

text

Character vector containing one preprocessed document per item.

document_ids

Optional unique document identifiers.

binary

Logical; store term presence instead of term frequency?

min_doc_freq

Minimum number of documents in which a term must occur.

max_doc_prop

Maximum proportion of documents in which a term may occur.

Value

A numeric matrix with class text_dtm.

Examples

docs <- preprocess_text(c("data science", "data analysis", "science"))
document_term_matrix(docs)

Extract text from a directory of HTML files

Description

Extract text from a directory of HTML files

Usage

extract_html_dir(path, pattern = "\\.html?$", recursive = FALSE, ...)

Arguments

path

Directory containing HTML files.

pattern

File-name regular expression.

recursive

Logical; search recursively?

...

Passed to extract_html_text().

Value

A data frame with document_id, path, and text.

Extract readable text from HTML

Description

Extracts text from an HTML file or character string. When the suggested xml2 package is installed, simple CSS selectors (tag, .class, ⁠#id⁠, or tag.class) or XPath can target a specific part of the page. A dependency-free fallback strips markup from the full document.

Usage

extract_html_text(
  x,
  selector = NULL,
  xpath = NULL,
  collapse = "\n",
  trim = TRUE
)

Arguments

x

Path to an HTML file or a length-one HTML character string.

selector

Optional CSS selector.

xpath

Optional XPath expression. Supply at most one of selector and xpath.

collapse

Character used to join matched nodes.

trim

Logical; normalize whitespace and trim the result?

Value

A length-one character vector containing extracted text.

Examples

html <- "<html><body><h1>Analyst</h1><p>Analyze data.</p></body></html>"
extract_html_text(html)

Extract top TF-IDF keywords

Description

Extract top TF-IDF keywords

Usage

extract_keywords(x, n = 1L, already_tfidf = FALSE)

Arguments

x

A document-term matrix or TF-IDF matrix.

n

Number of keywords per document.

already_tfidf

Logical; is x already weighted?

Value

A data frame with document, rank, term, and weight.

F-measure

Description

F-measure

Usage

f_measure(precision, recall, beta = 1)

Arguments

precision

Numeric precision.

recall

Numeric recall.

beta

Relative weight assigned to recall.

Value

Numeric F-measure.

Fit a multinomial Naive Bayes text classifier

Description

Fit a multinomial Naive Bayes text classifier

Usage

fit_naive_bayes(x, y, laplace = 1, prior = NULL)

Arguments

x

Nonnegative numeric document-term matrix.

y

Class labels with one value per row of x.

laplace

Nonnegative additive smoothing parameter.

prior

Optional named class probabilities.

Value

An object of class text_nb.

Examples

x <- rbind(c(3, 0), c(2, 0), c(0, 3), c(0, 2))
colnames(x) <- c("analysis", "care")
model <- fit_naive_bayes(x, c("data", "data", "health", "health"))
predict(model, x)

Normalize text for document-term analysis

Description

Normalize text for document-term analysis

Usage

preprocess_text(
  text,
  lowercase = TRUE,
  remove_punctuation = TRUE,
  remove_numbers = TRUE,
  stopwords = character(),
  min_token_length = 1L
)

Arguments

text

Character vector.

lowercase

Logical; convert text to lowercase?

remove_punctuation

Logical; replace punctuation with spaces?

remove_numbers

Logical; replace digits with spaces?

stopwords

Optional character vector of words to remove.

min_token_length

Minimum number of characters per token.

Value

A character vector of normalized documents.

Examples

preprocess_text(
  c("Analyze the data!", "Present 2 reports."),
  stopwords = c("the")
)

Read a one-label-per-line file

Description

Read a one-label-per-line file

Usage

read_label_file(path, type = c("character", "integer", "numeric", "factor"))

Arguments

path

Path to a text file.

type

Return labels as character, integer, numeric, or factor.

Value

A vector of labels.

Split text into sentences

Description

A lightweight sentence segmenter suitable for tutorials and clean prose. It splits at terminal punctuation followed by whitespace, and optionally at line breaks and vertical bars. For production multilingual segmentation, use a dedicated NLP tokenizer.

Usage

split_sentences(
  text,
  split_lines = TRUE,
  keep_punctuation = TRUE,
  drop_empty = TRUE
)

Arguments

text

Character vector.

split_lines

Logical; treat line breaks and | as boundaries?

keep_punctuation

Logical; retain terminal punctuation?

drop_empty

Logical; remove empty results?

Value

A character vector of sentences.

Examples

split_sentences("Analyze data. Present results! Work with teams?")

Create stratified cross-validation folds

Description

Create stratified cross-validation folds

Usage

stratified_folds(y, k = 5L, repeats = 1L, seed = NULL)

Arguments

y

Class labels.

k

Number of folds.

repeats

Number of repeated fold sets.

seed

Optional random seed. The caller's random-number state is restored.

Value

A list of integer test-set indices with class text_folds.

Calculate TF-IDF weights

Description

Calculate TF-IDF weights

Usage

tf_idf(dtm, normalize = c("length", "max", "none"), smooth_idf = TRUE)

Arguments

dtm

Numeric document-term matrix.

normalize

Term-frequency normalization: document length, maximum frequency, or none.

smooth_idf

Logical; use smoothed inverse document frequency?

Value

A numeric matrix of TF-IDF weights.