Classify free-text items against a user-supplied taxonomy using TF-IDF

Usage

classify_text(
  corpus,
  taxonomy,
  id_col = "id",
  text_col = "text",
  class_id_col = "class_id",
  class_label_col = "class_label",
  num_leaves = 1,
  method = c("tfidf_sum", "cosine"),
  max_dist = 0.1,
  string_dist = NULL,
  stopwords = NULL
)

Arguments

corpus: A data.frame with at minimum an id column and a text column.
taxonomy: Either: - A named character vector: names = class IDs, values = class labels/descriptions - A data.frame with columns identified by `class_id_col` and `class_label_col`
id_col: Name of the id column in `corpus`. Default "id".
text_col: Name of the text column in `corpus`. Default "text".
class_id_col: Name of the class ID column in `taxonomy` (if data.frame). Default "class_id".
class_label_col: Name of the class label column in `taxonomy` (if data.frame). Default "class_label".
num_leaves: Number of top-matching classes to return per item. Default 1.
method: "tfidf_sum" (mirrors labourR) or "cosine" (improved, length-normalised). Default "tfidf_sum".
max_dist: Maximum string distance for fuzzy token matching (used when `string_dist` is set).
string_dist: String distance method passed to `stringdist::amatch()`. NULL disables fuzzy matching.
stopwords: Character vector of stopwords to remove. NULL uses a built-in English set.

Value

A data.table with columns: <id_col>, class_id, class_label, score