pattern: Pattern for feature, token and keyword matching

In quanteda: Quantitative Analysis of Textual Data

Pattern for feature, token and keyword matching

Description

Pattern(s) for use in matching features, tokens, and keywords through a valuetype pattern.

Arguments

pattern

a character vector, list of character vectors, dictionary, or collocations object. See pattern for details.

Details

The pattern argument is a vector of patterns, including sequences, to match in a target object, whose match type is specified by valuetype. Note that an empty pattern ("") will match "padding" in a tokens object.

character

A character vector of token patterns to be selected or removed. Whitespace is not privileged, so that in a character vector, white space is interpreted literally. If you wish to consider whitespace-separated elements as sequences of tokens, wrap the argument in phrase().

⁠list of character objects⁠

If the list elements are character vectors of length 1, then this is equivalent to a vector of characters. If a list element contains a vector of characters longer than length 1, then for matching will consider these as sequences of matches, equivalent to wrapping the argument in phrase(), except for matching to dfm features where this does not apply.

dictionary

Values in dictionary are used as patterns, for literal matches. Multi-word values are automatically converted into phrases, so performing selection or compounding using a dictionary is the same as wrapping the dictionary in phrase().

collocations

Collocations objects created from quanteda.textstats::textstat_collocations(), which are treated as phrases automatically.

See Also

valuetype, case_insensitive

Examples

# these are interpreted literally
(patt1 <- c("president", "white house", "house of representatives"))
# as multi-word sequences
phrase(patt1)

# three single-word patterns
(patt2 <- c("president", "white_house", "house_of_representatives"))
phrase(patt2)

# this is equivalent to phrase(patt1)
(patt3 <- list(c("president"), c("white", "house"),
               c("house", "of", "representatives")))

# glob expression can be used
phrase(patt4 <- c("president?", "white house", "house * representatives"))

# this is equivalent to phrase(patt4)
(patt5 <- list(c("president?"), c("white", "house"), c("house", "*", "representatives")))

# dictionary with multi-word matches
(dict1 <- dictionary(list(us = c("president", "white house", "house of representatives"))))
phrase(dict1)

quanteda documentation built on June 8, 2025, 9:41 p.m.