Skip to contents

A small, fixed list of English function words (articles, prepositions, conjunctions, pronouns, auxiliaries) for the stop_words argument of text_hypergraph(). Deliberately minimal and versioned with the package: corpus-specific boilerplate (e.g. "study", "results" in an abstract corpus) should be added by the caller, as in c(stop_words_en(), "study", "results").

Usage

stop_words_en()

Value

A sorted character vector of lowercase English function words.

References

Manning, C. D., Raghavan, P., & Schütze, H. (2008). Introduction to Information Retrieval. Cambridge University Press. doi:10.1017/CBO9780511809071

Examples

hg <- text_hypergraph(
  c(a = "the salt and the soup", b = "the soup and the stars"),
  stop_words = stop_words_en()
)
hg_get(hg, what = "vocabulary")
#>    word count doc_freq
#> 1  salt     1        1
#> 2  soup     2        2
#> 3 stars     1        1