Clean open-ended text responses for analysis — clean_text

Extracts one text/textarea item's responses from a response data frame and applies light, configurable cleaning: lower-casing, punctuation removal, and optional number stripping. Blank and missing responses are dropped rather than kept as empty strings, since they carry no term-frequency signal and would otherwise inflate downstream response counts.

Usage

clean_text_responses(
  data,
  item_id,
  lowercase = TRUE,
  remove_punct = TRUE,
  strip_numbers = FALSE,
  instrument = NULL
)

Arguments

data: A data.frame of responses.
item_id: Character. The text/textarea item's column name.
lowercase: Logical. Lower-case the text. Default TRUE.
remove_punct: Logical. Strip punctuation. Default TRUE.
strip_numbers: Logical. Strip digits. Default FALSE.
instrument: Optional sframe instrument. When supplied, item_id is validated as a "text" or "textarea" item before cleaning.

Value

A character vector of cleaned responses, with an integer "respondent" attribute giving each entry's original row index in data, so quotes extracted later can cite a respondent.

Examples

demo <- sframe_demo_data()
cleaned <- clean_text_responses(demo$responses, "comments")
head(cleaned)
#> [1] "useful service information"            
#> [2] "useful service information"            
#> [3] "would like more sustainability details"
#> [4] "useful service information"            
#> [5] "would like more sustainability details"
#> [6] "clear online content"                  
attr(cleaned, "respondent")[1:5]
#> [1] 2 3 5 7 8