R/limpiar_repeat_chars.R

In jpcompartir/LimpiaR: LimpiaR

#' Clean repeated charaaaacters
#'
#' Removes multiple vowels (holaaaa) and normalises common laughing patterns (jajaja, jejeje, ajajaaaaja).
#' Useful for visualisations, and reducing the overall number of tokens present in the text variable.
#'
#' @param df Name of the Data Frame or Tibble object
#' @param text_var Name of the text variable/character vector. Default is mention_content
#'
#' @return Data Frame or Tibble object with most repeat vowels & laughing patterns removed from the text variable
#' @export
#'
#' @examples
#' limpiar_examples %>% dplyr::select(mention_content)
#'
#' limpiar_examples %>% limpiar_repeat_chars() %>% dplyr::select(mention_content)
#'
#'
limpiar_repeat_chars <- function(df, text_var = mention_content){
  #Checks there is a j + (e or i or a) and at least characters made from j + a i e.
  laughing_regex <- "\\b(?=.+j)((?=.+j)|(?=.+a)|(?=.+e)|(?=.+i))[(j|a|i|e)+(j|a|i|e)+]{4,}"
  #Replaces messy laughing strings with "jaja"
  laughing_replacement <- "jaja"

  #Creates a capture group for any vowel seen 2 or more times.
  #This part needs some work. holaaaaa will become hola, but holaa will not become hola.
  #If we want holaa to become hola, then we have to deal with annoying outcomes like:
  #leer' > 'ler', google > gogle
  #for now the best trade-off seems to be like this, but we may need to re-work.
  repeat_vowels_regex <-  "([a|e|i|o|u|y])\\1{2,}"

  #replaces two or more of the same vowel with one of that same vowel
  repeat_vowels_replacement <- "\\1"

  dplyr::mutate(df,
                {{ text_var }} := stringr::str_replace_all({{ text_var }}, repeat_vowels_regex, repeat_vowels_replacement),
                {{ text_var }} := stringr::str_replace_all({{ text_var }}, laughing_regex, laughing_replacement))
}