R/interface-sudachi.R

In paithiov909/krikinton: rJava Wrapper of Sudachi and Kintoki

#' Rebuild Dictionary
#'
#' @param json_string character scalar. valid json string
#' @return a binding to the instance of
#' `com.worksap.nlp.sudachi.JapaneseDictionary` is returned invisibly.
#'
#' @export
rebuild_dictionary <- function(json_string) {
  Dict(rJava::.jnew("com.worksap.nlp.sudachi.JapaneseDictionary", json_string))
  return(invisible(Dict()))
}

#' Rebuild Sudachi Tokenizer
#'
#' @return a binding to the instance of
#' `com.worksap.nlp.sudachi.JapaneseTokenizer` is returned invisibly.
#'
#' @export
rebuild_tokenizer <- function() {
  Tokenizer(rJava::J(Dict(), "create"))
  return(invisible(Tokenizer()))
}

#' Call Sudachi Tagger
#'
#' @param chr character vector
#' @return tibble
#'
#' @export
sudachi <- function(chr) {
  if (!is.character(chr) || is.na(chr)) {
    message("Invalid characters provided. Chr must be a character vector, not NA_character_.")
    return(invisible(tibble()))
  } else {
    texts <- stri_enc_toutf8(chr)
    if (is.null(Tokenizer())) rebuild_tokenizer()

    results <- imap(texts, function(sent, idx) {
      if (!is.na(sent)) {
        morphemes <- Tokenizer()$tokenize(sent)
        morph_size <- morphemes$size()

        if (!rJava::is.jnull(morph_size)) {
          df <- lapply(seq_len(morph_size), function(i) {
            morpheme <- morphemes$get(i - 1L)
            features <- morpheme$partOfSpeech()
            features <- sapply(seq_len(features$size()), function(itr) {
              return(features$get(itr - 1L))
            }, simplify = TRUE)
            res <- list(
              Surface = morpheme$surface(),
              Reading = morpheme$readingForm(),
              Normalized = morpheme$normalizedForm(),
              Original = morpheme$dictionaryForm(),
              Feature = paste(features, collapse = ",")
            )
            res <- lapply(res, function(elem) {
              Encoding(elem) <- "UTF-8"
              return(elem)
            })
            res <- list(
              sentence_id = idx,
              res,
              isOOV = morpheme$isOOV()
            )
            return(as.data.frame(res, stiringsAsFactors = FALSE))
          })
          df <- map_dfr(
            df,
            ~ separate(
              .,
              col = "Feature",
              into = c(
                "POS1",
                "POS2",
                "POS3",
                "POS4",
                "X5StageUse1",
                "X5StageUse2"
              ),
              sep = ","
            )
          )
          return(mutate(
            as_tibble(df),
            across(
              where(is.character),
              ~ dplyr::if_else(
                . == "*",
                NA_character_,
                .
              )
            )
          ))
        } else {
          return(tibble())
        }
      } else {
        return(tibble())
      }
    })

    return(map_dfr(
      discard(results, ~ is_empty(.)),
      ~.
    ))
  }
}