gibasa

A plain ‘Rcpp’ wrapper of ‘MeCab'

https://github.com/paithiov909/gibasa

Keywords

Repository

A plain ‘Rcpp’ wrapper of ‘MeCab'

README.Rmd

---
output: github_document
---



```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)
pkgload::load_all(export_all = FALSE)
```

# gibasa


[![gibasa status badge](https://paithiov909.r-universe.dev/badges/gibasa)](https://paithiov909.r-universe.dev/gibasa)
![GitHub](https://img.shields.io/github/license/paithiov909/gibasa)
[![R-CMD-check](https://github.com/paithiov909/gibasa/workflows/R-CMD-check/badge.svg)](https://github.com/paithiov909/gibasa/actions)
[![codecov](https://codecov.io/gh/paithiov909/gibasa/branch/main/graph/badge.svg)](https://app.codecov.io/gh/paithiov909/gibasa)
[![CRAN logs badge](https://cranlogs.r-pkg.org/badges/gibasa)](https://cran.r-project.org/package=gibasa)


## Overview

gibasa is a plain 'Rcpp' wrapper for 'MeCab', a morphological analyzer for CJK text.

Part-of-speech tagging with morphological analyzers is useful for processing CJK text data. This is because most words in CJK text are not separated by whitespaces and `tokenizers::tokenize_words` may split them into wrong tokens.

The main goal of gibasa package is to provide an alternative to `tidytext::unnest_tokens` for CJK text data. For this goal, gibasa provides three main functions: `gibasa::tokenize`, `gibasa::prettify`, and `gibasa::pack`.

![flowchart of a text analysis that combines gibasa and other packages](man/figures/tidytext_fig5_1_mod.drawio.png)

- `gibasa::tokenize` takes a TIF-compliant data.frame of corpus, returning tokens as format that known as 'tidy text data', so that users can replace `tidytext::unnest_tokens` with it for tokenizing CJK text.
- `gibasa::prettify` turns tagged features into columns.
- `gibasa::pack` takes a 'tidy text data', typically returning space-separated corpus.

## Installation

You can install binary package via [CRAN](https://cran.r-project.org/package=gibasa) or [r-universe](https://paithiov909.r-universe.dev/gibasa).

```r
## Install gibasa from r-universe repository
install.packages("gibasa", repos = c("https://paithiov909.r-universe.dev", "https://cloud.r-project.org"))

## Or build from source package
Sys.setenv(MECAB_DEFAULT_RC = "/fullpath/to/your/mecabrc") # if necessary
remotes::install_github("paithiov909/gibasa")
```

To use gibasa package requires the [MeCab](https://taku910.github.io/mecab/) library and its dictionary installed and available.

In case using Linux or macOS, you can install them with their package managers, or build and install from the source by yourself.

In case using Windows, use installer [built for 64bit](https://github.com/ikegami-yukino/mecab/releases/tag/v0.996.2). Note that gibasa requires a UTF-8 dictionary, not a Shift-JIS one.

As of v0.9.4, gibasa looks at the file specified by the environment variable `MECABRC` or the file located at `~/.mecabrc`. If the MeCab dictionary is in a different location than the default, create a mecabrc file and specify where the dictionary is located.

For example, to install and use the [ipadic](https://pypi.org/project/ipadic/) from PyPI, run:

```sh
$ python3 -m pip install ipadic
$ python3 -c "import ipadic; print('dicdir=' + ipadic.DICDIR);" > ~/.mecabrc
```

## Usage

### Tokenize sentences

```{r}
res <- gibasa::tokenize(
  data.frame(
    doc_id = seq_along(gibasa::ginga[5:8]),
    text = gibasa::ginga[5:8]
  ),
  text,
  doc_id
)
res
```

### Prettify output

```{r}
gibasa::prettify(res)
gibasa::prettify(res, col_select = 1:3)
gibasa::prettify(res, col_select = c(1, 3, 5))
gibasa::prettify(res, col_select = c("POS1", "Original"))
```

### Pack output

```{r}
res <- gibasa::prettify(res)
gibasa::pack(res)

dplyr::mutate(
  res,
  token = dplyr::if_else(is.na(Original), token, Original),
  token = paste(token, POS1, sep = "/")
) |>
  gibasa::pack() |>
  head(1L)
```

### Change dictionary

IPA, UniDic, [CC-CEDICT-MeCab](https://github.com/ueda-keisuke/CC-CEDICT-MeCab), and [mecab-ko-dic](https://bitbucket.org/eunjeon/mecab-ko-dic/src/master/) schemes are supported.

```{r}
## UniDic 2.1.2
gibasa::tokenize("あのイーハトーヴォのすきとおった風", sys_dic = file.path("mecab/unidic-lite")) |>
  gibasa::prettify(into = gibasa::get_dict_features("unidic26"))


## CC-CEDICT
gibasa::tokenize("它可以进行日语和汉语的语态分析", sys_dic = file.path("mecab/cc-cedict")) |>
  gibasa::prettify(into = gibasa::get_dict_features("cc-cedict"))


## mecab-ko-dic
gibasa::tokenize("하네다공항한정토트백", sys_dic = file.path("mecab/mecab-ko-dic")) |>
  gibasa::prettify(into = gibasa::get_dict_features("ko-dic"))
```

## Build dictionaries

### Build a system dictionary

```{r}
## build a new ipadic in temporary directory
build_sys_dic(
  dic_dir = file.path("mecab/ipadic-eucjp"), # replace here with path to your source dictionary
  out_dir = tempdir(),
  encoding = "euc-jp" # encoding of source csv files
)

## copy the 'dicrc' file
file.copy(file.path("mecab/ipadic-eucjp/dicrc"), tempdir())

dictionary_info(sys_dic = tempdir())
```

### Build a user dictionary

```{r}
## write a csv file and compile it into a user dictionary
writeLines(
  c(
    "月ノ,1290,1290,4579,名詞,固有名詞,人名,姓,*,*,月ノ,ツキノ,ツキノ",
    "美兎,1291,1291,8561,名詞,固有名詞,人名,名,*,*,美兎,ミト,ミト"
  ),
  con = (csv_file <- tempfile(fileext = ".csv"))
)
build_user_dic(
  dic_dir = file.path("mecab/ipadic-eucjp"),
  file = (user_dic <- tempfile(fileext = ".dic")),
  csv_file = csv_file,
  encoding = "utf8"
)

tokenize("月ノ美兎は箱の中", sys_dic = tempdir(), user_dic = user_dic)
```

## License

GPL (>=3).

Owner

GitHub Events

Committers

Last synced: 7 months ago

Issues and Pull Requests

Last synced: 6 months ago

Packages