Skip to content

ikegami-yukino/neologdn

Repository files navigation

neologdn

PyPI DownloadsPyPI - VersionPyPI - Python VersionPyPI - License

neologdn is a Japanese text normalizer for mecab-neologd.

The normalization is based on the neologd's rules: https://github.com/neologd/mecab-ipadic-neologd/wiki/Regexp.ja

And also some optional features are added.

Contributions are welcome!

NOTE: Installing this module requires C++11 compiler.

Installation

pip install neologdn

If setuptools is not installed, you must install it:

pip install setuptools

If you encountered the following error:

ERROR: Could not find a version that satisfies the requirement setuptools (from versions: none)

Then execute the following commands to may solve this error:

pip install wheel
pip install --no-build-isolation neologdn

Usage

import neologdn
neologdn.normalize("ハンカクカナ")
# => 'ハンカクカナ'
neologdn.normalize("全角記号!?@#")
# => '全角記号!?@#'
neologdn.normalize("全角記号例外「・」")
# => '全角記号例外「・」'
neologdn.normalize("長音短縮ウェーーーーイ")
# => '長音短縮ウェーイ'
neologdn.normalize("チルダ削除ウェ~∼∾〜〰~イ")
# => 'チルダ削除ウェイ'
neologdn.normalize("いろんなハイフン˗֊‐‑‒–⁃⁻₋−")
# => 'いろんなハイフン-'
neologdn.normalize("   PRML  副 読 本   ")
# => 'PRML副読本'
neologdn.normalize(" Natural Language Processing ")
# => 'Natural Language Processing'
neologdn.normalize("かわいいいいいいいいい", repeat=6)
# => 'かわいいいいいい'
neologdn.normalize("無駄無駄無駄無駄ァ", repeat=1)
# => '無駄ァ'
neologdn.normalize("1995〜2001年", tilde="normalize")
# => '1995~2001年'
neologdn.normalize("1995~2001年", tilde="normalize_zenkaku")
# => '1995〜2001年'
neologdn.normalize("1995〜2001年", tilde="ignore")  # Don't convert tilde
# => '1995〜2001年'
neologdn.normalize("1995〜2001年", tilde="remove")
# => '19952001年'
neologdn.normalize("1995〜2001年")  # Default parameter
# => '19952001年'

Benchmark

# Sample code from
# https://github.com/neologd/mecab-ipadic-neologd/wiki/Regexp.ja#python-written-by-hideaki-t--overlast
import normalize_neologd

%timeit normalize(normalize_neologd.normalize_neologd)
# => 9.55 s ± 29.4 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

import neologdn
%timeit normalize(neologdn.normalize)
# => 6.66 s ± 35.8 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

neologdn is about x1.43 faster than sample code.

details are described as the below notebook: https://github.com/ikegami-yukino/neologdn/blob/master/benchmark/benchmark.ipynb

License

Apache Software License.

CHANGES

0.5.4 (2025-03-15)

  • Support Python 3.13
  • Fix tilde loss after latin and whitespace (Many thanks @a-lucky)

0.5.3 (2024-05-03)

  • Support Python 3.12

0.5.2 (2023-08-03)

  • Support Python 3.10 and 3.11 (Many thanks @polm)

0.5.1 (2021-05-02)

  • Improve performance of shorten_repeat function (Many thanks @yskn67)
  • Add tilde option to normalize function

0.4 (2018-12-06)

  • Add shorten_repeat function, which shortening contiguous substring. For example: neologdn.normalize("無駄無駄無駄無駄ァ", repeat=1) -> 無駄ァ

0.3.2 (2018-05-17)

  • Add option for suppression removal of spaces between Japanese characters

0.2.2 (2018-03-10)

  • Fix bug (daku-ten & handaku-ten)
  • Support mac osx 10.13 (Many thanks @r9y9)

0.2.1 (2017-01-23)

  • Fix bug (Check if a previous character of daku-ten character is in maps) (Many thanks @unnonouno)

0.2 (2016-04-12)

  • Add lengthened expression (repeating character) threshold

0.1.2 (2016-03-29)

  • Fix installation bug

0.1.1.1 (2016-03-19)

  • Support Windows
  • Explicitly specify to -std=c++11 in build (Many thanks @id774)

0.1.1 (2015-10-10)

Initial release.

Contribution

Contributions are welcome! See: https://github.com/ikegami-yukino/neologdn/blob/master/.github/CONTRIBUTING.md

Cited by

Book

  • 山本 和英. テキスト処理の要素技術. 近代科学者. P.41. 2021.

Blog