Skip to content

Latest commit

 

History

History
133 lines (89 loc) · 4.97 KB

README.md

File metadata and controls

133 lines (89 loc) · 4.97 KB

Current PyPI packages Test Status PyPI - Downloads Supported Platforms

mecab-python3

This is a Python wrapper for the MeCab morphological analyzer for Japanese text. It works with Python 3.6 and greater; if you need to use Python 2.7, use v1.0.2.

issueを英語で書く必要はありません。

Note that Windows wheels require a Microsoft Visual C++ Redistributable, so be sure to install that.

Basic usage

>>> import MeCab
>>> wakati = MeCab.Tagger("-Owakati")
>>> wakati.parse("pythonが大好きです").split()
['python', 'が', '大好き', 'です']

>>> tagger = MeCab.Tagger()
>>> print(tagger.parse("pythonが大好きです"))
python  python  python  python  名詞-普通名詞-一般
                        助詞-格助詞
大好き  ダイスキ        ダイスキ        大好き  形状詞-一般
です    デス    デス    です    助動詞  助動詞-デス     終止形-一般
EOS

The API for mecab-python3 closely follows the API for MeCab itself, even when this makes it not very “Pythonic.” Please consult the official MeCab documentation for more information.

Installation

Binary wheels are available for MacOS X, Linux, and Windows (64bit) are installed by default when you use pip:

pip install mecab-python3

These wheels include an internal (statically linked) copy of the MeCab library, but not dictionary. In order to use MeCab you'll need to install a dictionary. unidic-lite is a good one to start with:

pip install unidic-lite

To build from source using pip,

pip install --no-binary :all: mecab-python3

Dictionaries

There are many different dictionaries available for MeCab. These UniDic packages, which include slight modifications for ease of use, are recommended:

  • unidic: The full UniDic 2.3.0.
  • unidic-lite: The older, much smaller, but not much less useful 2.1.2.

The dictionaries below are not recommended due to being unmaintained for many years, but they are available for use with legacy applications.

For more details on the differences between dictionaries see here.

Common Issues

If you get a RuntimeError when you try to run MeCab, here are some things to check:

Windows Redistributable

You have to install this to use this package on Windows.

Installing a Dictionary

Run pip install unidic-lite and confirm that works. If that fixes your problem, you either don't have a dictionary installed, or you need to specify your dictionary path like this:

tagger = MeCab.Tagger('-r /dev/null -d /usr/local/lib/mecab/dic/mydic')

Note: on Windows, use nul instead of /dev/null. Alternately, if you have a mecabrc you can use the path after -r.

Specifying a mecabrc

If you get this error:

error message: [ifs] no such file or directory: /usr/local/etc/mecabrc

You need to specify a mecabrc file. It's OK to specify an empty file, it just has to exist. You can specify a mecabrc with -r. This may be necessary on Debian or Ubuntu, where the mecabrc is in /etc/mecabrc.

You can specify an empty mecabrc like this:

tagger = MeCab.Tagger('-r/dev/null -d/home/hoge/mydic')

Using Unsupported Output Modes like -Ochasen

Chasen output is not a built-in feature of MeCab, you must specify it in your dicrc or mecabrc. Notably, Unidic does not include Chasen output format. Please see the MeCab documentation.

Alternatives

  • fugashi is a Cython wrapper for MeCab with a Pythonic interface, by the current maintainer of this library
  • SudachiPy is a modern tokenizer with a maintained dictionary, though it's slower than MeCab
  • KoNLPy is a library for Korean NLP that includes a MeCab wrapper

Licensing

Like MeCab itself, mecab-python3 is copyrighted free software by Taku Kudo [email protected] and Nippon Telegraph and Telephone Corporation, and is distributed under a 3-clause BSD license (see the file BSD). Alternatively, it may be redistributed under the terms of the GNU General Public License, version 2 (see the file GPL) or the GNU Lesser General Public License, version 2.1 (see the file LGPL).