Skip to main content

CJK Line Break Organizer

Project description

English uses spacing and hyphenation as cues to allow for beautiful and legible line breaks. Certain CJK languages have none of these, and are notoriously more difficult. Breaks occur randomly, usually in the middle of a word. This is a long standing issue in typography on the web, and results in degradation of readability.

Budou automatically translates CJK sentences into organized HTML code with lexical chunks wrapped in non-breaking markup so as to semantically control line breaks. Budou uses word segmenter to analyze the input sentence, and it concatenates proper words in order to produce meaningful chunks utilizing part-of-speech (pos) tagging and syntactic information. Processed chunks are wrapped with SPAN tag, so semantic units will no longer be split at the end of a line by specifying their display property as inline-block in CSS.


The package is listed in the Python Package Index (PyPI), so you can install it with pip:

$ pip install budou

How to use

Budou outputs a HTML snippet wrapping chunks with span tags like below.

<span><span class="ww">常に</span><span class="ww">最新、</span>
<span class="ww">最高の</span><span class="ww">モバイル。</span></span>

Semantic chunks in the output HTML will not be split at the end of line by conditioning each span tag with display: inline-block in CSS.

.ww {
  display: inline-block;

By using the output HTML from Budou and above CSS conditioning, the sentence on your webpage will be rendered with legible line breaks like shown below.

Using as a command-line app

You can process your texts by running budou command like below.

$ budou 渋谷のカレーを食べに行く。

The output is:

<span><span class="ww">渋谷の</span><span class="ww">カレーを</span>
<span class="ww">食べに</span><span class="ww">行く。</span></span>

You can also configure the command with optional parameters. For example, you can change the backend segmenter to MeCab and change the class name to wordwrap by running the command below.

$ budou 渋谷のカレーを食べに行く。 --segmenter=mecab --classname=wordwrap

The output is:

<span><span class="wordwrap">渋谷の</span><span class="wordwrap">カレーを</span>
<span class="wordwrap">食べに</span><span class="wordwrap">行く。</span></span>

Run help command budou -h to see other available options.

Using programmatically

You can use budou.parse method in your python scripts.

import budou
results = budou.parse('渋谷のカレーを食べに行く。')
# <span><span class="ww">渋谷の</span><span class="ww">カレーを</span>
# <span class="ww">食べに</span><span class="ww">行く。</span></span>

You can also make a parser instance to reuse the segmenter backend with the same configuration. If you want to integrate Budou into your web development framework in a form of a custom filter or a build process, this would be the way to go.

import budou
parser = budou.get_parser('mecab')
results = parser.parse('渋谷のカレーを食べに行く。')
# <span><span class="ww">渋谷の</span><span class="ww">カレーを</span>
# <span class="ww">食べに</span><span class="ww">行く。</span></span>

for chunk in results['chunks']:
# 渋谷の 名詞
# カレーを 名詞
# 食べに 動詞
# 行く。 動詞

(deprecated) authenticate method

authenticate, which have been the method to create a parser in the previous releases, is now deprecated from this release. authenticate method is now a wrapper of get_parser method and returns a parser with Google Cloud Natural Language API segmenter backend. The method is still available, but it may be removed in a future release.

import budou
parser = budou.authenticate('/path/to/credentials.json')

# This is equivalent to:
parser = budou.get_parser(
    'nlapi', credentials_path='/path/to/credentials.json')

Available backend segmenters

You can choose the backend segmenter considering your environmental needs. Currently, the segmenters below are supported.

Name Identifier Supported Languages
Google Cloud Natural Language API nlapi Chinese, Japanese, Korean
MeCab mecab Japanese
TinySegmenter tinysegmenter Japanese

Specify the segmenter when you run budou command or load a parser. For example, you can run budou command with MeCab segmenter by passing --segmenter=mecab parameter like below.

$ budou 今日も元気です --segmenter=mecab

You can pass segmenter parameter when you load a parser otherwise.

import budou
parser = budou.get_parser('mecab')

If no segmenter is specified, Google Cloud Natural Language API is used as the default segmenter.

Google Cloud Natural Language API Segmenter

Google Cloud Natural Language API ( (NL API) analyzes the input sentence using machine learning technology. The API can extract not only syntax but also entities included in the sentence, which can be used for better quality segmentation (see more at Entity mode). Since this is a simple REST API, you don’t need to maintain the dictionary and can support multiple languages with one single source.

Supported languages

  • Simplified Chinese (zh)
  • Traditional Chinese (zh-Hant)
  • Japanese (ja)
  • Korean (ko)

For those considering to use Budou for Korean sentences, please also refer to Korean support section.


NL API requires authentication to use. Firstly, create a Google Cloud Platform project and enable Cloud Natural Language API. Billing also need to be enabled for the project. Then, download a credentials file for a service account by accessing Google Cloud Console and navigating through “API & Services” > “Credentials” > “Create credentials” > “Service account key” > “JSON”.

Budou will handle authentication once the path to the credentials file is set as GOOGLE_APPLICATION_CREDENTIALS environment variable.

$ export GOOGLE_APPLICATION_CREDENTIALS='/path/to/credentials.json'

You can also pass the path to the credentials file when you initialize the parser.

parser = budou.get_parser(
    'nlapi', credentials_path='/path/to/credentials.json')

NL API segmenter uses Syntax Analysis and incurs cost according to monthly usage. NL API has free quota to start testing the feature at free of cost. Please refer to for more detailed pricing information.

Caching system

Parsers on NL API segmenter cache responses from the API in order to save unnecessary requests to the API and make the processing faster. If you want to force refresh the cache, set use_cache as False.

parser = budou.parse('明日は晴れるかな', segmenter='nlapi', use_cache=False)

In Google App Engine Python 2.7 Standard Environment, Budou tries to use memcache service in order to cache the outputs efficiently across instances. If not, Budou creates a cache file in python pickle format in your file system.

Entity mode

Default parser only uses results from Syntactic Analysis for parsing, but you can also utilize results from Entity Analysis by specifying use_entity=True. Entity Analysis will improve the accuracy of parsing for some phrases, especially proper nouns, so it is recommended to use if your target sentences include a name of an individual person, place, organization etc.

Please note that Entity Analysis will results in additional pricing because it requires additional requests to NL API. For more detail about API pricing, please refer to for more detail.

import budou
# Without Entity mode (default)
result = budou.parse('六本木ヒルズでご飯を食べます。', use_entity=False)
# <span class="ww">六本木</span><span class="ww">ヒルズで</span>
# <span class="ww">ご飯を</span><span class="ww">食べます。</span>

# With Entity mode
result = budou.parse('六本木ヒルズでご飯を食べます。', use_entity=True)
# <span class="ww">六本木ヒルズで</span>
# <span class="ww">ご飯を</span><span class="ww">食べます。</span>

MeCab Segmenter

MeCab ( is an open source text segmentation library for Japanese language. MeCab Segmenter does not require any billed API calling unlike Google Cloud Natural Language API Segmenter, so you can process the sentences without internet connection free. You can also customize the dictionary by building your own.

Supported languages

  • Japanese


You need to have MeCab installed to use MeCab segmenter in Budou. You can install MeCab with IPA dictionary by running

$ make install-mecab

in the project’s home directory after cloning this repository.

TinySegmenter-based Segmenter

TinySegmenter ( is a compact Japanese tokenizer originally created by (c) 2008 Taku Kudo. This tokenizes sentences by a combination of pattern matchings carefully designed using machine learning. It means you can use this backend without any additional setup!

Supported languages

  • Japanese

Korean support

Korean has spaces between chunks, so you can organize line breaking simply by putting word-break: keep-all in your CSS. We recommend you to use that technique instead of using Budou.

Where to use

Budou is designed to be used mostly in eye-catching sentences such as titles and headings assuming split chunks would be more stood out negatively in larger typography.


Some screen reader software read wrapped chunks one by one when Budou is applied, which may degrades user experience for those who need audio support. You can attach any attribute to the output chunks to enhance accessibility. For example, you can make screen readers to read undivided sentences by combining aria-describedby and aria-label attribute in the output.

<p id="description" aria-label="やりたいことのそばにいる">
  <span class="ww" aria-describedby="description">やりたい</span>
  <span class="ww" aria-describedby="description">ことの</span>
  <span class="ww" aria-describedby="description">そばに</span>
  <span class="ww" aria-describedby="description">いる</span>

This functionality is currently down due to html5lib sanitizer’s behavior which strips aria related attributes from the output HTML. The progress on this issue is tracked at


Shuhei Iitsuka


This library is authored by a Googler and copyrighted by Google, but is not an official Google product.


Copyright 2018 Google LLC

Licensed under the Apache License, Version 2.0 (the “License”); you may not use this file except in compliance with the License. You may obtain a copy of the License at

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
budou-0.9.3-py2.py3-none-any.whl (29.2 kB) Copy SHA256 hash SHA256 Wheel py2.py3

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page