This is a pre-production deployment of Warehouse. Changes made here affect the production instance of PyPI (
Help us improve Python packaging - Donate today!

Python bindings around Google Chromium's embedded compact language detection library (CLD2)

Project Description

This package contains the CLD (Compact Language Detection) library as maintained by Dick Sites ( The first fork was done at revision r161. It also contains python bindings that were originally created by Mike McCandless. The bindings here differ than upstream by making the full set of languages the default option supporting more than 165 languages!

The goal of this project is to consolidate the upstream library with its bindings, so the user can pip install one package instead of two.

The LICENSE is the same as Chromium’s LICENSE and is included in the LICENSE file for reference.



$ git clone
$ cd pycld2
$ ./ install


import pycld2 as cld2

detectedLangName, detectedLangCode, isReliable, textBytesFound, details = cld2.detect("This is my sample text", pickSummaryLanguage=True, removeWeakMatches=False)
print '  detected: %s' % detectedLangName
print '  reliable: %s' % (isReliable != 0)
print '  textBytes: %s' % textBytesFound
print '  details: %s' % str(details)

# The output look lie so:
#  detected: ENGLISH
#  reliable: True
#  textBytes: 25
#  details: [('ENGLISH', 'en', 64, 20.25931928687196), ('FRENCH', 'fr', 36, 8.221993833504625)]


First, you must get your content (plain text or HTML) encoded into UTF8 bytes. Then, detect like this:

topLanguageName, topLanguageCode, isReliable, textBytesFound, details = cld2.detect(bytes)

The code and name of the top language is returned. isReliable is True if the top language is much better than 2nd best language. textBytesFound tells you how many actual bytes CLD analyzed (after removing HTML tags, collapsing areas of too-many-spaces, etc.). details has an entry per top 3 languages that matched, that includes the percent confidence of the match as well as a separate normalized score.

The detect method takes optional params:

  • isPlainText (default is False): set to True if you know your bytes don’t have any XML/HTML markup
  • includeExtendedLanguages (default is True): set to False to exclude “extended” languages added by Google
  • hintTopLevelDomain (default is None): set to the last part of the domain name that the content came from (for example if the URL was, pass the string ‘cz’). This gives a hint that can bias the detector somewhat.
  • hintLanguageCode (default is None): set to the possible language. For example, if the web-server declared the language, or the content itself embedded an http-equiv meta tag declaring the language, pass this (for example, “it” for Italian). This gives a hint that can bias the detector somewhat.
  • hintEncoding (default is None): set to the original encoding of the content (note you still must pass UTF-8 encoded bytes). This gives a hint that can bias the detector somewhat. NOTE: this is currently not working.
  • pickSummaryLanguage (default is False): if False, CLD will always return the top matching language as the answer. If True, it will sometimes pick 2nd or 3rd match (for example, if English and X match, where X (not UNK) is big enough, assume the English is boilerplate and return X). In simple testing accuracy seems to suffer a bit (XX to YY %) when this is True so I’ve defaulted to False.
  • removeWeakMatches (default is True): if a match isn’t strong enough, delete it. This ensures some amount of confidence when a language is returned.

The module exports these global constants:

  • cld2.ENCODINGS: list of the encoding names CLD recognizes (if you provide hintEncoding, it must be one of these names).
  • cld2.LANGUAGES: list of languages and their codes (if you provide hintLanguageCode, it must be one of the codes from these codes).
  • cld2.EXTERNAL_LANGUAGES: list of external languages and their codes. Note that external languages cannot be hinted, but may be matched if you pass includeExtendedLanguages=True (the default).
  • cld2.DETECTED_LANGUAGES: list of all detectable languages, as best I can determine (this was reverse engineered from a unit test, ie it contains a language X if that language was tested and passes for at least one example text).
Release History

Release History

This version
History Node


History Node


History Node


History Node


Download Files

Download Files

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

File Name & Checksum SHA256 Checksum Help Version File Type Upload Date
pycld2-0.31.tar.gz (14.3 MB) Copy SHA256 Checksum SHA256 Source Mar 15, 2015

Supported By

WebFaction WebFaction Technical Writing Elastic Elastic Search Pingdom Pingdom Monitoring Dyn Dyn DNS Sentry Sentry Error Logging CloudAMQP CloudAMQP RabbitMQ Heroku Heroku PaaS Kabu Creative Kabu Creative UX & Design Fastly Fastly CDN DigiCert DigiCert EV Certificate Rackspace Rackspace Cloud Servers DreamHost DreamHost Log Hosting