Denis Bilenko
and 1 contributors


Lingua::Identify::CLD2 - CLD2 wrapper for Perl


This module is an XS wrapper around the CLD2 "compact language detection" library.

Optionally, you may choose to import a any or all of the functions and constants discussed below into your namespace using normal Exporter semantics. You can import all of them with the ":all" tag. You can choose to import only the functions or the (large number of) constants using ":functions" and ":constants" respectively.

The constants that correspond to the Language enum values in CLD2 have a CLD2_ prefix in Perl. For example CLD2::GERMAN in C++ becomes CLD2_GERMAN in Lingua::Identify::CLD2 in Perl. Unlike the Language enum values, the ULScript values already have a name prefix in C++, so they are exposed as is, eg. ULScript_Balinese.

The documentation of this module might be a bit spotty. If in doubt, refer to the CLD2 documentation of the respective functions and please submit patches after you do.


The main API function that, given a text and some other parameters, will attempt to detect the language(s) of the text. An example output is reproduced below. For details on its interpretation, please refer to the CLD2 manual. Patches welcome.

The first input parameter should be a string containing the text to analyse.

The optional second parameter is a hash reference that allows to specify various options for CLD2.

The following options enable corresponding CLD2 flags:

     Normally, several languages are detected solely by their Unicode script.
     Combined with appropritate lookup tables, this flag forces them instead
     to be detected via quadgrams. This can be a useful refinement when looking
     for meaningful text in these languages, instead of just character sets.
     The default tables do not support this use.

     For each detection call, write an HTML file to stderr, showing the text
     chunks and their detected languages.

     In that HTML file, force a new line for each chunk.

     In that HTML file, show every lookup entry.

     In that HTML file, suppress most of the output detail.

     Echo every input buffer to stderr.

     Give best-effort answer, instead of empty string. May be useful for
     short text if the caller prefers an approximate answer over none.

The following options are translated into CLDHints structure.

Pass in hints whenever possible; doing so improves detection accuracy. The set of passed-in hints are all information that is external to the text itself.

     This is intended to come from an HTTP header Content-Language: field.
     Example: "mi,en" boosts Maori and English

     This is from the hostname of the URL.
     Example: "id" boosts Indonesian

     Language code from any other context you might have.
     Example: "it" boosts Italian.

Finally, these options control how CLD2 is called:

     By default CLD2 skips html tags and expands html entities. Setting this option disables this behaviour.

     When enabled, the result will contain an extra field 'resultchunkvector'.


    $res = DetectLanguage("Привет", {bestEffort => 1});

    $VAR1 = {
          'language_code' => 'ru',
          'language_name' => 'RUSSIAN',
          'languages' => [
                             'language_code' => 'ru',
                             'percent' => 92,
                             'score' => 630
          'is_reliable' => 1,
          'text_bytes' => 14


Given a CLD2 language id, converts it to a human readable language name.


Given a CLD2 language id, converts it to a language code. Quoting the CLD2 documentation:

    Given the Language, return the language code, e.g. "ko"
    This is determined by the following (in order of preference):
    - ISO-639-1 two-letter language code
      (all except those mentioned below)
    - ISO-639-2 three-letter bibliographic language code
      (Tibetan, Dhivehi, Cherokee, Syriac)
    - Google-specific language code
      (ChineseT ("zh-TW"), Teragram Unknown, Unknown,
      Portuguese-Portugal, Portuguese-Brazil, Limbu)
    - Fake RTypeNone names.



Convert a language name or code back to a CLD2 id.

Quoting the CLD2 documentation:

    Name can be either full name or ISO code, or can be ISO code embedded in
    a language-script combination such as "en-Latn-GB".


Given a language code or CLD2 language id, returns which set of statistically-close languages lang is in. 0 means "none".


For both portability (CLD2 uses a bunch of ummm.. shell scripts as a build system) AND for consistency of the exposed constants, Lingua::Identify::CLD2 ships its own copy of CLD2. Newer versions of CLD2 thus require updating this module.

The encoding functionality for hints is mostly not exposed. But if needed, that should be a rather simple matter of (relatively little) programming.


At the time of this writing, CLD2 still lived on Google Code:



Steffen Mueller, <>

Denis Bilenko


The Lingua::Identify::CLD2 module (but not the CLD2 library) is

Copyright (C) 2015 by Steffen Mueller

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself, either Perl version 5.8.0 or, at your option, any later version of Perl 5 you may have available.

At the time of this writing, the CLD2 library code carries the following license and author notice:

    Copyright 2013 Google Inc. All Rights Reserved.
    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.
    Author: (Dick Sites)

1 POD Error

The following errors were encountered while parsing the POD:

Around line 178:

Non-ASCII character seen before =encoding in 'DetectLanguage("Привет",'. Assuming UTF-8