Friday, August 2, 2013

A new version of the Compact Language Detector

It's been almost two years since I originally factored out the fast and accurate Compact Language Detector from the Chromium project, and the effort was clearly worthwhile: the project is popular and others have created additional bindings for languages including at least Perl, Ruby, R, JavaScript, PHP and C#/.NET.

Eric Fischer used CLD to create the colorful Twitter language map, and since then further language maps have appeared, e.g. for New York and London. What a multi-lingual world we live in!

Suddenly, just a few weeks ago, I received an out-of-the-blue email from Dick Sites, creator of CLD, with great news: he was finishing up version 2.0 of CLD and had already posted the source code on a new project.

So I've now reworked the Python bindings and ported the unit tests to Python (they pass!) to take advantage of the new features. It was much easier this time around since the CLD2 sources were already pulled out into their own project (thank you Dick and Google!).

There are a number of improvements over the previous version of CLD:
  • Improved accuracy.

  • Upgraded to Unicode 6.2 characters.

  • More languages detected: 83 languages, up from 64 previously.

  • A new "full language table" detector, available in Python as a separate cld2full module, that detects 161 languages. This increases the C library size from 1.8 MB (for 83 languages) to 5.5 MB (for 161 languages). Details are here.

  • An option to identify which parts (byte ranges) of the text contain which language, in case the application needs to do further language-specific processing. From Python, pass the optional returnVectors=True argument to get the byte ranges, but note that this requires additional non-trivial CPU cost. This wiki page shows very interesting statistics on how frequently different languages appear in one page, across top web sites, showing the importance of handling multiple languages in a single text input.

  • A new hintLanguageHTTPHeaders parameter, which you can pass from the Content-Language HTTP header. Also, CLD2 will spot any lang=X attribute inside the <html> tag itself (if you pass it HTML).
In the new Python bindings, I've exposed CLD2's debug* flags, to add verbosity to CLD2's detection process. This document describes how to interpret the resulting output.

The detect function returns up to 3 top detected languages. Each detected language includes the percent of the text that was detected as the language, and a confidence score. The function no longer returns a single "picked" summary language, and the pickSummaryLanguage option has been removed: this option was apparently present for internal backwards compatibility reasons and did not improve accuracy.

Remember that the provided input must be valid UTF-8 bytes, otherwise all sorts of things could go wrong (wrong results, segmentation fault).

To see the list of detected languages, just run this python -c "import cld2; print cld2.DETECTED_LANGUAGES", or python -c "import cld2full; print cld2full.DETECTED_LANGUAGES" to see the full set of languages.

The README gives details on how to build and install CLD2.

Once again, thank you Google, and thank you Dick Sites for making this very useful library available to the world as open-source.

11 comments:

  1. Hi!
    Today I spent some time to compare the quality of few python libs for detecting languages from tracks name (really short text).
    Unfortunately, according to my tests, CLD2 gave me really weird results. :(
    Some examples:

    "Meadowlake Street" is PL

    cld2.detect("13 Años")
    (False, 10, (('Unknown', 'un', 0, 0.0), ('Unknown', 'un', 0, 0.0), ('Unknown', 'un', 0, 0.0)))

    CLD2 was not able to detect language for "I love music" too.
    Other libs are way slower than CLD2, but they gave me better results.
    I'm wondering if something was bad with my installation, or someone else gets the same bad matching for the strings I posted here.

    I run my tests against millions of strings, not just few.

    ReplyDelete
    Replies
    1. Indeed I see the same results as you; I think CLD2 is just not designed for short text.

      You could try opening an issue at https://code.google.com/p/cld2/ and see what they say?

      Delete
    2. so what is the best library for small text lang recognition?

      Delete
    3. Maybe try https://github.com/saffsd/langid.py ? But in general, working well on short text is challenging for any language detector.

      Delete
  2. Hi ,
    How to retrieve the matched tokens against the Lucene Query fired from index file ...This is am doing for auto complete ...suggest me is there any alternative way for auto complete in lucene ...

    ReplyDelete
    Replies
    1. Hi, I don't fully understand the question. Can you re-ask this, with more details, on java-user@lucene.apache.org?

      Delete
  3. HI,
    Thank you for this tutorial, it works perfectly !
    question, how to interface with PHP ?
    I searched the documentation, but I can not find anything with CLD2 …
    when i test with python the file test.py, it's OK, but with php:
    Fatal error: Class 'CLD\Detector' not found
    an idea ?

    thx for your help.

    Eric

    ReplyDelete
    Replies
    1. Maybe you can try to compile compile_full.sh from "internal" folder and call it from command line using PHP? That's what I did with Perl...

      Delete
    2. (call it = call the compiled file - compact_lang_det_test_full)*

      Delete
  4. You mentioned .NET bindings. Are they available somewhere publicly? I haven't been able to find any.

    ReplyDelete
    Replies
    1. Alas I can't find the .NET bindings either ... I remember seeing them at one point ...

      Delete