[libvoikko] Libvoikko 3.8rc1

Harri Pitkänen hatapitk at iki.fi
Sun Sep 6 14:51:26 EEST 2015


Release candidate for new version of libvoikko is available for testing:

  http://www.puimula.org/htp/testing/libvoikko-3.8rc1.tar.gz

New in this version:

* Dictionary path specified at compile time, through VOIKKO_DICTIONARY_PATH or
  through Windows registry keys can now contain more than one directory.
  Separator character is ";" on Windows and ":" on other platforms (so this
  works now just as regular PATH environment variable).
* BCP 47 scripts subtags are now supported:
  - Four letter script subtag can now be appended when requesting a dictionary
    for specific language. This should only be used when more than one script
    is actually used for a language since libvoikko does not know what the
    "default" script for any specific language is.
  - New API function voikko_dict_script has been added.
  - Script attribute has also been added to Python and Java Dictionary objects
* HFST backend code has been re-licensed under GPL/LGPL/MPL (hfstospell itself
  is under the Apache license).
* Lots of improvements related to VFST format:
  - A new weighted variant of VFST format has been added. This variant also
    supports larger transducers with more symbols and states than the
    unweighted variant.
  - Language independent "simple" VFST spellers now use weighted transducers
    and support spelling suggestions that are sorted using the weight
    information from the transducers. Finnish VFST format (format 5) continues
    to use unweighted transducer format.
  - Finnish VFST format now reads grammar checker autocorrect data from a
    transducer instead of using built-in autocorrect data shipped with
    libvoikko. The autocorrect data within libvoikko is still included if
    Malaga backend is enabled but it is only used for grammar checking with
    Malaga dictionaries (format 2).
  - Support for WORDIDS and WORDBASES attributes has been added to Finnish
    VFST format.
  - Dictionary format 5 has reached feature parity with format 2 and is now
    good enough to be used for beta testing.
* Various small improvements have been made in Finnish grammar checker and
  morphological analysis.
* Extended support for character sets affecting at least Plains Cree written
  in Canadian syllabics, Skolt Sami and Romanian.
* Fixed a crash that appeared when the library was used within 64 bit Firefox
  on Windows.

Please test this release if you can. I will release the final version in about 
one week.

Harri


More information about the Libvoikko mailing list