These are an important part of creating a complete language specific view of the operating system. Even English speakers prefer to see correct spell checkers for their locale (UK vs US vs South African). Certain languages by their nature do not need or cannot use the traditional wordlist type spellcheckers found on Linux.
Hunspell is the main spell checker on Linux .
There is a legacy of previous spell checkers on *nix platforms and these are:
If starting from scratch your best bet is to focus on a hunspell checker and make use of its language specific features if needed. If you maintain aspell and ispell checker you might want to migrate them. If you maintain a current myspell checker you should probably wait until hunspell infrastructure is widely deployed.
For spell checker development on OSX, these links might be useful. Just check that things are still current
A corpus is a body of text used by language researchers and spell checker builder. You can find missing or new words for your spell checker by scanning the web. There are free tools that you can use to build your own web-based corpus. We developed CorpusCatcher for this purpose.
Other possibilities: corpusbuilder and text_cat (FIXME How to use these). The former searches the web using a public search engine and the later uses a statistical model to determine if the text found is indeed in your target language.
Once you have a list of potential words you can use the new-words script in src/wordlist in the Translate Toolkit SVN to identify words that are not in your language. Review these words and add them to you master wordlist.
This Python code could easily be used to develop language detection for a webcrawler: http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/326576
The translate project has a simple python script that creates letter frequencies that can be used in the MySpell affix files TRY line. See translate/src/wordlist/letter-frequency.py in the Translate Toolkit CVS
The easiest way to build your spellcheckers is to use our project spellchecker build framework. This will build MySpell and Aspell (Ispell temporarily disabled) spellcherckers from a common wordlist or wordlists. Look at the Afrikaans and Zulu dictionaries for a template of the process. Again this is in SVN in the dict module of the zaf project.
Checkout the dict/ module from Subversion:
svn co https://zaf.svn.sourceforge.net/svnroot/zaf/trunk/dict dict
Simple make instructions:
make– makes all spell checkers
make aspell– makes the respective dictionary
make count– gives spell checker word counts
make clean– removes all autogenerated files
make wordlist– create and packs the wordlists
Make sure that your language is included in: http://cvs.gnome.org/viewcvs/gnome-spell/gnome-spell/dictionary.c
So that Gnome applications such as Evolution can make use of your aspell spellchecker.
To get the spellchecker onto the OpenOffice.org pages and thus downloadable from within OpenOffice.org. You will need to submit a bug report. Here is and example issue: http://www.openoffice.org/issues/show_bug.cgi?id=23201
Mozilla dictionaries must be tri-licensed (GPL/LGPL/MPL) for inclusion in the source tree, which results in inclusion in a language build. For many spell checkers this will probably be a problem.
Alternatively you can create a dictionary extension and upload it to Mozilla Addons. Users who upgrade Firefox are directed to the dictionary download page ensuring rapid adoption of your spell checker.
|||Many distributions consolidated spell checking around Hunspell to some extent, for example Fedora, Firefox, Thunderbird and OpenOffice.org use Hunspell.|