У нед, 23. 11 2008. у 07:14 +0900, JiHui Choi пише: > > So- I'd like to make one. > I guess at first I need a wordlist and affix file for hunspell or > myspell or ispell. > > But I couldn't find any information about the first step to make those. > > Is anyone who helps me? >
Affix file specifies rules that can be applied to words in a word list to create new words. For some languages this can significantly reduce word list and provide better coverage, and for others it is useless. There is syntax for rules that delete, add or insert characters. Rules are applied to specific class of words (think about making plural for english nouns by adding 's' or removing 'y' and adding 'ies') Minimal affix file (without any rules) consists of two lines: SET UTF-8 TRY asfjlasjfl If you want to write some rules you can look here for syntax: http://tinyurl.com/hunspell-manual First is encoding used for word list, and second should list every letter from your language sorted by frequency. Word list is just that, one word in a row (optionally followed by affix file rule classes that should be applied to that rule). In first line you should write approximate number of words. You can use Wikipedia or other texts in your language available on the Internet to create basic corpora. Once you have kr.aff and kr.dic you need to package it as extension. There is Wiki page that describe what additional files you should make: http://wiki.services.openoffice.org/wiki/Extension_Dictionaries You will need dictionaries.xcu with one DICT_SPELL entry, and description.xml to name your extension. The best way to do packaging is to download some dictionary extension from http://extensions.services.openoffice.org/dictionary rename it to ".zip", extract it and then edit it's content. I doubt that there is ispell/aspell Korean dictionary as they don't support full Unicode. I will be glad to help you with packaging, but word list is something you have to make. Regards, Goran Rakic OpenOffice.org Serbian native-lang project lead --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
