Thanks for all discussions. It looks like it could become top my head and time to follow the directions proposed here.
I tried a simple approach: in the indexing method of SpellChecker, I store each word (except noisy/stop word) in a relational db table ( current with the embedded Derby). When user types letters(ajax style), I just take the last part (part after last space) for Western languages (for CJK it has to be the whole part), query the table like "ac%" to get a list of words and prepend the current words before the space. Prototyping seems workable. The table can be designed to take some statistics and users can put more phrases (not just words) in. Not sure if the table can be too big -- powerful DB system and proper indexing should be OK by my reasoning. One part I don't know is how to remove some words from the table when some docs that have been removed from the repository. I don't know how SpellChecker removes the words that don't appear in any docs in repository any more. -- View this message in context: http://jackrabbit.510166.n4.nabble.com/AutoCompelete-tp3058073p3059263.html Sent from the Jackrabbit - Users mailing list archive at Nabble.com.
