Have you tried the Lucene spellchecker first? I think it could be adapted to do want, esp with the help of LUCENE-400 to produce N-word shingles (which you can then index with the Spellchecker). I'm quite sure this could be done, in fact, and would be a nice addition to Spellchecker in general.
Otis -- Sematext -- http://sematext.com/ -- Lucene - Solr - Nutch ----- Original Message ---- From: solr_user <[EMAIL PROTECTED]> To: java-user@lucene.apache.org Sent: Tuesday, January 15, 2008 1:14:06 PM Subject: Re: spell checking for combined words I don't have a list of common "combined word" queries. Splitting of words seem to be quite a standard thing, most search engines and spell checkers have this ability. It would be nice if Lucene provides this out of the box. karl wettin-3 wrote: > > > 14 jan 2008 kl. 19.47 skrev solr_user: > >> Does Lucene spell checker have the ability to suggest splitting of >> combined >> words. So for e.g. if I have got the word "apple" and "computer" in >> my >> index and if I type "applecomputer" then how can I make it suggest >> "apple computer" > > > It would probably be very expensive to analyze each query token and > decompose it using some word list or index. > > If "apple computer" is a phrase that is common in your query, then you > can create a new field with chained tokens: "think diffrent apple > computer" -> "thinkdiffrent diffrentapple applecomputer". This would > at least allow you to search for such typos. You might want to set > some threadholds like min/max token size and such in your TokenFilter. > > > -- > karl > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > > -- View this message in context: http://www.nabble.com/spell-checking-for-combined-words-tp14809197p14843700.html Sent from the Lucene - Java Users mailing list archive at Nabble.com. --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
