On Saturday 08 March 2003 04:31, Rishabh Bajpai wrote: > are we sure of this?? i was under the impression that Lucene does > "first-found-first-returned", and as a result I ended up writing a sorting > method on the results? so can i actually do away with that? > secondly, is it an ascending order by relevence or descending order by > relevence - as in the term with the highest score comes at the top or at > the bottom?
I have never done sorting, and results have come in descerding order of relevance... and I doubt that could happen by just pure coincidence. :-) I haven't looked at the sources, but I believe at least the 'high-level' access methods do return results in best-match-first order. > finally, my queryString looks like: > ( title:rishabh^4 OR title:bajpai^4 ) OR ( description:rishabh^3 OR > description:bajpai^3 ) OR ( authcontrib:rishabh^2 OR authcontrib:bajpai^2 ) > OR ( keywords:rishabh^1 OR keywords:bajpai^1 ) > > so, will the appropriate score be returned and terms sorted according to > that? If I have understood ranking code properly (from FAQ and previous messages on this list; check archives for those), there's no guarantee that term boosted by '4' would get 40% weight of resulting score, nor that '1' would get 10%. Rather, 4 would mean it gets (relatively speaking) 4 times as big boost before normalization as it would by default. Default weighing depends on frequence of the key term in indexed content as far I have understood. It should be possible to override this if you just want equal weighings independent of search term (before boosting; normalization does not alter relative weighing), although I haven't gone through ranking code. -+ Tatu +- --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
