On Tue, Dec 28, 2010 at 5:26 AM, Biedermann,S.,Fa. Post Direkt <[email protected]> wrote: > Hi Robert, > > Thanks for your hint about LevensteinAutomata. Are AutomatonQueries planned > for an upcoming release?
yes, but its in trunk, so you can use it now... > > At the moment, we build the reference to boost documents those at query time > which contain fuzzily seldom used tokens within a queried region, in a manner > of speaking a fuzzied localised idf() .The boosts are injected via payloads. > Since levenstein must be calculated within a (fuzzied) region only, O(mn) > applies "only" to each region. On the outside, we have O(#region). > > The problem could be equivalently solved query time. But this would mean to > count the matched documents of each fuzzy query within a more complex queries. > In Release 3.0.2. it looks quite complicated to me to incorporate a different > scoring model that first count matches of each fuzzy sub-query and then apply > the boosts to the matched tokens. I haven't seen a Scorer doing this so far. > Furthermore we are sensible about query time. > > Do you have any ideas? Not sure I fully understand what your app needs to do, but you can take a look at using a different rewrite method. for example, it seems like rewriting to span queries (see SpanMultiTermQueryWrapper) might be close to what you want, except it suffers from the problem that boosting is completely broken in Lucene's span queries (since they don't combine with real Scorers but instead Spans)... --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
