After spending more time on this, it seems more likely a problem from FunctionQuery. If using boost = log(100) takes 100ms, log(log(100)) adds another 100ms, log(log(log(100))) adds another 100ms, and so on. The time goes up almost linearly instead of being constant. Any ideas?
Thanks, Guangwei On Sat, May 9, 2009 at 12:31 PM, Guangwei Yuan <guy...@gmail.com> wrote: > Hi, > > I'm trying the BoostQParserPlugin and FunctionQuery to enable query time > boosting. It works better than bf (boost function) because it multiplies > the relevancy score by the boosts. However I noticed significant > performance issues with it. The more functions I use as boosts, the slower > it gets. For example, if a query without boosts takes 300ms, adding a boost > function like log(popularity) becomes 600ms, using pow(log(popularity),2) > takes 800ms, using > product(pow(log(score_total),2),recip(rord(days),1,90,90)) takes 1100ms. > Ideally I'd like to pass in a reasonable amount of functions to adjust the > search ranking dynamically without sacrificing much performance. > > Thanks, > > Guangwei >
