Dear Upayvira, Thank you very much. I think probably in long-term the fluctuation of changing score becomes really small. Probably I can reduce that with considering percentage instead of filtering by raw score. (however it is still possible that percentage changes during a period of time) Anyway, I have another question about MLT which is not related to score calculation, "what is the difference of calculating MLT on a field that is set to has 'termVectore=true' with the one that is not?" Best regards.
On Wed, Feb 4, 2015 at 12:05 AM, Upayavira <u...@odoko.co.uk> wrote: > I've seen this done (encouraged against it, but didn't win). It works. > Except, sometimes things change in the index, and the scores change > subtly. We get complaints that documents that previously were above the > threshold now aren't, and visa-versa. I try to explain that the score > has no meaning between two search requests, but unfortunately, there's > *enough* similarity between requests to make it work, *sometimes*. But > when it doesn't work, people get baffled, and don't accept the truth as > an answer (you can't use scores to compare separate sets of search > results). > > Upayavira > > On Tue, Feb 3, 2015, at 08:01 PM, Ali Nazemian wrote: > > Dear Markus, > > Hi, > > Thank you very much for your response. I did check the reason why it is > > not > > recommended to filter by score in search query. But I think it is > > reasonable to filter by score in case of finding similar documents. I > > know > > in both of them (simple search query and mlt query) vsm of tf-idf > > similarity is used to calculate the score of documents, but suppose you > > indexed news as document in solr and you want to find all enough similar > > news for the specific one. In this case I think it is reasonable to > > filter > > similar documents by score threshold. Please correct me if I am wrong. > > Thank you very much. > > Regards. > > On Feb 3, 2015 7:00 PM, "Markus Jelsma" <markus.jel...@openindex.io> > > wrote: > > > > > Hi - sure you can, using the frange parser as a filter: > > > > > > > > > > https://cwiki.apache.org/confluence/display/solr/Other+Parsers#OtherParsers-FunctionRangeQueryParser > > > > > > > http://lucene.apache.org/solr/4_10_3/solr-core/org/apache/solr/search/FunctionRangeQParserPlugin.html > > > > > > But this is very much not recommended, at all, so don't do it: > > > > > > https://wiki.apache.org/lucene-java/LuceneFAQ#Can_I_filter_by_score.3F > > > > > > -----Original message----- > > > > From:Ali Nazemian <alinazem...@gmail.com> > > > > Sent: Tuesday 3rd February 2015 16:22 > > > > To: solr-user@lucene.apache.org > > > > Subject: MoreLikeThis filter by score threshold > > > > > > > > Hi, > > > > I was wondering how can I limit the result of MoreLikeThis query by > the > > > > score value instead of filtering them by document count? > > > > Thank you very much. > > > > > > > > -- > > > > A.Nazemian > > > > > > > > -- A.Nazemian
