Is this the "Paris - France + Italy = Rome" thing? I think I've seen a couple of discussions about that.
When you say a "term" are you actually looking for one keyword? Because that's very different from a query that can be infinitely complex. Regards, Alex. Personal: http://www.outerthoughts.com/ and @arafalov Solr resources and newsletter: http://www.solr-start.com/ and @solrstart Solr popularizers community: https://www.linkedin.com/groups?gid=6713853 On 24 November 2014 at 13:44, Robust Links <pey...@robustlinks.com> wrote: > @ Alexandre.. the business case is the following. I am using lucene/solr to > compose vectors. I want to apply different functions (addition, > multipication, tensor products,etc) to these vectors. The end result of > that is a new vector. I would like to then go back from that vector back to > term that might have generated that vector..i.e. the following pipeline > > 1) term i -> query i -> N docs_of_i > 2) term j -> query j -> N docs_of_j > 3) F(docs_of_i, docs_of_j) -> N docs > 4) N docs -> term_k > > 1 & 2 are usual query path. steps 3 and 4 are what i am trying to do > > @walter. if I am correct, MLT will generate a term for *EACH* element of > the vector ("N docs" above).. but I am interested in a single term > > > > On Mon, Nov 24, 2014 at 1:31 PM, Walter Underwood <wun...@wunderwood.org> > wrote: > >> This feature is called “more like this”. I think it only works for a >> single document, but it probably could be extended. >> >> wunder >> Walter Underwood >> wun...@wunderwood.org >> http://observer.wunderwood.org/ >> >> On Nov 24, 2014, at 10:26 AM, Alexandre Rafalovitch <arafa...@gmail.com> >> wrote: >> >> > Very unlikely. What's the business case? >> > >> > Regards, >> > Alex. >> > Personal: http://www.outerthoughts.com/ and @arafalov >> > Solr resources and newsletter: http://www.solr-start.com/ and @solrstart >> > Solr popularizers community: https://www.linkedin.com/groups?gid=6713853 >> > >> > >> > On 24 November 2014 at 13:23, Robust Links <pey...@robustlinks.com> >> wrote: >> >> Hi >> >> >> >> Is there a way (in Lucene / SOLR) to go from a vector of documents (IDs, >> >> and optionally their scores) to a query? i.e. the reverse process of >> query >> >> -> document? >> >> >> >> thank you >> >> >> >> Peyman >> >>
