2011/6/18 Dmitry Kan <dmitry....@gmail.com> > Do you mean you would like to boost the facets that contain the most of the > lemmas? >
That would be good, but I'd prefer getting facets, for example, from first 50 of 500 docs only . > What is the user query in this case and if possible, what is the use case > (may be some other solution exists for what you are trying to achieve)? the use case is to help the user refining a query with the "most relevant" facets, which in theory come from the most relevant documents. So with 500 results being sorted by score (desc) the facet counts would come also from the documents ranked 490 to 500, which contain less relevant information. 2011/6/18 lee carroll <lee.a.carr...@googlemail.com> > Hi Tommaso > > I don't think you can achieve what you want using vanilla solr. > Facet counts will be for the result set matching not for the top n > result sets matching. > > However what is your use case ? Assuming its for faceted navigation > showing facets for the > top n result sets could be confusing to your users. As the next > incremental filter applied by the user would change the "relevancy > focus" of the user and produce another set of top n facet counts with > a document set un-related to the last result set. This could be a very > bad user experience producing a fluctuating facet counts (ie a filter > narrowing the search could produce an increase in a facet term count - > very odd) also the result set could change strangely with docs > floating in and out of the result list. > Right :-) Thanks for pointing this out. > > relevancy seems to be the answer here - if your docs are scored > correctly then counting all docs in the result set for the facet > counts is correct. do you need to improve relevancy? I have a quite good relevance obtained after playing a bit with dismax and bq. I think the problem is just in how the facets are being used, I think a customized SpellChecker sounds like the right component to provide smart suggestions. 2011/6/20 Toke Eskildsen <t...@statsbiblioteket.dk> > On Thu, 2011-06-16 at 12:39 +0200, Tommaso Teofili wrote: > > Do you know if it is possible to show the facets for a particular field > > related only to the first N docs of the total number of results? > > It collides with the inner working in Solr, as faceting does not process > the doc-IDs from the matching documents in result order. It also uses > all the hits, but that could be hacked. > > What is N? If it is a fairly low number (hundreds) and your documents > are indexed with an unique ID, you can extract the IDs and perform a > facet-request with the ORed IDs as query. > > > I am a bit curious about what you're trying to achieve here. > Conventionally, faceting provides an overview of all data, often > prioritized by occurrence count. While I understand the idea of trying > to use weights to prioritize, limiting the faceting to a subset of the > result set seems very much like a standard ranked document search. > > my use case (that is my customer's) sounds like a mixed one; as I said I suspect that an interesting try would be mixing the spellcheck's result with facets using spellcheck's suggestions as facet queries. Thanks all for your responses as they were very useful to understand how to face my use case. Regards, Tommaso
