Index-time synonym expansion maximizes recall (not missing any documents, especially partial matches), but minimizes precision and relevancy - you are unable to select or boost exact matches. Ditto for ngrams.
As Erik indicates, using edismax with separate fields for precision (exact matches) and recall (even the most remote partial match to avoid missing any documents) with a much higher boost for exact matches. -- Jack Krupansky On Tue, Dec 1, 2015 at 10:10 AM, Erik Hatcher <erik.hatc...@gmail.com> wrote: > One technique that works well is to use copyField to end up with two > indexed fields, one with synonyms, one without. Then you can qf=title^5 > title_with_synonyms^1 with edismax and weight the “exacter” field higher > than one with synonyms. > > — > Erik Hatcher, Senior Solutions Architect > http://www.lucidworks.com <http://www.lucidworks.com/> > > > > > On Nov 30, 2015, at 8:36 PM, Brian Narsi <bnars...@gmail.com> wrote: > > > > I am using edismax with mm=1 and qs=6 > > > > I have a field type with synonyms attached to it. > > A sample synonym is: > > > > toothbrush tbrush > > > > For the following data: > > > > 1) Phillips toothbrush > > > > 2) Oral-B tbrush > > > > 3) Phillips Sonicare toothbrush > > > > > > If a user searches for > > > > q = tbrush > > > > I am getting > > > > 1), 3), 2) > > > > i.e. the one with exact synonym match is last > > > > If a user searches for > > > > q = toothbrush > > > > Result is > > > > 2), 1), 3) > > > > i.e. one with exact match is not first > > > > > > So > > > > 1) I need the exact matches to be first and then the other matches > (because > > of being a synonym) > > > > 2) If a user searches for Phillips Sonicare toothbrush; I need the > results > > in the following order 3), 1), 2) i.e. most relevant results should be > > higher > > > > What changes do I need to make in order to accomplish this? > > > > Thanks in advance, > >
