This an unsolved problem in general. The TREC folks try this, see: http://trec.nist.gov/
but in general I've found that each domain has such specific needs that "correctness" isn't an easy thing to pin down. Of course you can, with a known set of data, define the "best" response and try to tune Solr to return those, but that's a static snapshot not a general test of correctness. Best Erick On Fri, Sep 2, 2011 at 11:25 AM, Nemani, Raj <raj.nem...@turner.com> wrote: > All, > > I was wondering if anybody has any information on approaches to testing > and verification search results from Solr. Most of the time we end up > manually verifying the results from a search but the verification is not > necessarily scientific. > > The main question is what are we verifying these search results against? > As an example, how can I be sure that that the relevancy calculated at > any given time for a given document in an index is accurate? > > > > Hope the question makes sense. > > > > Any feedback is really appreciated. > > > > Thanks > > Raj > > > >
