MoreLikeThis similarity ranking


(Itamar Syn-Hershko) #1

Hi all,

We are using MoreLikeThis and trying to measure how similar each document
really is from Elasticsearch's point of view.

Our current way of doing that is indexing the actual content, and by
looking at its score in the results (we don't filter it out) we can
establish some sort of a pivot we can work based on. That is, all results
with a higher score than the original are considered perfect matches, and
results with a lower score are handled based on their distance from that
pivot score.

As Lucene scores are extremely subjective I'm pretty sure this is the only
way to rank MLT results reliably, but I was wondering whether any of you
know of a better way to do this?

Thanks,

Itamar.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


(system) #2