Elasticsearch clickthrough relevance ranking


I have been reading about lucidworks search. One interesting feature I
found on it was what they call "click tracking and boosting". I haven't
find anything like it in elastic search world (hopefully because of lack of
google querying skills). The feature works like this: the user makes a
query and clicks on a result. That click is tracked and it boosts the score
of that document for the query.

Reading more about lucid's implementation, it seems like they create a
"clicks" index along with the main index. The click index has a clicks_term
field which holds queries for that particular document. Each query has a
weight proportional to the number of times the document was clicked on that
given query. They make it like that so they can reindex the documents
faster than a full reindex of all the fields for each document. Both
indices are read using ParallelReader in order to merge the fields (that
was pre-4.0, now ParallelReader seems to have two flavours: Atomic and
Composite. I don't know the difference). I'm not completely sure on any of
this, but that's what I have understood so far.

Anyway, is there any work on doing that sort of thing with elastic search?
How easy would it be to use ParallelReader on ES? Would it make sense?


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.

I am wondering about the same question now. I am curious to learn how you solved your problem.