How can I multiply the score of two queries together in Elasticsearch?

Question

In Solr I can use the query function query to return a numerical score for a query and I can user that in the context of a bf parameter something like bf=product(query('cat'),query('dog')) to multiply two relevance scores together.

Elasticsearch has search API that is generally more flexible to work with, but I can't figure out how I would accomplish the same feat. I can use _score in a script_function of a function_query but I can only user the _score of the main query. How can I incorporate the score of another query? How can I multiply the scores together?

Compression

even better you can name those queries and do anything you want with them in the context of Solr's query DSL, such as catQuery={!edismax qf=title^10 text v=$q} then refer to that query in a function query: product($catQuery...). Disappointed Elasticsearch lacks this fairly powerful capability – Endo

Answer 1

You could script a TF*IDF scoring function using a function_score query. Something like this (ignoring Lucene's query and length normalization):

"script": "tf = _index[field][term].tf(); idf = (1 + log ( _index.numDocs() / (_index[field][term].df() + 1))); return sqrt(tf) * pow(idf,2)"

You'd take the product of those function results for 'cat' and 'dog' and add them to your original query score.

Here's the full query gist.

Chatterbox

And there is a bevy of other index-level information you can use to test out and construct your own scoring script directly in the DSL: elastic.co/guide/en/elasticsearch/reference/1.6/… – Chatterbox

pretty awesome answer... but I would hop to avoid this type of low level detail. Especially if I'm thinking about a multiterm or multifield query. The function_score's functions supports a filter section. Ideally we could also just have a query section. – Compression

Or even simpler, add a score_mode argument to the should block of a Bool Query. (As of now, ES scores boolean queries like this: must clause score + sum-of(should clause scores)). What you're trying to do is must clause score + prod-of(should clause scores). – Chatterbox

Answer 2

Alternately, if you've got something in that bf that's heavyweight enough you'd rather not run it across the entire set of matches, you could use rescore requests to modify the score of the top N ranked ORIGINAL QUERY results using subsequent scoring passes with your (cat, dog, etc...) scoring-queries.

Chatterbox

why did you answer the same question twice? If you wish to present an alternative solution you can add it into the same answer – Striper

I like having two different answers presented as two different answers so that the upvotes can differentiate between them – Compression