Minimum_should_match odd behaviour when mixed with stopword fields and cross_field searches

Elastic Stack Elasticsearch
1

Using ES 5.1.1 (Edit: Just upgraded to 5.5.0, same behaviour)

I'm getting some odd behaviour when mixing minimum_should_match with differing analyzers, and was wondering if anyone had any insights.

Simplified down, my issue is this:

I have two analysers. One has stopwords (A), and one doesn't(B).

There are two fields.
"title", which uses analyser A (the one with stopwords)

"keywords", which uses analyser B.

The stopwords list includes the word "cat".

There is one record:
{"title": "Hello world", "keywords": "dog"}

I'm doing the following query:

"query" : {
	"multi_match": {
		"query" :  "hello cat",
		"type" : "cross_fields",
		"fields": ["title", "keywords"],
		"minimum_should_match" : 2
	}
}

Now, in theory there should be no results. The word "cat" doesn't appear anywhere in the data, and minimum_should_match is set to 2. Yet this result is returned.

If I change the query term to:
"hello cheese", I rightly get 0 results.

Does anyone know why is this? Any suggestions on how can I work around it to get the desired result would be appreciated?
Thanks

2

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.