match_phrase查询的模糊行为

Question

Days ago I got this "problem". I was running a match_phrase query in my index. Everything was as expected, until I did the same search with a multiple words nouns (before I was using single word nouns, eg: university). I made one misspelling and the search did not work (not found), if I removed a word (let's say the one that was spelled correctly), the search work (found).

Here there are the example I made:

The settings

PUT index1
{
  "mappings": {
    "myType": {
      "properties": {
        "field1": {
          "type": "string",
          "analyzer": "standard"
        }
      }
    }
  }
}

POST index1/myType/1
{
  "field1": "Commercial Banks"
}

Case 1: Single noun search

GET index1/myType/_search
{
  "query": {
    "match": {
      "field1": {
        "type": "phrase", 
        "query": "comersial",
        "fuzziness": "AUTO"
      }
    }
  }
}

{
  "took": 16,
  "timed_out": false,
  "_shards": {
    "total": 5,
    "successful": 5,
    "failed": 0
  },
  "hits": {
    "total": 1,
    "max_score": 0.19178303,
    "hits": [
      {
        "_index": "index1",
        "_type": "myType",
        "_id": "1",
        "_score": 0.19178303,
        "_source": {
          "field1": "Commercial Banks"
        }
      }
    ]
  }
}

Case 2: Multiple noun search

GET index1/myType/_search
{
  "query": {
    "match": {
      "field1": {
        "type": "phrase", 
        "query": "comersial banks",
        "fuzziness": "AUTO"
      }
    }
  }
}

{
  "took": 1,
  "timed_out": false,
  "_shards": {
    "total": 5,
    "successful": 5,
    "failed": 0
  },
  "hits": {
    "total": 0,
    "max_score": null,
    "hits": []
  }
}

So, in the second case, why am I not finding the document when performing the match_phrase query? Is there something I am missing? Those result just make doubt about what I know. Am I using the fuzzy search incorrectly? I'm not sure if this is a problem, or I'm the one who do not understand the behavior.

Many thanks in advance for reading my question. I hope you can help me with this.

Answer 1

Fuzziness is not supported in phrase queries.

Currently, ES is silent about it, ie it allows you to specify the parameter but doesn't warn you that it is not supported. A pull request (#18322) (related to issue #7764 ) exists that will remedy to this problem. Once merged into ES 5, this query will error out.

In the breaking changes document for 5.0, we can see that this won't be supported:

The multi_match query will fail if fuzziness is used for cross_fields , phrase or phrase_prefix type. This parameter was undocumented and silently ignored before for these types of multi_match .