How can I allow for fuzziness in a multimatch query for elasticsearch?

Elastic Stack Elasticsearch
1

Consider an index containing thousands of flowers and their descriptions in my ES. For example, a redacted json document would look like this:

{
  "name": "rose",
  "description": "some long description about roses",
  "categories": [
    "flowers",
    "red",
    "thorn",
    "perfume"
  ]
}

For some other needs, I have the following mapping to convert all the categories into a text field like this:

{
	"mappings": {
		"properties": {
			"categories": {
				"type": "keyword",
				"copy_to": "m_cats"
			},
			"m_cats": { "type": "text" }
		}
	}
}

The issue that I am facing is that when I run a multimatchquery, i can get a hit if i search for red flowers, but if i search for red flower, i cannot get a match because the s in flowers is missing.

I have tried to solve this by using various fuzziness values including 0 and AUTO, but I can never get the rose document to show up for red flower, only shows up for red flowers. I am aware that the fuzziness applies to either keyword or text fields, and my index for flowers is showing m_cats as text correctly.

My code is using the olivere elasticsearch for golang, and is structured in the following way currently:

fields := []string{"m_cats", "description"}
q := elastic.NewMultiMatchQuery(query, fields...).Fuzziness("AUTO:1,6").MinimumShouldMatch("1")
result, err := helpers.Client.Search().Index(index).Size(limit).Query(q).Do(helpers.Ctx)

The JSON query roughly breaks down to

{
  "query": {
    "multi_match": {
      "query": "red flower",
      "fields": ["m_cats", "description"],
      "fuzziness": "AUTO:1,6",
      "minimum_should_match": 1
    }
  }
}

What else can I do/try so that I can get the rose document as a hit when I search for red flower? Notice the missing s.

Please note, I am not having issues with keyword searches, but because I want to provide additional results whenever possible, I am using minimum_should_match and avoiding using operator.

2

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