How to control scoring or ordering of results while using ngram in Elasticsearch?

Question

I am using Elasticsearch 6.X. .

I have created an index test_index with index type doc as follow:

PUT /test_index
{
  "settings": {
    "number_of_shards": 1,
    "number_of_replicas": 0,
    "analysis": {
      "analyzer": {
        "my_analyzer": {
          "type": "custom",
          "tokenizer": "my_ngram_tokenizer"
        }
      },
      "tokenizer": {
        "my_ngram_tokenizer": {
          "type": "nGram",
          "min_gram": "1",
          "max_gram": "7",
          "token_chars": [
            "letter",
            "digit",
            "punctuation"
          ]
        }
      }
    }
  },
  "mappings": {
    "doc": {
      "properties": {
        "my_text": {
          "type": "text",
          "fielddata": true,
          "fields": {
            "ngram": {
              "type": "text",
              "fielddata": true,
              "analyzer": "my_analyzer"
            }
          }
        }
      }
    }
  }
}

I have indexed data as follow:

PUT /text_index/doc/1
{
    "my_text": "ohio"
}
PUT /text_index/doc/2
{
    "my_text": "ohlin"
}
PUT /text_index/doc/3
{
    "my_text": "john"
}

Then I used search query:

{
  "query": {
    "bool": {
      "must": [
        {
          "multi_match": {
            "query": "oh",
            "fields": [
              "my_text^5",
              "my_text.ngram"
            ]
          }
        }
      ]
    }
  }
}

And got the response:

{
  "took": 1,
  "timed_out": false,
  "_shards": {
    "total": 1,
    "successful": 1,
    "failed": 0
  },
  "hits": {
    "total": 5,
    "max_score": 1.0042334,
    "hits": [
      {
        "_index": "test_index",
        "_type": "doc",
        "_id": "1",
        "_score": 1.0042334,
        "_source": {
          "my_text": "ohio"
        }
      },
      {
        "_index": "test_index",
        "_type": "doc",
        "_id": "3",
        "_score": 0.97201055,
        "_source": {
          "my_text": "john"
        }
      },
      {
        "_index": "test_index",
        "_type": "doc",
        "_id": "2",
        "_score": 0.80404717,
        "_source": {
          "my_text": "ohlin"
        }
      }
    ]
  }
}

Here, we can see the when I searched for oh , I got results in the order:

-> ohio
-> john
-> ohlin

But, I want to have scoring and order of the results in a way which gives higher priority to matching prefix:

-> ohio
-> ohlin
-> john

How can I achieve such result ? What approaches can I take here ? Thanks in advance.

Answer 1

You should add a new subfield with a new analyzer using the edge_ngram tokenizer then add the new subfield in your multimatch.

You need then to use the type most_fields for your multimatch query. Then only the documents starting by the search term will match on this subfield and then will be boosted against others matching documents.