Elastic Search: Boost Results with "Exact match" in "query_string" without specifying the field

1
votes

I am using query_string to search for records from Elastic,

Example Query : 

GET /stories_qa/_search
{
  "query": {
    "query_string": {
      "query": "Johnson && Johnson"
    }
  }
}

This query gives me relevant records, but the exact match record is not on top, I figured out on how we can boost records based on specific fields, but in my case, I don't want to provide field and make search restrictive, Is there an option in ElasticSearch to achieve the same.

2
javaelasticsearch
if you don't want to make your search restrictive/specific, it will be hard to make it precise. But maybe someone can give you some hints – Pierre Mallet
Can you add at least two sample docs, mentioning the one that should be on top? – Nishant

2 Answers

1
votes

You can probably do something like below, but do thorough testing and test multiple scenarios before you move it to production. 

POST <your_index_name>/_search
{
  "query": {
    "bool": {
      "should": [
        {
          "bool": {
            "must": [
              {
                "query_string": {
                  "query": "\"Karnataka Delhi\"",
                  "default_operator": "AND"
                }
              }
            ]
          }
        },
        {
          "query_string": {
            "query": "Karnataka Delhi",
            "default_operator": "OR",
            "boost": 2
          }
        },
        {
          "query_string": {
            "query": "Karnataka Delhi",
            "default_operator": "AND",
            "boost": 4
          }
        }
      ]
    }
  }
}

Hope this helps!

0
votes

It would help with examples, but you can use phrase_match query by adding double quote. If you add a boost to the phrase_match query you will be able to get to exact match ranked higher.

The following query should do what you want.

GET /stories_qa/_search
{
  "query": {
    "query_string": {
      "query": "(\"Johnson && Johnson\")^2 OR (Johnson && Johnson)"
    }
  }
}