25_Query_scoring.asciidoc

Manipulating relevance with query structure

The Elasticsearch Query DSL is immensely flexible. You can move individual query clauses up and down the query hierarchy to make a clause more or less important. For instance, imagine the following query:

quick OR brown OR red OR fox

We could write this as a bool query with all terms at the same level:

GET /_search
{
  "query": {
    "bool": {
      "should": [
        { "term": { "text": "quick" }},
        { "term": { "text": "brown" }},
        { "term": { "text": "red"   }},
        { "term": { "text": "fox"   }}
      ]
    }
  }
}

But this query might score a document which contains quick, red and brown the same as another document which contains quick, red and fox. Red and brown are synonyms and we probably only need one of them to match. Perhaps we really want to express the query as:

quick OR (brown OR red) OR fox

According to standard Boolean logic, this is exactly the same as the original query but, as we have already seen in [bool-query], a bool query does not only concern itself with whether a document matches or not, but also with how well it matches.

A better way to write this query would be as follows:

GET /_search
{
  "query": {
    "bool": {
      "should": [
        { "term": { "text": "quick" }},
        { "term": { "text": "fox"   }},
        {
          "bool": {
            "should": [
              { "term": { "text": "brown" }},
              { "term": { "text": "red"   }}
            ]
          }
        }
      ]
    }
  }
}

Now, red and brown compete with each other at their own level and quick, fox and red OR brown are the top-level competitive terms.

We have already discussed how the match, multi_match, term, bool, and dis_max queries can be used to manipulate scoring. In the rest of this chapter, we are going to look at three other scoring-related queries: the boosting query, the constant_score query and the function_score query.