CODEWITHSUNDEEP
elasticsearchlucene
shantanuo
shantanuo

Reputation: 32306

Change the search order based on word index

Is there any way to add weight to the terms found at the beginning of the document? For e.g. I have 3 documents.

Medicine XXX
Sulpher This medicine contains sulpher and should be taken only after consultation with your doctor.

Medicine YYY
contains: sulpher Not recommended by most physicians

Medicine ZZZ
This medicine works like sulpher but does not contain sulpher at all.

The document XXX should be listed at the top for the search term "Sulpher" because that is the first word in that document. It is OK if YYY is listed at the top because that is same as XXX. But ZZZ should always be the last. In other words terms found at "left" should have higher priority than the terms found on "right" side of the document.

Upvotes: 1

Views: 79

Answers (1)

Joe - Check out my books
Joe - Check out my books

Reputation: 16905

You could boost by the lowercase-normalized term position:

PUT sulphur
{
  "settings": {
    "analysis": {
      "normalizer": {
        "keyword_lowercase": {
          "type": "custom",
          "filter": ["lowercase"]
        }
      }
    }
  },
  "mappings": {
    "properties": {
      "text": {
        "type": "text",
        "fields": {
          "keyword": {
            "type": "keyword",
            "normalizer": "keyword_lowercase"
          }
        }
      }
    }
  }
}

POST sulphur/_doc
{"text":"This medicine works like sulpher but does not contain sulpher at all."}
POST sulphur/_doc
{"text":"contains: sulpher Not recommended by most physicians"}
POST sulphur/_doc
{"text":"Sulpher This medicine contains sulpher and should be taken only after consultation with your doctor."}

and then 

GET sulphur/_search
{
  "query": {
    "bool": {
      "must": [
        {
          "function_score": {
            "query": {
              "match": {
                "text": "sulpher"
              }
            },
            "script_score": {
              "script": """
                def pos = doc['text.keyword'].value.indexOf('sulpher');
                return Math.exp((2.0/(pos+1)))
              """
            },
            "boost_mode": "replace"
          }
        }
      ]
    }
  }
}

yielding

[
  {
    "_index":"sulphur",
    "_type":"_doc",
    "_id":"sf5S2nEBW-D5QnrWODvB",
    "_score":7.389056,
    "_source":{
      "text":"Sulpher This medicine contains sulpher and should be taken only after consultation with your doctor."
    }
  },
  {
    "_index":"sulphur",
    "_type":"_doc",
    "_id":"sP5S2nEBW-D5QnrWNjtw",
    "_score":1.1993961,
    "_source":{
      "text":"contains: sulpher Not recommended by most physicians"
    }
  },
  {
    "_index":"sulphur",
    "_type":"_doc",
    "_id":"r_5S2nEBW-D5QnrWNDuw",
    "_score":1.079959,
    "_source":{
      "text":"This medicine works like sulpher but does not contain sulpher at all."
    }
  }
]

Upvotes: 2

Related Questions