[Recipes] Slop, Fuzziness and Proximity Searches in Elasticsearch

Missing Terms (Without Slop)

Assume we need to search for address "246 Beverly Road", but forgot "Beverly":

GET /accounts/_search
{
  "query": {
    "match_phrase": {
      "address": "246 Road"
    }
  }
}

Note: This will not return any result.

Missing Terms (With Slop)

Assume we need to search for address "246 Beverly Road", but forgot "Beverly":

GET /accounts/_search
{
  "query": {
    "match_phrase": {
      "address": {
        "query": "246 Road",
        "slop" : 1
      }
    }
  }
}

Note:

1. This will return the record with address "246 Beverly Road" as we are telling we are ok to move one word for getting a match. 

2. You may specify a very high slop value (e.g. 100) for proximity searches ( return documents with both words, but words which are closer have higher relevance).

Spelling mistake (without fuzziness)

Assume we are searching for "Hamilton" and we typed "Hamiltn" instead with a missing o:

GET /accounts/_search
{
  "query": {
    "match": {
      "city":"Hamiltn"
    }
  }
} 

Note: This will not return any value.

Spelling mistake (with fuzziness)

Assume we are searching for "Hamilton" and we typed "Hamiltn" instead with a missing o:

GET /accounts/_search
{
  "query": {
    "fuzzy": {
      "city":{
        "value": "Hamiltn", "fuzziness": "2"
      }
    }
  }
}

Note:

1. This will return record with city "Hamilton".

2. We use "fuzziness" as 2, because "Hamilton" is stored as "hamilton" in the inverted index. One change for capitalizing H and one for missing o.

3. You may also use "fuzziness": "AUTO" which means:

   • 0 corrections for 1-2 character strings.

   • 1 correction for 3-5 character strings.

   • 2 corrections for strings with length more than 5.

4. The fuzziness within Elasticsearch follows the levelshtein edit distance. 

TODO

1. Add example to demo proximity search relevance.