Skip to main content

Improving Search Results

Proximity Searches​

GET /proximity/_search
{
    "query": {
        "match_phrase": {
            "title": {
                "query": "spicy sauce",
                "slop": 1
            }
        }
    }
}
note

Suppose that we want to allow a term in between the spicy and sauce terms, we can accomplish this by adding a parameter named slop for the match phrase query.

The value for this slop parameter should be an integer representing how far apart terms are allowed to be while still being considered a match.

How far apart refers to how many times a term may be moved for a document to match.

This example above satisfies a slop value of one which allows a single move to be made to make a document match.

e.g. It allows "spicy tomato sauce" to match

Affecting Relevance scoring with proximity​

  • The closer the proximity, the higher the relevance scores.
    • In other words, the lower the edit distance, the higher the relevance scores.
  • Proximity is one of the factors of relevance scores but is not the only one that makes a big difference.
GET /proximity/_search
{
    "query": {
        "bool": {
            "must": [
                "match": {
                    "title": {
                        "query": "spicy sauce"
                    }
                }
            ],
            "should": [
                {
                    "match_phrase": {
                        "title": {
                            "query": "spicy sauce",
                            "slop": 5
                        }
                    }
                }
            ]
        }
    }
}

Fuzzy Match Query (Handling Typos)​

  • This fuzziness parameters account for when users make typos or mispellings
GET /product/_search
{
    "query": {
        "match": {
            "name": {
                "query": "l0bster",
                "fuzziness": "auto"
            }
        }
    }
}
  • Without the fuzziness parameter, "l0bster" will not show "lobster" search results.
  • Maximum fuzziness value is 2.
    • Most misspellngs are just 1 or 2 characters.
GET /product/_search
{
    "query": {
        "match": {
            "name": {
                "query": "l0bster love",
                "operator": "and",
                "fuzziness": 1
            }
        }
    }
}
  • fuzziness of 1 can also be applied for each word in the query.
GET /product/_search
{
    "query": {
        "match": {
            "name": {
                "query": "lvie",
                "fuzziness": 1,
                "fuzzy_transpositions": false
            }
        }
    }
}
  • Transpositions example: AB -- BA
  • fuzzy_transpositions are enabled true by default.
    • They count as 1 edit instead of 2.
    • e.g. live -- lvie
    • If fuzzy_transpositions is false, need to set fuzziness to 2 to find search results with "live" value.

Fuzzy Query​

GET /product/_search
{
    "query": {
        "fuzzy": {
            "name": {
                "value": "LOBSTER",
                "fuzziness": "auto"
            }
        }
    }
}
  • Fuzzy Query is a term-level query.
    • Thus, fuzzy query is not analyzed.
tip

Therefore, prefer to use the match query with a fuzziness parameter.

Adding Synonyms​

PUT /synonyms
{
    "settings": {
        "analysis": {
            "filter": {
                "synonym_test": {
                    "type": "synonym",
                    "synonyms": [
                        "awful" => "terrible",
                        "awesome" => "great, super",
                        "weird, strange"
                    ]
                }
            },

            ...
        }
    }
}
note

All "awfuls" being search will show up terrible, too.

"weird, strange" are placed at the same position, so no replacement occurs.

Adding synonyms from file​

  • For words with so many synonyms, store the synonyms on a text file to reference
...

"synonym_test": {
    "type": "synonym",
    "synonyms_path": "analysis/synonyms.txt"
}

...
  • The synonym files should be available on all notes storing documents for the index using the analyzer.
warning

Elasticsearch picks up the new synonym when restarting a node but it does not re-index documents when starting back up.

Highlighting matches in fields​

GET /highlighting/_search
{
    "_source": false,
    "query": {
        "match": { "description": "Elasticsearch story" }
    },
    "highlight": {
        "fields": {
            "description": {}
        }
    }
}
  • Next example, wrap matches within a different tag.
    • Use pre-tags and post-tags parameters.
GET /highlighting/_search
{
    "_source": false,
    "query": {
        "match": { "description": "Elasticsearch story" }
    },
    "highlight": {
        "pre-tags": ["<strong>"],
        "post-tags": ["</strong>"],
        "fields": {
            "description": {}
        }
    }
}

Stemming​

  • Use stemming to improve the matches of search queries.
  • A stemmer is applied to a field.
  • Elasticsearch will highlight the original words even after stemming has been applied to a text field.
PUT /stemming_test
{
    ...

    "stemmer_test": {
        "type": "stemmer",
        "name" : "english"
    }
}