When a fuzzy query is executed in Elasticsearch, it searches for documents that contain terms that are similar to a specified term, based on a fuzzy matching algorithm. The fuzzy query searches the inverted index for terms that match the specified term within a certain degree of similarity and returns any documents that contain a matching term.
The degree of similarity is determined by the `fuzziness` parameter, which specifies the maximum number of changes allowed between the searched term and the matched terms. The allowed changes can include insertion, deletion, or substitution of characters in the term. The `fuzziness` parameter can be a number or a string, such as “AUTO”, which uses a heuristic algorithm to determine the most appropriate fuzziness value based on the length of the term being searched.
The fuzzy query also supports the `max_expansions` parameter, which limits the number of terms that are searched in the inverted index. This parameter can be used to improve the performance of the query and to prevent it from returning too many results.
Here’s an example of a fuzzy query in Elasticsearch:
GET /my_index/_search
{
"query": {
"fuzzy": {
"title": {
"value": "elastiksearch",
"fuzziness": "AUTO",
"max_expansions": 10
}
}
}
}
In this example, we are searching the `title` field in the `my_index` index for terms that are similar to the term “elastiksearch”, based on a fuzzy matching algorithm. The fuzzy query will return any documents that contain a term in the `title` field that matches the specified term within a certain degree of similarity, with a maximum of 10 terms searched in the inverted index.
The fuzzy query can also be used in combination with other query types, such as the bool query, to construct more complex search queries. By allowing for searching for terms that are similar to a specified term, based on a fuzzy matching algorithm, the fuzzy query provides a powerful and flexible way to search for text in Elasticsearch. However, it’s important to note that fuzzy queries can be computationally expensive and may not be suitable for large datasets or high-traffic applications.