CODEWITHSUNDEEP
elasticsearch
user606521
user606521

Reputation: 15474

Why elastic does not find my search text?

I have multiple instances of one type, each has field displayName. This field is:

"Contributor1"
"Contributor2"
...
"Contributor49"

I have all mappings/analyzers/etc set to defaults.

I try to find this:

fuzzy_like_this_field: { "displayName": { like_text: "49" } }

But it does not return any matches. When I try following search texts:

"c49" -> nothing
"co49" -> nothing
"con49" -> nothing
"cont49" -> nothing
"contr49" -> nothing
"contri49" -> nothing
"contrib49" -> CORRECT MATCH

How I can improve the search? Strange that elastic does not find "49" - it is unique along all sources...

Upvotes: 0

Views: 341

Answers (1)

John Petrone
John Petrone

Reputation: 27515

Elasticsearch fuzzy searches on string fields are based on the Levenshtein Edit Distance:

String

When querying string fields, fuzziness is interpreted as a Levenshtein Edit Distance — the number of one character changes that need to be made to one string to make it the same as another string.

http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/common-options.html#fuzziness

A detailed explanation of the Levenshtein Edit Distance can be found here: http://en.wikipedia.org/wiki/Levenshtein_distance

For purposes of your example, it's the total number of character adds and removals that would be needed to change the term you are searching for to the term you find. The string "contrib49" is much closer to "Contributor49" than "49" is and it falls within the default distance or fuzziness for this field and search.

You can increase the fuzziness with the fuzziness parameter which defaults to .5 : http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/query-dsl-flt-field-query.html

Increasing it (let's say .7 or .8) will increase the overall fuzziness it will match.

Overall though are you sure you are using the right approach here? If all you are looking for is wildcard search, a fuzzy search may not be the best way to go - you might want to look at wildcards and ngram analyzers instead.

Upvotes: 1

Related Questions