, "filter": "phonetic_filter": "type": "phonetic", "encoder": "double_metaphone", "replace": false
Elasticsearch was doing exactly what we told it to do: matching text. But the text was "dirty," and our users expected "fuzzy" intelligence without sacrificing precision. elasticsearch plugin
Standard search would have failed. But the double_metaphone encoder in our plugin mapped "Indemnifikation" and "Indemnification" to the same phonetic code: INTM . But the double_metaphone encoder in our plugin mapped
An Elasticsearch plugin is a custom extension that adds functionality to the core engine. Because Elasticsearch is written in Java and uses Google Guice for dependency injection, developers can "hook" into various parts of the server's lifecycle. He then tried a trickier one
He then tried a trickier one. He typed (a misspelling of "Indemnification").
The lesson? For every plugin you add—even the innocuous analysis-phonetic —you add a dependency that must be maintained, upgraded, and tested across every future cluster upgrade.
The correct document popped up.