All products

Improve Elasticsearch search relevance with AI

Tune full-text catalogue results using confirmed mappings, representative queries and measured before-and-after rankings. Use scoped query rules for explicit exceptions.

Improve Elasticsearch search relevance with AI when the expected product consistently appears below weaker matches. The elasticsearch-search-relevance skill inspects the current query and field mappings before choosing a lever. A general ranking problem may call for matching across fields or adjusting their weights. A specific promotion may instead need a narrowly scoped rule.

Use a small set of representative searches to check that improving one result does not undo another. Keep the original query and the top hits from both versions. The comparison concerns relevance rather than simply reordering everything by price, popularity or date.

What it does

  • Check whether the query searches the intended text fields.
  • Compare title and description boosts.
  • Test a rule that pins a particular document for a known query.
  • Review several top hits before and after a change.
  • Record the scope and reversal route for a ruleset change.

Install and use

  1. Install the skill and configure the official elastic CLI 0.2 or later for the authorised index. Query rules require Elasticsearch 8.10 or later and appropriate write access. Use representative, non-sensitive queries.
  2. Ask, “Compare title weighting on these catalogue searches. Show the original top hits and the proposed ranking before touching production rules.”
  3. Test the examples in a suitable environment and review permissions and rollback before saving any persistent change.

FAQ

Source

Made by Elastic. Original skill: elasticsearch-search-relevance, published on GitHub under the Apache-2.0 licence.

View on GitHub