Engineering Semantic Search at Elastic

Amsterdam meetup, November 2022 | Users increasingly expect text search to work like “google search” where relevant results are not just based on keyword matches, but on user intent and query meaning. For example, searching for “How to set up Elasticsearch?” should return documents such as “Elasticsearch configuration”, “Elasticsearch installation: first steps” which may not contain the words “how to set up”.
Traditionally, information retrieval in Elasticsearch is based on the BM25 ranking function but with the addition of vector search and the ability to perform inference using large NLP models there are now additional options to improve search relevance in the Elastic platform.
This talk will give an overview of the problem and the solution space. We will then give an introduction to how Elastic engineering is approaching the problem, and an overview of our roadmap and the steps we are taking to improve relevance.
Speaker: Steve Dodson, Distinguished Engineer, Elastic

[@vladiraress]

Nice content as always.
I have a question about indexing. If the entity I want to index has a list of other entity, what would be the best approach for indexing?
1. Indexing a document for every combination of the fields and then make aggregations when querying?
2. Index the document with nested field types
The implementation also needs to support pagination (I think the first approach won't be able to do this)

[@shaktisd]

Can you please share link for the tutorial of the same

[@hamzashabbir3589]

Do you guys have a Slack community 🤔

[@philippkrenn5779]

yes, ela.st/slack

[@hamzashabbir3589]

@@philippkrenn5779 Thanks so much. Joined the Slack Channel.