Introducing OpenSearch’s Search Relevance Workbench! - Eric Pugh & Stavros Macrakis

Introducing OpenSearch’s Search Relevance Workbench! - Eric Pugh, OpenSource Connections & Stavros Macrakis, Amazon Web Services

This talk introduces you to OpenSearch’s integrated new capability to measure and enhance search quality: Search Relevance Workbench (SRW).

We’ll introduce the mental model for improving search quality espoused by OpenSource Connections. We'll show you how SRW in OpenSearch 3 implements this approach.

We’ll teach you the three phases of improving search using SRW.

Understand Search Quality:
* Define explicit Search Configurations. This is a new concept in OpenSearch for structuring query plans
* Run evaluation metrics like NDCG.
* Scale up with the “LLM as a Judge”.
Improve Search Quality:
* SRW makes it easy to evaluate changes to your query templates. We’ll show you how to make new search configurations in OpenSearch and quickly test them.
* We’ll introduce the new Hybrid Optimizer capability in SRW that goes beyond manual experimentation by introducing automated experimentation.
Deploy new Search Configurations:
* We’ll demonstrate how to A/B test your search configurations using the User Behavior Insights event tracking system.

At the end of this talk, you’ll leave understanding how to take use these new capabilities in OpenSearch 3 to enhance your own Search.

Видео Introducing OpenSearch’s Search Relevance Workbench! - Eric Pugh & Stavros Macrakis канала OpenSearch