OpenSearch Search Relevance Workbench

The Search Relevance Workbench is a set of tools that support users improving search relevance. Users can find all features to optimize search result quality in one place: analyze track user behavior with UBI, manage judgements, run search quality experiments, analyze experiment results, explore search quality tuning options, …

This repository covers the backend part which is implemented as an OpenSearch plugin.

Example Usage

./gradlew build
docker compose build && docker compose up

After the container is running:

./scripts/initialize-ubi-indexes.sh

Then index the ESCI data:

cd data-esci
./index-ubi-queries-events.sh
./index-ecommerce-products.sh

Then to make judgments:

cd scripts
./create-coec-judgments.sh

Then a query set still under scripts):

./create-query-set-using-random-sampling.sh

To see the judgments

./get-judgments.sh

And to see the query sets:

./get-query-sets.sh

All of the scripts under scripts/ can be used similarly.