Revisiting the Basics: How Round Robin Improved Search Relevancy

In the rapidly evolving landscape of search technologies, the allure of novel, complex algorithms often overshadows the potential of simple, traditional solutions. This talk aims to shift the focus back to the basics, demonstrating how a standard Round Robin algorithm improved search relevance in an enterprise-level application. Our journey begins with a dual challenge: aligning the comprehensive view of our data with individual verticals and improving its underperforming search relevance metrics. We experimented with three methods to merge our vertical datasets into the comprehensive view, finding a standard Round Robin algorithm to be the most effective. We also tailored this algorithm to prioritize certain datasets, aligning with business requirements and customer expectations. The results were striking. All our internal relevance metrics, for both implicit and explicit human judgements, improved. Notably, our explicit judgement ERR score jumped from 0.49 to 0.61, underscoring the effectiveness of our approach. Join us as we explore how revisiting the basics led to gains, and why the Round Robin algorithm, a simple yet powerful tool, can still hold surprising value in today’s advanced search technology landscape.

Joelle Robinson

Moody's

Joelle Robinson is a Senior Software Engineer at Moody's with three years of experience on the search team. She played a crucial role on many projects during her tenure at Moody’s like facilitating the transition from Solr to AWS OpenSearch. Currently, Joelle is focused on optimizing search relevance for keyword and semantic search functionalities for Moodys.com.