STELLA enables researchers and coders to deploy and evaluate search and recommendation algorithms in a real-world scenario. These so-called "Living Labs" provide a user-focused and thus more realistic evaluation approach.
What are the crucial steps when implementing a participant-system in STELLA?
In academia, we often rely on standard test collections with static relevance judgments that somewhat abstract away the specific information need of single users. Thus, evaluating search engine algorithms with real-world users is a desiderata. Finally, we want to know how users can benefit from innovative search engines. The STELLA project tries to bridge this gap with the help of an innovative infrastructure that makes it possible to evaluate experimental search algorithms in live environments. In this post we want to outline the general idea of our infrastructure.