“Engage moi”: From retrieval effectiveness, user satisfaction to user engagement

The effective prediction of a click remains a primary challenge in the areas of search, digital media and online advertising. In the context of search, satisfying a userâĂŹs information need by returning results that they will click on is an important objective in any information retrieval system. Consequently, information retrieval systems have had a long and varied history of how to evaluate their effectiveness of responding to a given query. However, building such a system that not only only returns relevant results to a user query but also encourages a long-term relationship between the user and the system is far more challenging. In this talk, we review the current state-of-the-art evaluation approaches for search before exploring other ways of quantifying more long-term engagement measures. Finally, the talk ends with a proposal of how the two approaches can be considered together to create a service that optimises for the query and the longer term engagement aspects.