Abstract: We are going to start with a brief refresher on the concepts of optimising sequential decision making via Reinforcement Learning (RL). We will then look at some recent success of RL for generating human-like relevant answers. We then will discuss a what types of discrete optimisation problems occur in the broad field of Information Retrieval, some already tackled by RL approaches, some tackled with simpler heuristic-based approaches. This will be followed up by some of my own thoughts of what other research questions in Information Retrieval may be benefitted from RL-based approaches.
10:10 - 10:30
Coffee break
10:30 - 11:30
Invited talk2: Modeling, Evaluation, and Mitigation of Filter Bubbles and Matthew Effects in Offline RL-based Recommender Systems
