Multi Period Information Retrieval and Optimal Relevance Feedback using Dynamic Programming

In Multi Period Information Retrieval we consider retrieval as a stochastic yet controllable process, the ranking action during the process continuously controls the retrieval system's dynamics, and an optimal ranking policy is found in order to maximise the overall users' satisfaction. Our derivations show interesting properties about how the posterior probability of the documents relevancy evolves from users feedbacks through clicks, and provides a plug-in framework for incorporating different click models. Based on the Multi-Armed Bandit theory, we propose a simple implementation of our framework using a dynamic ranking rule that takes rank bias and exploration of documents into account. We also look at relevance feedback, where we use dynamic programming to make a ranking decision at each iteration according to the overall future payoff, rather than the instant reward. We show that document correlations used in result diversification have a significant impact on relevance feedback and its effectiveness.