Relevance Feedback Methods 2 Using Rf in the Probabilisitic Model 3 Reranking with Rf

For the purpose of this lecture, let us assume that Relevance Feedback (abbreviated RF) is available in the form of explicit binary (yes/no) judgments for each of the top-k retrieved documents. Although RF can potentially be used in any setting, it is most natural in the probabilistic retrieval model since there is a notion of relevance in the form of a random variable. So let us first discuss how we can use feedback in the probabilistic case.