Probabilistic Information Retrieval

This lecture introduces the concept of probabilistic information retrieval, which aims to provide a principled approach to dealing with uncertainty in user information needs. It covers probabilistic IR models, query likelihood models, language modeling, and the use of probabilistic language models. The lecture also discusses the probability of creating a query, learning the model, issues with maximum likelihood estimation, smoothing techniques, and probabilistic retrieval. Additionally, it explores the Rocchio algorithm for relevance feedback and the SMART algorithm for practical relevance feedback approximation. Practical considerations and assumptions underlying these algorithms are also addressed.