Abstract

A number of real-world applications require modeling uncertainty. Examples include information extraction, information integration, data collected from sensor networks, social networks and scientific databases just to name a few. Traditionally, most of these applications have been tackled in isolation even though the core operations of modeling and reasoning under uncertainty are common to all of them. Probabilistic databases, an area of research in the intersection of machine learning and large-scale data management, hold the promise to provide a common framework that solves all applications requiring uncertainty management and enhance their utility by providing an easy-to-use, declarative querying interface. In this talk, I will describe some of my recent contributions to this area of research touching upon efficient query evaluation, different representation schemes in increasing order of complexity and efficient inference algorithms. Designing a common framework requires that we first gain an in-depth understanding of the targetted range of applications. I harbour an independent interest in a number of such applications and will also be describing my recent work in information retrieval, information extraction and information integration.