Abstract

Wikipedia’s impact has been revolutionary. The collaboratively edited encyclopedia has transformed the way many people learn, browse new interests, share knowledge and make decisions. Its information is mainly represented in natural language text. However, for many tasks more structured information is useful because it better supports pattern analysis and decision-making. In this talk I will describe multiple research components useful for building large, structured knowledge bases, including information extraction from text, entity resolution, joint inference with conditional random fields, probabilistic databases to manage uncertainty at scale, robust reasoning about human edits, tight integration of probabilistic inference and parallel/distributed processing, and probabilistic programming languages for easy specification of complex graphical models. I will also discuss applications of these methods to scientometrics and a new publishing model for science research.

Joint work with Michael Wick, Sameer Singh, Karl Schultz, Sebastian Riedel, Limin Yao, Brian Martin and Gerome Miklau.