Generating hypotheses from the web

Wei Jin, Rohini K. Srihari, Abhishek Singh

Abstract: Hypothesis generation is a crucial initial step for making scientific discoveries. This paper addresses the problem of automatically discovering interesting hypotheses from the web. Given a query containing one or two entities of interest, our algorithm automatically generates a semantic profile describing the specified entity or provides the potential connections between two entities of interest. We implemented a prototype on top of the Google search engine and the experimental results demonstrate the effectiveness of our algorithms.