Cornell News; Bill Steele

Researchers from Cornell University, the University of Pittsburgh, and the University of Utah have launched a project seeking to train computers to scan text and make a determination as to whether its contents are fact or fiction. The Department of Homeland Security created the consortium of three universities as one of four that are exploring sophisticated techniques for information analysis and security-related computational technologies. "Lots of work has been done on extracting factual information--the who, what, where, when," said Cornell computer science professor Claire Cardie. "We're interested in seeing how we would extract information about opinions." The research aims to bridge the gap between the distinctly human form of intuitive intelligence and the more literal machine intelligence by giving meaning to sentences through novel machine-learning algorithms. Cardie says her team is also working to rate the sources of a work that a writer might cite. "We're making sure that any information is tagged with confidence. If it's low confidence, it's not useful information," she said.