A software engine that pulls together facts by combing through more than 500 million Web pages has been developed by researchers at the University of Washington. The tool extracts information from billions of lines of text by analyzing basic relationships between words.