Raytheon BBN Technologies

Though search engines provide documents in response to a query, the goal of our research in open-domain question answering is to find answers to queries using textual sources, e.g., the World-Wide Web, without the benefit of a narrow domain to constrain interpretation. Our approach is a hybrid of information extraction, natural language parsing, and information retrieval. Answers are assembled across multiple documents to answer complex questions. For example, we assemble facts from across many documents in answer to a request for biographical information, an organizational profile, or an extended definition. The work is multi-lingual, supporting English questions across English, Arabic, and Chinese sources.