Abstract

Information retrieval has traditionally at the granularity of the document - Web search engines are an obvious example of this. In large (> 10**6 document) collections, this poses a serious difficulty to users, given that the result set for many queries measure in the thousands or tens of thousands of documents. Question answering has gained significant interest as a means of reducing the result set for fact-based user queries to, literally, the answer. I will introduce the general problem of question answering and then discuss my approach, which involves a number of key assumptions to scale question answering to Web search-level responsiveness. In particular, by shifting natural language parsing forward in the process, we can amortize this very expensive step against a number of downstream extraction processes that mine the text for named entities, relationships, etc. Redefinition of extraction specifications hence does not require reparsing of the source text. I will also discuss an implementation of a grep-like extraction grammar designed for predicate-based extensibility and how it is employed in mapping sentence parse trees to relational structure. Finally, I'll demonstrate the overall approach is really feasibly through a short demonstration.

David Eichmann is the Director of the School of Library and Information Science at The University of Iowa and a member of the faculty in the Department of Computer Science. He received his B.S., M.S. and Ph.D. in Computer Science from Iowa before taking positions at Seattle University, West Virginia University and the University of Houston - Clear Lake, where his research and teaching focused on software reuse and repository management. His current research involves extraction and retrieval of information across a variety of media, including text, imagery and video.