Building Embedded Conceptual Parsers



for the degree


Field of Computer Science


William Alan Fitzgerald


Dissertation submitted and approved December 1994

(c) Copyright by William Alan Fitzgerald 1995

All Rights Reserved


Building Embedded Conceptual Parsers

William Alan Fitzgerald

This dissertation describes how to build conceptual parsers (that is, natural language understanding systems built on semantic and pragmatic principles) that are embedded into application programs. A new architecture for building such parsers, indexed concept parsing, is described. Indexed concept parsing is a case-based reasoning approach to parsing, in which underlying target concepts (that is, those conceptual representations of the application program identified as important to recognize) are associated with sets of index concepts. Each index concept is associated with sets of phrasal patterns. At run time, the parser looks for phrasal patterns in input text, and the index concepts recognized thereby are used to appraise the best matching target concepts. The architecture defines a range of parsers, in which the complexity of the index concept representations can vary according to the needs of the application program: index concepts can be key words, synonym sets, representations in an abstraction hierarchy, or representations in a partonomic hierarchy. Indexed concept parsing was developed to build parsers for Casper, an interactive learning environment designed to teach customer service representatives how to solve customer problems, and TransAsk, a multimedia system for transportation planners. Indexed concept parsing proved robust (for example, the Casper parser had an accuracy rate ranging from 83-96%), yet required minimal knowledge representation. A methodology for building an indexed concept parser is given, and evaluation metrics are described. Another parser, based on Direct Memory Access Parsing (DMAP), and developed for the Creanimate biology tutor, is also described, as well as a DMAP parser for Casper. Indexed concept parsing and DMAP are contrasted as architectures for building embedded conceptual parsers.


Chris Riesbeck was the first member of the faculty I met at the Institute for the Learning Sciences. He fascinated me with the idea of parsers that could get bored if they had to, and then (patiently, over many months) taught me why one might have to. He has modelled many things for me--elegant programming style; how to do research in artificial intelligence and natural language processing; how to write and act as a member of a community of scholars. His advice, patience, example and humor have shaped my past few years.

Larry Birnbaum and Ken Forbus also served as members of my dissertation committee. As others before me have said, I cannot underestimate the importance of Larry's introduction to the ways of artificial intelligence research--how hard, how intriguing, how much fun AI research can be. Ken works very hard, and is very insightful--a combination before which I've often stood in awe. It was an honor to have them on my committee.

Roger Schank has gathered an astoundingly wonderful group of researchers at the Institute for the Learning Sciences, and, through his vision for new ways of educating, has provided many interesting opportunities for working on intriguing projects. I am in his debt. I worked on building parsers that were embedded in application programs, and I am indebted to the people who built those application programs. Danny Edelson and Alex Kass were both graduate students before me, and I profited from their advice and example, and am glad for their friendship. Ray Bareiss encouraged and challenged me as I worked on the challenge of adding a parser to a large-scale system. Other faculty and staff at Northwestern have had their impact in ways too numerous to mention--but I cannot fail to mention, by name at least, three others who were my teachers: Gregg Collins, Paul Cooper, and Dedre Gentner.

Grad school often provides a place for friendships to develop, and this was true among those who took the qualifying exam the same year as I: Chip Cleary, John Cleave, David Foster, John Hubbel, Steven McGee, Mike Sang, Nikitas Sgouros, Eric Shafto, Ian Underwood, and Chris Wisdo. Chip has been a steady friend and source of insights, suggestions and good coffee. Eric can convince me of anything (even when we both know it's wrong). David, conservative/rocker/professional volleyball player/AI hacker, is someone I never have enough time to talk to. Chris has been a constant intellectual sparring partner and colleague. Eric Goldstein always had an encouraging word about my research.

Louise Pryor was a master at the mot juste, including wishing me the best of British luck exactly when I needed it. Andy Fano helped me sharpen the distinctions among different types of conceptual parsers. The members of the Millard Fillmore Memorial (Film Noir) Lunch group patiently listened to my ramblings even as I impatiently listened to theirs. Members of the administrative and support staffs, including Elizabeth Brown, Amy Juhl, Teri Lehmann, Heidi Levin, Sharon Lewis, Doug MacFarlane, Paul Ring, Michelle Suran and Tina Turnbull made the Institute a (usually) efficient as well as an (always) humane place to be. Heidi Levin also did the momentous favor of proof-reading the technical report version of the dissertation. Friends outside the Institute also need to be acknowledged, especially my fellow dissertator Terry Wolfer; fellow household members Mark Knoper, Paul Tretbar, Bev Wiebe and Dan Yutzy; and the members of the Elmwood Small Group and the JJSG at Reba Place Church.

It's impossible to acknowledge how important Bess Fitzgerald (my wife) has been to my getting here. She has been constant in her support and love along the way, never letting me quit or become overly discouraged. Our children, Mark William Fitzgerald and Jane Elizabeth Fitzgerald, are just plain delightful and a joy to come home to.

This work was supported in part by the Defense Advanced Research Projects Agency, monitored by the Office of Naval Research under contracts N00014-91-J-4092 and N00014-90-J-4117. The Institute for the Learning Sciences was established in 1989 with the support of Andersen Consulting, part of The Arthur Andersen Worldwide Organization. The Institute receives additional support from Ameritech and North West Water Group plc, Institute Partners.


For my father, Harold Fitzgerald,who gave me a love for books,and to the memory of my mother,Muriel Shaughnessy,who first took me to college.