Susan W. McRoy

Computer Science
University of Wisconsin-Milwaukee


Computer Science, University of Wisconsin-Milwaukee, 3200 North Cramer Street, Milwaukee, WI 53211.

(414) 229-6695, (414) 229-6958.



Speech and Natural Language Understanding


Dialogue, misunderstanding, abduction, medical applications of artificial intelligence


One of the biggest roadblocks preventing people from using natural language to interact with computers is that understanding language perfectly is difficult, even for people. Recently, theories have been proposed explaining how interactors can behave appropriately---even if they differ about the meaning of some part of the discourse. This project aims to extend these ideas so that their effectiveness can be tested and their insights put to use.

Extending earlier work comprises four tasks: elicitation and collection of examples; theoretical extensions; evaluation of the extended model within its logical framework; and the construction of applications supporting a user study. The planning activities are to include a review of the relevant literature; preliminary experimentation with candidate methods of data collection and validation; discussions with researchers whose expertise bears on these issues; and the construction of two prototype applications that discuss medical information with doctors and patients, respectively. Upon completion of these activities, a detailed plan for extending prior approaches to misunderstanding will be developed. If we can enable computers to address misunderstanding, they will be better able to understand natural language reliably and robustly, which is essential if they are to help those who need them most, non-computer-experts.


McRoy, S. W., and G. Hirst, ``The repair of speech act misunderstandings by abductive inference'', Computational Linguistics. To appear December 1995.

McRoy, S. W. ``Misunderstanding and the negotiation of meaning'', Knowledge-Based Systems. March 1995.

Hirst, G., S. McRoy, P. Heeman, P. Edmonds, and D. Horton, ``Repairing conversational misunderstandings and non-understandings,'' Speech Communication. 15, December 1994, 213--229.

McRoy, S. W., ``Abductive Interpretation and Reinterpretation of Natural Language Utterances'', CSRI Technical Report No. 288, University of Toronto, Department of Computer Science.

McRoy, S. W. and G. Hirst, ``Abductive explanations of dialogue misunderstandings'', Proceedings, Sixth Conference of the European Chapter of the Association for Computational Linguistics, Utrecht, April 1993.

McRoy, S. W., ``Using multiple knowledge sources for word sense discrimination'', Computational Linguistics, 18(1), March 1992, 1--30.


The general goal of this researcher is to specify the reasoning that underlies communicative interaction. To begin, we have looked at recognizing and repairing speech-act misunderstandings, because any theory that is sufficient to capture general dialogue must be able to account for this phenomena.

Speech acts are the communicative actions that speakers perform by saying things in a certain way in a certain context. For example, communicators can make statements, ask questions, or request actions of others. Moreover, conventional usage allows them to use the same surface form in different contexts to do different things. Thus, the declarative sentence ``It is cold in here'' might be used either to inform or to request that the recipient address the problem. Similarly, clicking on a menu button marked ``Explain'' might be meant as a request to clarify the system's most recent action or it might be meant as a request to explain the reasoning that led to that action.

Typically, the interactants' expectations about the conventional structure of interaction make it easy for them to interpret such actions as they were intended. However, when one participant fails to consider the other's expectations or to notice apparent incoherence, communicative breakdown results. Misunderstandings occur when one participant obtains an interpretation that is complete and correct from her perspective, but which is, however, not the one that the other participant expected her to obtain. Having different beliefs about interpretation (rather than errors in prior belief) is what distinguishes misunderstanding from misconception.


James Allen and Raymond Perrault. Plans, inference, and indirect speech acts. In 17th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference, pages 85--87, 1979.

Kent Bach and Robert~M. Harnish. Linguistic Communication and Speech Acts. The MIT Press, 1979.

Sandra Carberry. Plan Recognition in Natural Language Dialogue The MIT Press, Cambridge, MA, 1990.

H. Paul Grice. Meaning. The Philosophical Review, 66:377--388, 1957.

Diane J. Litman. Linguistic coherence: A plan-based alternative. In 24th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference, pages 215--223, New York, 1986.


Adaptive Human Interfaces; Usability and User-Centered Design; Other Communication Modalities