-
Department Info
-
-
Admissions
-
-
Academics
-
-
People
-
-
Research
-
-
-
CSE Profile: William Schuler
![]() Associate Professor (612) 626-7502 Office: EE/CS 5-225F InterestsSpeech and natural language processing, spoken language interfaces, statistical (Bayesian) models of language, and in particular statistical models of natural language meaning that are grounded in perception EducationPh.D. 2003, Computer and Information Science, University of Pennsylvania M.S. 1997, Computer and Information Science, University of Pennsylvania B.S., 1995, Computer Science, University of Michigan B.A., 1992, English Literature, University of Michigan | ResearchMost of my research is directed toward the development of statistical models of language for spoken language interfaces that not only generate probabilities for the words and syntactic structure of input utterances, but also generate probabilities for the meanings or denotations of utterances in the context of an interface's underlying application environment or world model. These denotations are then used to condition the probability estimates of hypothesized syntactic structures and words of an utterance. My students and I are currently working on further extending these models to incorporate uncertainty in the agent's perception of its environment, as an interface to a robot, sensor, or other real-world agent would require. These more plausible models of natural language utterances, in which strings of words are generated from some notion of intended meaning, are expected to improve the recognition accuracy of spoken language interfaces to artificial agents (e.g. database assistants, simulation characters, or mobile robots) -- applications which can provide a spoken language interface with a ready environment context in which to interpret input directions (e.g. an application database, 3-D simulation, or perceived model of the real world). Such interfaces offer the promise of a flexible, intuitive means for untrained users to interact with artificial agents, particularly in mobile or hands-free applications where keyboard-, mouse-, and monitor-based interfaces are not convenient. Personal Home Page |
Related Links


