Ninth Biennial Rice University Symposium on Linguistics

Speech Perception in Context:

Beyond Acoustic Pattern Matching

March 13-16, 2002

Prospectus

The satisfactory modeling of the processes involved in the perception of speech has proven to be both a fascinating and elusive ambition. Decades of research have shown that the quest for acoustic features that are uniquely correlated with speech segments is not adequate: units that are composed of very different acoustic properties can be perceived as the "same" in different contexts, and units that are composed of similar acoustic properties can be perceived as "different". Furthermore, humans are able to make sense of acoustic signals that vary according to physical and social characteristics of the speaker, that vary according to the speech rate and social characteristics of the speech context, and that often occur in noisy environments.

The ability to recognize and categorize speech sounds in these imperfect and variable conditions is a direct result of the fact that humans use many sources of information in the process of speech perception. Humans interpret speech much better than computers because humans use, for example, cross-modal cues (such as vision), context cues, speaker normalization processes, their expectations, and categorical perception, each c ontributing to the human capacity to make sense of the acoustic stream of information they are confronted with. The way that these sources of information interact, however, has proven difficult to describe, and new prospective sources of information cont inue to be discovered.

The purpose of this symposium is to bring together researchers who investigate what knowledge systems are involved in the perception process, and how these systems interact. Although they are from such different fields as linguistics, psychology, neurology, and electrical engineering, they share an interest in examining speech perception as a phenomenon that moves beyond mere acoustic phonetic pattern matching (such as those used by computerized speech recognition system), and instead appeal to the entire range of cognitive systems involved in perceiving speech in more natural contexts.

This symposium is sponsored by the Rice Department of Linguistics, the School of Humanities,
the Douglas Mitchell Fund and by the Center for the Study of Cultures

Program Rice Linguistics Campus Map

Site Designed and maintained by Hilary Young
Last updated: January 25, 2002