|
|
| Fall 2003: CLSP Seminar Series | Tuesday, May 13, 2008 |
Spontaneous speech contains a variety of different kinds of speech errors. Speech errors are one of the things that makes spoken language processing so difficult, and any robust system for extracting "deep" information from speech will probably have to find a way of dealing with them. Similar problems with speech errors occur in a simpler setting in syntactic parsing of transcribed speech, and methods for dealing with them in this setting may generalize to real speech recognition. It turns out that speech repairs, which are the focus of this talk, are problematic even for parsing transcribed speech. The talk describes a general architecture for detecting speech repairs and discusses features that turn out to be useful for this task.
Biographical Information
Mark Johnson is Professor of Cognitive and Linguistic Sciences and Computer Science at Brown University, and is currently a visitor at MIT CSAIL. He is also currently President of the Association for Computational Linguistics.
| The Center for Language and Speech Processing The Johns Hopkins University 3400 North Charles Street, Barton Hall Baltimore, MD 21218 | |||||
 | Telephone: (410) 516-4237 | | Fax: (410) 516-5050 | | E-mail: clsp@clsp.jhu.edu |