Natural language research announcements

["Alexandre Klementiev" <aklement AT gmail.com>]

Title: 
Articulatory Feature-based Pronunciation Modeling and Applications to Lipreading


Speaker:  Karen Livescu, Toyota Technological Institute - Chicago





Date: 
4/22/08, 
2:00pm
Location:     
2269 Beckman

Abstract:
Spoken language technologies, such as automatic speech recognition and synthesis, typically treat speech as a string of phones.  In contrast, humans produce speech through a complex combination of semi-independent articulatory trajectories.  Some recent theories of phonology acknowledge this, and treat speech as a combination of multiple streams of variables such as articulatory features.  I will present a feature-based approach to pronunciation modeling, using dynamic Bayesian networks.  In this class of models, the great variety of pronunciations seen in conversational speech is explained as the result of asynchrony among feature streams and changes in individual feature values.  I will present ways in which the factorization of speech into features can be useful in speech recognition, with emphasis on recent experiments in visual speech recognition (lipreading).