Objective
The aim of this RTN is to elucidate the interaction of knowledge and sensation in speech perception, by developing a psychological/computational model of speech recognition that reflects the flexible response to the availability of information in the sign al that is comparable to that of humans listening to their native language (L1). A central focus is to track how phonetic information, especially fine phonetic detail that varies systematically with linguistic/interactional structure and function, is used under different circumstances: when the listener has appropriate linguistic-phonetic knowledge (listening to L1), possibly inappropriate or inadequate knowledge (listening to a foreign language (L2)), and inadequate access to the signal (listening in adverse conditions). A related focus is to elucidate how phonetic information contributes to understanding: modelling will reflect known attributes of auditory processing, and linguistic and statistical properties of language units. The ultimate aim is to provide models of speech perception that closely reflect the flexibility and robustness of human speech recognition, that pave the way for the next generation of robust automatic speech recognition and text-to-speech machines, and promise to offer a new theoretical basis for L2 teaching and for diagnosis and treatment of speech disorders. Cross-disciplinary training is essential to achieve these aims: no single discipline or pair of disciplines currently teaches all the information and skills required to achieve them, many work independently of the others, and some assume old models that have been superseded in other disciplines. By nurturing a sufficiently large group of scientists who will lead a sustained, cross-disciplinary effort to understand human speech processing, we hope to provide the basis for more integrated work and faster progress in speech understanding in the future.
Fields of science
Call for proposal
FP6-2005-MOBILITY-1
See other projects for this call