Browse Bar: Browse by Author | Browse by Category | Browse by Citation | Advanced Search
Author(s): Hoffberg, S.M. | Hawkins, S. |
Year: 2003
Citation: JOURNAL OF PHONETICS Volume: 31 Issue: 3-4 Pages: 373-405
Abstract: This paper aims to show how we can make progress in elucidating how people understand speech by changing our focus of inquiry from abstraction of formal units of linguistic analysis to a detailed analysis of global aspects of the communicative situation, of which speech is just one part. It uses evidence of (a) the communicative importance of fine phonetic detail and (b) exemplar memory for speech to explore the idea that, in certain normal, easy conversations at least, one may interpret the meaning of an utterance directly from the global sound pattern; reference to formal linguistic units of analysis, such as phonemes, words, and grammar, is incidental; circumstances dictate whether such reference takes place at all, and if it takes place, whether it does so after the meaning has been understood, before it has been understood, or simultaneously with the construction of meaning. The implications of this position are that speech perception does not demand early reference to abstract linguistic units, but instead, to flexible, dynamic organization of multi-modal (and modality-specific) memories; and that models of speech perception should reflect the multi-purpose function of phonetic information, and the polysystemic nature of speech within language. A preliminary model that reflects this theoretical position, Polsyp, is described. Polysp has intellectual antecedents in Hebbian principles, and current relevance to adaptive resonance theory (ART). Neuronal bases for initial processing of exemplars are briefly discussed. Hierarchical and more abstract processing arises in an ART-like, self-organizing dynamic system in which, once processing has begun, the sensory input is not effectively distinguishable from top-down knowledge. Understanding meaning is more important than identifying linguistic structure, and processing is strongly guided by rhythmic and attentional factors.
Topics:
Speech and Hearing,
Applications:
Information Fusion,
Models:
ART 2 / Fuzzy ART,