WHO:	Harlan Harris , Graduate Student University of Illinois at Urbana-Champaign
TOPIC:	The HORROR: Speech Errors and Phonological Production Models
ABSTRACT:	Computational modeling of speech errors has been an important aspect of the study of language production. Models developed over the last twenty years have proposed mechanisms for the sequential production of speech segments, along with explanations for the particular patterns of speech errors seen in corpus and experimental data. Different models have supported different theories of linguistic representation and information processing in the language system. In this talk I overview the phonological production system and the patterns of speech errors that have been observed, then survey the various sorts of models that have been proposed previously. I then introduce HORROR (Holographic Reduced Representations for Oscillator Recall), a new model of phonological production that accounts for many speech error patterns by combining a hierarchical control signal with a distributed associative memory. I will argue that HORROR provides support for a view of phonological production in which planning a nd memory processes are used to combine linguistic items (phonemes of a word) with temporal schemas (syllable and prosodic frames). ›
WHEN:	3/17/2003 4:00:00 PM
WHERE:	Meliora 418

Events Homepage