Computational model of speech understanding
Daniel Müller, Philippe O. A. Navaux · International conference on Artificial intelligence and applications · 2006
This paper proposes a speech comprehension computational model based on neurocognitive researches. The computational representation uses techniques as wavelets transform and connectionist models. The speech signal codification and data prosodic extraction are derived from wavelet coefficients. Moreover, the connectionist models are used to perform syntactic parsing and prosodic-semantic mapping. Thus, the computational model applies three approaches: the application of wavelet coefficients as input in connectionist language analysis, the use of SARDSRN-RAAM system to syntactic analysis as well as the proposition of prosodic-semantic maps to language contexts definition.