A computational model of sound perception in a second language

Abstract

This thesis reposts on an investigation based around a number of studies that use computational statistical pattern recognition and machine learning techniques to simulate listeners ¿non-native consonant perception and learning development.More specifically, listeners¿ second language (L2) to native language (L1) consonant assimilations as well as the early stage development of learnes¿L2 consonant acquisition have been modeled.The computational modeling approaches used in this thesis have been shown to provide quantitative predictions and data visualization for L2 perception studies, and as such act as a valuable complementary approach to those used in traditional behavioral and theoretical studies in sound acquisition.