Pronunciation training with non-native automatic speech recognition

Abstract

In this thesis I investigate the way that automatic speech recognition can be best implemented so as to be smoothly integrated in the classes of teaching a foreign language. The main goal of this thesis was to implement a speech recogniser who would identify pronunciation mistakes made by Greek native speakers when speaking German. In particular, I am investigating the recognition of duration of vowels. The proposed recogniser shows that there is a difference between conventional recognisers and those who are directed to teaching of pronunciation, given that more principles need to be taken into consideration and more fields need to work together for a satisfactory result.