Abstract:
There is an increasing demand for computer software which can provide useful personalised feedback to English as a Second Language (ESL) speakers on prosodic aspects of their speech, to supplement the shortage of ESL teachers and reduce the cost of learning.
This thesis concentrates on constructing such an Intelligent Computer Aided Instruction (ICAI) prototype system, particularly focusing on one component — the Speech Analyser. The speech analyser recognises a user's speech, identifies the rhythmic stress pattern in the speech, discovers stress and rhythm errors in the speech, and provides reports for the other component generating personalised feedback to the user on ways of effectively improving the prosodic aspects of the speech.
We build an Hidden Markov Model (HMM) based speech recogniser to recognise a user's speech. A set of parameters for constructing the recogniser is investigated by an exhaustive experiment implemented in a client/server computing network. The exploration suggests that the choice of parameters is very important. We build stress detectors to detect the rhythmic stress pattern in the user's speech by using both Support Vector Machine (SVM) and Decision Tree (DT) techniques. The detector using SVM outperforms the one using DT. It suggests that SVM is more suitable for a relatively large data set with all numeric data than DT. We build an error identifier to automatically identify stress and rhythm errors in the user's speech. A two-layer phoneme alignment algorithm using the Needleman/Wunsch technique is developed to facilitate the prosodic error identification problem. Our study also suggests that the foot comparison method is better than Vowel Onset Point comparison method for automatically identifying the main rhythm errors in the user's speech.
