Performance Measures for Phone - Level Pronunciationteaching in Calls

This work presents a general development framework for automatic pronunciation assessment within computer-assisted language learning (CALL) together with several reenements of a previously described pronunciation scoring method. This method utilises a likelihood-based`Goodness of Pronunciation' (GOP) measure which in this work has been extended to include individual thresholds for each phone based on both averaged native con-dence scores and on rejection statistics provided by human judges. These statistics where provided through a speciically recorded and annotated database of non-native speech. Since pronunciation assessment is highly subjective, a set of four performance measures has been designed, each of them measuring diierent aspects of how well computer-derived phone-level scores agree with human scores. These performance measures are used to cross-validate the reference annotations and to assess the basic GOP algorithm and its reenements. The experimental results suggest that a likelihood-based pronunciation scoring metric can achieve usable performance, especially after applying the various enhancements.