Abstract

Abstract: The Spoken CALL Shared Task is a series of tasks that assess English responses of German-speaking Swiss teenagers according to grammar and meaning correctness. This talk presents the systems developed by the University of Birmingham (UoB) for the Spoken CALL Shared Tasks. UoB systems consist of an Automatic Speech Recognition (ASR) system and a Text Processing (TP) system. The target speakers of the ASR are German-speaking children aged between 12 and 15 years, both factors—children and non-native—make ASR more challenging. This work investigates different techniques to compensate for the effects of non-native and children on the performance of ASR systems. The study mainly utilises hybrid DNN -HMM systems with conventional DNNs, LST Ms and TDNN models. For the text processing component, we focused on providing accept/reject feedback to learners based on the text generated by the ASR system. A rule-based and a machine learning-based system were proposed for making the judgement, several aspects of the systems were evaluated. The influence of the ASR system on the text processing system was explored.

Bio: Mengjie Qian is a Research Assistant in the Speech Group of the Machine Intelligence Laboratory, Engineering Department of Cambridge University. Mengjie did her PhD in the Speech Group at the University of Birmingham, supervised by Prof. Martin Russell. She submitted her thesis in January 2021. The title of her thesis is “Computer Analysis of Children’s Non-Native English Speech for Language Learning and Assessment”. Her research interests include speech recognition, speaker diarisation and natural language processing.