A high performance automatic speech recognition (ASR) system is an important constituent component of an automatic language assessment system for free speaking language tests. The ASR system is required to be capable of recognising non-native spontaneous English speech and to be deployable under real-time conditions. The performance of ASR systems can often be significantly improved by leveraging upon multiple systems that are complementary, such as an ensemble. Ensemble methods, however, can be computationally expensive, often requiring multiple decoding runs, which makes them impractical for deployment. In this paper, a lattice-free implementation of sequence-level teacher-student training is used to reduce this computational cost, thereby allowing for real-time applications. This method allows a single student model to emulate the performance of an ensemble of teachers, but without the need for multiple decoding runs. Adaptations of the student model to speakers from different first languages (L1s) and grades are also explored. TEACHER-STUDENT TRAINING OF ACOUSTIC MODELS FOR AUTOMATIC FREE SPEAKING LANGUAGE ASSESSMENT
There is a high demand around the world for the learning of English as a second language. Assessment of a learner’s language proficiency is a key part of learning both in measuring progress made and for formal qualifications required e.g. for entrance to university or to obtain a job. Given the high demand from English learners, it will be very difficult to train sufficient examiners and the introduction of automatic markers will be beneficial especially for practice situations.