Multiple ensemble learning methods for Mandarin multisyllabic pronunciation assessment

The automatic pronunciation assessment technology based on artificial intelligence can well help language learners understand their pronunciation level and give timely feedback. The main reasons for the lack of relevant datasets and the immaturity of multisyllabic pronunciation assessment techniques are that there are few studies on the assessment of Mandarin multisyllabic pronunciation. In order to solve the above problems, this paper establishes a phoneme- level Mandarin multi-syllable assessment annotated dataset, designs various acoustic models with different network structures, and combines ensemble Learning algorithms to conduct a multifaceted pronunciation assessment comparison experiment on the self-built Mandarin multi-syllable assessment dataset. The experimental results show that the performance of the model with ensemble Learning algorithms is significantly improved compared with the traditional regression model.