Linguistic features and automatic scoring of Aptis speaking performances

The purpose of this study is to investigate linguistic features of spoken performances on Aptis, which can be automatically extracted and further utilised for automated speaking assessment, and to actually predict Aptis English proficiency levels through an automated scoring system. The project aims to adapt automatic speaking proficiency rating software, initially developed by the research team, which automatically scores unconstrained English speech proficiency (e.g., Johnson, Kang & Ghanem, 2016; Johnson & Kang, 2016; Kang & Johnson, 2015; 2018). The current project further plans to augment the existing computer model to employ lexico-grammatical features along with suprasegmental/fluency measures to predict speaking proficiency scores in the Aptis speaking test.