In this paper, we report some preliminary experiments on automated scoring of non-native English speech and the prompt spe-cific nature of the constructed models. We use ICNALE, a publicly available corpus of non-native speech, as well as a vari-ety of non-proprietary speech and natural language processing (NLP) tools. Our re-sults show that while the best performing model achieves an accuracy of 73% for a 4-way classification task, this performance does not transfer to a cross-prompt evalu-ation scenario. Our feature selection ex-periments show that most predictive fea-tures are related to the vocabulary aspects-of speaking proficiency.
展开▼
