Methods for assessing foreign language proficiency

Abstract

The invention relates to a method for assessing foreign language proficiency using multimodal deep learning models and large-scale language models to automatically evaluate speaking and writing skills in large-scale standardized foreign language proficiency tests. Through a strategy of refining the model according to specific task-oriented criteria of the foreign language proficiency framework, the method described in the invention aims to standardize the testing process, minimize human assessment bias, and provide quantitative explanations for the scoring results.