Enhancing English language teaching quality evaluation via dynamic multimodal cognitive transfer models

This paper proposes a dynamic assessment method based on multimodal cognitive transfer modelling to address the limitations of static and unidimensional analysis in English teaching quality assessment. A two-channel LSTM-cognitive state space model is constructed by synchronously collecting four-dimensional data of speech, vision, text and physiological signals in the teaching scene, quantifying students' cognitive state transfer trajectories based on the ACT-R cognitive architecture in the knowledge transfer channel, and adopting a dynamic causal map to model the feedback mechanism of teachers' strategy adjustment in the teaching intervention channel. A time-varying weighted assessment function was designed to dynamically fuse cognitive state vectors with intervention intensity. In a 136-lesson experiment in 12 schools, the causal attribution rate of this method was improved by 41.2%, and the adoption rate of intervention suggestions reached 83.7%, which verified the effectiveness and universality for dynamic quality assessment of English teaching.