Web-Based Multimodal Dynamic Assessment System for International Business English Proficiency Optimization

Abstract

With the advancement of economic globalization and web-based educational technologies, traditional static assessments fail to meet practical demands of international business English teaching and learning due to scenario disconnection and lack of real-time, personalized feedback. This study develops an artificial intelligence-driven, web-based dynamic assessment system with a cloud-edge collaborative, four-layer architecture. It collects multimodal data from simulated business scenarios, uses domain-adaptive Transformer models for language analysis, and applies reinforcement learning to adjust assessment parameters dynamically. Experiments show that system scores correlate highly with expert ratings (r = 0.91), with an average module accuracy of 91.8%. The system effectively tracks learners' ability development and narrows individual differences. The findings innovate business English assessment from summative evaluation to formative diagnosis, supporting the optimization of web-based business English learning and teaching.