The construction and practice of a hybrid learning evaluation system of humanistic value in business Japanese learning under the background of big data

With the development of big data technologies and the increasing adoption of blended learning in higher education, systematic and data-informed evaluation frameworks have become increasingly important. The rapid growth of business Japanese education in China further highlights the need for evaluation approaches that can account for both linguistic accuracy and humanistic values in professional communication. This study proposes a hybrid learning evaluation framework that integrates humanistic value considerations and develops a core linguistic analysis component based on an optimized Nivre dependency parsing algorithm. Rather than claiming a fully implemented evaluation system, the framework conceptualizes blended learning as the object of evaluation and aims to support teachers by providing structured insights into learners’ language use in business contexts. Drawing on learning theory, the study also discusses potential instructional implications for business Japanese learning from the perspectives of institutions, teachers, and students. Experimental results show that the optimized Nivre-based model achieves over 92% accuracy across multiple test sets in dependency analysis and performs effectively in detecting culturally relevant linguistic features such as honorific usage. It should be emphasized that this study primarily proposes a framework and validates the feasibility of its core computational module. The reported performance metrics reflect the module ‘s efficacy on specific linguistic analysis tasks, not an empirical evaluation of a complete pedagogical system or learning outcomes. The reported performance metrics reflect the technical capability of the core computational module and do not constitute empirical evidence of educational or instructional effectiveness.