NLP-Powered Curriculum Material Generation for Multilingual Maritime Technical Education

Abstract

Maritime education instructors spend 60-70% of professional time developing curriculum materials rather than teaching, with 85% of technical maritime resources available only in English creating comprehension barriers for Indonesian students whose English proficiency averages IELP 1.0-2.0 levels insufficient for technical content understanding. This research presents the design and validation of natural language processing systems trained on maritime education corpora capable of automatically generating lesson plans, assessment questions, case studies, and learning materials while adapting content complexity to diverse learner English proficiency levels. Employing design science research methodology with qualitative stakeholder evaluation, the study engaged maritime educators (n=16), instructional designers (n=8), and students (n=20) through structured interviews examining content quality, pedagogical effectiveness, and usability. Transformer-based NLP architecture incorporating BERT models fine-tuned on IMO Model Courses, STCW requirements, and maritime technical manuals achieved 78% content quality ratings from expert evaluators while reducing material development time 62% from 12 hours to 4.5 hours per lesson. Thematic analysis revealed strong support for AI-assisted content generation, identifying critical themes of instructor workload reduction, content accessibility enhancement, and curriculum updating acceleration. Pilot implementation with 450 students across Navigation and Engineering departments demonstrated equivalent learning outcomes compared to manually-developed materials while enabling rapid curriculum adaptation to regulatory changes, contributing validated NLP architectures and empirical evidence supporting intelligent content development in technical maritime education contexts addressing instructor time constraints and multilingual accessibility challenges.