Chatbot Personalisation for EFL Learning:‬ ‭ Integrating BKT and Task Context‬

Abstract

English proficiency is critical in today’s globalised world, yet many EFL learners face challenges in traditional classrooms due to large class sizes, rigid pedagogy, and varying proficiency levels. While AI chatbots offer accessible, low-pressure language practice, their effectiveness is limited by a lack of personalisation. This paper proposes a novel framework for dynamically personalising EFL chatbot interactions by integrating Bayesian Knowledge Tracing with semantic task analysis. The system aggregates learner data across multiple task-specific chatbots (e.g., translation, writing) to model vocabulary and grammar mastery. By analysing task context, the framework generates recommendations tailored to each learner, ensuring both proficiency-appropriate and activity-relevant support. This work advances personalised language learning by bridging the gap between isolated chatbots and a unified, adaptive learning ecosystem.