Work

BPMN-Based Design of Multi-Agent Systems: Personalized Language Learning Workflow Automation with RAG-Enhanced Knowledge Access

Year: 2025

Type: article

Abstract: The intersection of Artificial Intelligence (AI) and education is revolutionizing learning and teaching in this digital era, with Generative AI and large language models (LLMs) heralding even greater possibilities for the future. This paper presents a novel approach to design a personalized language learning by combining Business Process Model and Notation (BPMN) with Multi-Agent System (MAS) architectures and Retrieval Augmented Generation (RAG) knowledge bases. Addressing the specific challenges of teaching Luxembourgish, a low-resource language, we design a modular system where BPMN diagrams play a central role in designing and orchestrating the workflows of intelligent agents. Each agent is responsible for a specific learning activity, such as reading, listening, grammar, or conversation, and is equipped with access to a dynamically retrieved, context-rich knowledge base powered by RAG. To ensure realism in learner interaction, we integrate speech-to-text and text-to-speech technologies, creating an immersive, human-like learning environment. The system simulates intelligent tutoring through agents’ collaboration and dynamic adaptation to learner progress. We demonstrate our method through a Luxembourgish learning platform that integrates GPT-based agents and educational content from textbooks of the National Institute of Languages of Luxembourg. Our Results demonstrate that BPMN functions not merely as a modeling tool but as a pivotal framework for designing intelligent and adaptive agent workflows in language learning systems. By structuring the sequence, logic, and interdependencies of diverse learning activities, BPMN enables the development of coordinated, goal-oriented agent behaviors. This systematic design approach ensures that each agent, whether dedicated to reading, listening, grammar, or conversation, operates within a coherent and pedagogically aligned flow, thereby supporting greater personalization, learner engagement, and instructional effectiveness.

Source: Information

DOI: 10.3390/info16090809