Scholarlink

This article describes the design and validation of a hybrid chatbot aimed at student support in higher education institutions. The proposed system integrates a Dynamic Decision Tree (DDT) and an Adaptive Bayesian Model (ABM), enabling the management of frequent, ambiguous, or unstructured queries through a process of semantic classification and incremental learning. The development was based on an institutional corpus and a set of academic jargon, incorporating natural language processing techniques for lexical expansion and intent detection. A Module for Argumentation and Negotiation (MAN) was implemented, employing Toulmin’s model and multi-turn negotiation schemes to generate justified responses and resolve conflicting requests. In addition, a blockchain-based traceability system (Hyperledger Fabric) was integrated to anchor sensitive interactions, ensuring data integrity and privacy. Complementarily, a light gamification approach was applied to encourage proper query formulation and the completion of administrative procedures. The tests demonstrated a Top-1 accuracy of 92% in the DDT and a 100% resolution rate in the ABM, confirming the feasibility of the proposed approach to enhance institutional efficiency and user experience in automated support environments.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Design of a Hybrid Model Based on Dynamic Decision Trees and Adaptive Bayesian Algorithms for Student Support and the Integration of Emerging Technologies

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Samuel Efren Viñas Alvarez,
  • María Esmeralda Arreola Marín,
  • José Iraic Alcantar Alcantar,
  • Mariela Chávez Marcial,
  • Edgar Gonzalo Cossio Franco

摘要

This article describes the design and validation of a hybrid chatbot aimed at student support in higher education institutions. The proposed system integrates a Dynamic Decision Tree (DDT) and an Adaptive Bayesian Model (ABM), enabling the management of frequent, ambiguous, or unstructured queries through a process of semantic classification and incremental learning. The development was based on an institutional corpus and a set of academic jargon, incorporating natural language processing techniques for lexical expansion and intent detection. A Module for Argumentation and Negotiation (MAN) was implemented, employing Toulmin’s model and multi-turn negotiation schemes to generate justified responses and resolve conflicting requests. In addition, a blockchain-based traceability system (Hyperledger Fabric) was integrated to anchor sensitive interactions, ensuring data integrity and privacy. Complementarily, a light gamification approach was applied to encourage proper query formulation and the completion of administrative procedures. The tests demonstrated a Top-1 accuracy of 92% in the DDT and a 100% resolution rate in the ABM, confirming the feasibility of the proposed approach to enhance institutional efficiency and user experience in automated support environments.