editor@bjddjournal.org +91 8600071634, +91 98903 82132
Open-Access, Transparent Peer-Reviewed, Refereed

BODHIVRUKSHA JOURNAL OF DIVERSE DISCIPLINE

ISSN: 3139-1486 (Online) | ISSN: 3139-9819 (Print) Impact Factor : 6.21 ESTD Year : 2025
Bi-monthly Multilingual Academic Research Multidisciplinary

*** Important Announcement: Call for Papers! Submissions are now open for the upcoming issue. Please visit the "Submit Paper" section to learn more. ***
← Back to Special Issue on Emerging Frontiers in Artificial Intelligence and Multidisciplinary Intelligent Systems

AI-POWERED VIRTUAL TEACHING ASSISTANTS AND CHATBOTS: A Survey and a Retrieval-Augmented, Personalized, Multilingual Framework

Department of Artificial Intelligence and Data Science, Ajeenkya DY Patil School of Engineering, Pune, India · India

Abstract

The state-of-the-art AI-driven VTAs and educational chatbots are advancing from rule-based FAQ bots to LLM tutors with explanation capability, feedback and adaptability. Nevertheless, there is a range of issues that persists according to the literature: hallucination, generic answers to non-trivial questions, lack of good multilingual support, weak LMS integration, privacy concerns, and a poor track record in the long run. This paper (i) conducts a structured review of twelve recent surveys and empirical works on educational chatbots, ITS and ChatGPT in education; (ii) develops a taxonomy of VTA generations and a unified research-gap analysis; and (iii) introduces EduAssist, a five-layered VTA architecture with retrieval augmented generation (RAG) on course content, lightweight learner profile, multilingual question understanding, guardrails for academic honesty, and LMS/privacy ser-vices. The reproducible evaluation procedure and design validation results on the synthetic bench-mark of 600 questions and 200 simulated learners are described. In the simulation, the proposed approach increases grounded answers’ accuracy from 71.4% (base LLM only) to 90.2%, lowers un-supported statements’ rate from 18.6% to 4.1%, and deflects 63% of trivial questions from faculty. These figures are synthetic and illustrative; they inspire, but do not substitute, classroom experiments.

Keywords

Virtual teaching assistanteducational chatbotgenerative AIlarge language modelsretrieval-augmented generationintelligent tutoring systemspersonalized learningAI in education.

How to Cite This Article

Harsh Waghchoure, Prof. Swarupa Kamble. (2026). AI-POWERED VIRTUAL TEACHING ASSISTANTS AND CHATBOTS: A Survey and a Retrieval-Augmented, Personalized, Multilingual Framework. Special Issue on Emerging Frontiers in Artificial Intelligence and Multidisciplinary Intelligent Systems.

Rights

Copyright
BODHIVRUKSHA JOURNAL OF DIVERSE DISCIPLINE