Skip to main navigation Skip to search Skip to main content

EduX-RAG: Retrieval Augmented Generation Framework for Cross-Lingual Educational Chatbots

  • Johanne Medina*
  • , Soon Gyo-Jung
  • , Joni Salminen
  • , Kholoud Khalil Aldous
  • , Bernard J. Jansen
  • *Corresponding author for this work
  • Hamad bin Khalifa University
  • University of Vaasa

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Developing cross-lingual educational chatbots presents significant challenges when handling dynamic language switching and content retrieval across different languages. We introduce EduX-RAG, a framework that integrates Retrieval-Augmented Generation (RAG) with prompt engineering to enable cross-lingual conversations in educational chatbots. EduX-RAG addresses the complexities of managing multilingual content and cross-lingual queries within a single interaction. We validate the effectiveness of EduX-RAG using 13 languages, including high- and low-resource languages and commonly confused language pairs. Deployed in a fully functional educational chatbot, EduX-RAG demonstrates strong cross-lingual retrieval capabilities with a high answer rate of 92%, even when query and material languages differ. Overall, EduX-RAG delivers robust cross-lingual capabilities, achieving a Language Identification Accuracy of 90%, Material Citation Accuracy of 99%, and a Response Similarity Score of 70%.

Original languageEnglish
Title of host publication2025 3rd International Conference on Foundation and Large Language Models, FLLM 2025
EditorsKai Erenli, Christian Guetl, Yaser Jararweh, Jim Jansen
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages465-473
Number of pages9
ISBN (Electronic)9798331594091
DOIs
Publication statusPublished - 2025
Event2025 3rd International Conference on Foundation and Large Language Models, FLLM 2025 - Vienna, Austria
Duration: 25 Nov 202528 Nov 2025

Publication series

Name2025 3rd International Conference on Foundation and Large Language Models, FLLM 2025

Conference

Conference2025 3rd International Conference on Foundation and Large Language Models, FLLM 2025
Country/TerritoryAustria
CityVienna
Period25/11/2528/11/25

Keywords

  • chatbot
  • education
  • LLM
  • prompt-engineering
  • RAG

Fingerprint

Dive into the research topics of 'EduX-RAG: Retrieval Augmented Generation Framework for Cross-Lingual Educational Chatbots'. Together they form a unique fingerprint.

Cite this