AcademicRAG: Knowledge Graph Enhanced Retrieval-Augmented Generation for Academic Resource Discovery

Speakers: Shuhua Chen (KTH) and Zhuchenyang Liu (KTH & Aalto). The project is a collaboration led by Elias Zea (KTH), Shiva Sander Tavallaey (ABB), and Fredrik Heintz (LiU).

Title: AcademicRAG: Knowledge Graph Enhanced Retrieval-Augmented Generation for Academic Resource Discovery.

Abstract: AcademicRAG is a novel framework that integrates knowledge graphs with large language models to enhance academic information retrieval and knowledge discovery. By embedding semantic relationships into a structured representation, AcademicRAG overcomes the limitations of traditional Retrieval-Augmented Generation (RAG) systems, offering improved query precision, contextual understanding, and reduced hallucinations. The system was evaluated across multiple metrics using the UltraDomain dataset, showing consistent improvements over state-of-the-art baselines. Two test cases were then considered as applications of AcademicRAG. One focused on academic course discovery in acoustics at KTH, offering personalized learning paths and curriculum optimization tools for students and faculty. The other focused on research literature assistance, enabling structured literature surveys by identifying key papers and extracting relevant insights. Together, these contributions demonstrate the versatility and impact of graph-based retrieval systems in academic contexts, supporting more intuitive, accurate, and efficient engagement with scholarly content.