A Knowledge-Augmented Generation Framework for Question Answering Systems: Formalization, Architecture, and Evaluation
Abstract
This thesis presents the formalization, design, and evaluation of a Question Answering (QA) system based on the Knowledge-Augmented Generation (KAG) paradigm for the domain of police reports. The work addresses the need for reliable and explainable access to complex document collections characterized by sensitive information, rich relational structures, and strict accuracy requirements. The proposed framework integrates hybrid retrieval, graph-based evidence extraction, and answer generation grounded on retrieved subgraphs. A key contribution is the design of a synthetic data generation procedure that produces realistic police reports from a limited set of original documents, preserving privacy while supporting both knowledge base construction and benchmark creation. The system is evaluated through controlled experiments on answer quality, grounding, and explainability, and is compared with a baseline RAG pipeline. The results show that the KAG-based approach improves the reliability and traceability of generated answers, providing a generalizable architecture for knowledge-grounded question answering in complex information-intensive domains.
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.