University of Twente Student Theses

Login

KNOCK: Knowledge-driven Ontology and Chain-of-Thought-based Knowledge Graph Question Answering

Huang, B. (2025) KNOCK: Knowledge-driven Ontology and Chain-of-Thought-based Knowledge Graph Question Answering.

[img] PDF
4MB
Abstract:Knowledge graph question answering(KGQA) has garnered significant research interests and has evolved quickly in both academia and industry. It combines the strengths of artificial intelligence and linked data to connect humans with heterogeneous knowledge-intensive data sources in an interactive way. However, the current approaches of KGQA are mostly difficult to adapt to knowledge graph with different schemas universally. A lot of approaches also rely on substantial training to build the question answering model, which is difficult to maintain as the data sources iteratively update. In this thesis, we propose KNOCK, which stands for KNowledge-driven Ontology and Chain-of-Thought-based KGQA, an approach utilizing the synergy strengths of ontology embedding and Chain-of-Thought(CoT) prompting to construct a Retrieval-augmented Generation(RAG) system to enhance the performance of question answering on knowledge graphs. Via a bibliometric survey, we define our methodological framework and implement our prototype modeling in a CRISP-DM process. We ground our research in the question answering system of Kadaster Knowledge Graph(KKG), a knowledge database of the Dutch national cadastral and mapping agency. By analyzing the current question answering system of KKG, we identify its limitations and implement our approach on KKG to verify its feasibility. By conducting this research project, we aim to validate our approach and answer how we can establish a KGQA system more efficiently with less training cost and ensure the system is more resilient to knowledge source update. This research also makes contributions to the KGQA application in the domain of cadastral data management and geographical information system(GIS).
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science, 70 social sciences in general, 85 business administration, organizational science
Programme:Business Information Technology MSc (60025)
Link to this item:https://purl.utwente.nl/essays/106058
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page