DBIS

To infer the most out of data, the integration of ontologies and knowledge graphs has become essential for enhancing data interoperability and queryability. Ontologies are a formal representation of knowledge within a domain, structuring entities into classes with properties and relationships among them. With a knowledge graph, data can be visually represented in a graph, with the nodes representing the entities and the edges representing the relationships between them. Ontologies and knowledge graphs naturally work out in combination, as ontologies provide the entity structure which knowledge graphs can build upon.

For researchers, applying ontologies and knowledge graphs to their research data, and integrating this workflow with their everyday work commonly presents a significant challenge. Looking up or creating an ontology for a researcher’s specific context, as well as transforming their research data to conform to such an ontology, requires time and support. This thesis explores the use of large language models (LLMs) in the process of creating ontologies, transforming data into a format conforming to such an ontology, and writing such data into knowledge graphs. It aims to bridge the gap from heterogeneous, distinct, and unconnected data to structured semantic representations, leveraging the advanced natural language processing capabilities of LLMs to automate and optimize this transformation. The expected research output is an LLM-based tool supporting researchers on this pipeline.

Tasks:

• Conduct a comprehensive literature review on ontologies, knowledge graphs, and researchers’ needs with integration the practices to their RDM processes.
• Define the pipeline, design the architecture, define tool requirements and implement the LLM-based tool.
• Evaluate the tool for functionality, performance, and usability with real-world datasets and researchers.

If you are interested, please do not hesitate to contact: kim@dbis.rwth-aachen.de