Masterarbeit / Bachelorarbeit
Knowledge Graph Engineering using Large Language Models
Research Area
Advisers
![samuel](/svnproxy/images/people/samuel.jpeg)
![gaedke](/svnproxy/images/people/gaedke.png)
Description
Knowledge graphs are powerful tools for organizing and
representing complex relationships between entities in a structured form. They are widely
used across various domains, from enhancing search engines to enabling complex data
analytics and supporting artificial intelligence applications. The construction and
maintenance of knowledge graphs, however, are labor-intensive processes that require
significant expertise and manual effort. Large Language Models (LLMs) offer a promising
solution to automate these processes, significantly enhancing the efficiency and accuracy
of knowledge graph engineering. LLMs, with their advanced capabilities in natural language
understanding and generation, can be leveraged to automatically extract, integrate, and
update information from vast amounts of unstructured data sources.
This thesis aims to harness the power of LLMs to develop a robust system for
knowledge graph engineering, addressing the key challenges of entity recognition,
relationship extraction, and knowledge graph population. This thesis aims to develop a
LLM-bases solution, designed to automate the engineering of knowledge graphs. The key
feature of the solution will include Entity Recognition and Linking, where LLMs are used
to accurately identify and link entities from unstructured text, ensuring that all
relevant entities are captured and correctly integrated into the knowledge graph. Wikidata
and DBpedia are used to the entity recognition and linking. The solution provides a way to
automatically extract relationships between entities using LLMs, capturing the complex
interactions and dependencies that exist within the data. The solution then aims to
seamlessly populate the knowledge graph with extracted entities and relationships,
ensuring the graph remains up-to-date with the latest information. The solution provides a
web-based UI that allows users to easily input data sources, review extracted entities and
relationships, and make necessary adjustments or annotations. The solution supports
seamless integration with existing knowledge management systems and data sources,
facilitating smooth data transfer and utilization. The system is designed to be easily
extendable to accommodate new data sources, entity types, and relationship types, ensuring
the knowledge graph can evolve with changing needs.
The objective
of this thesis is to analyze the current state of knowledge graph engineering methods,
identify existing challenges, and develop a comprehensive LLM-based solution that
addresses these needs. This includes designing and implementing the software tool,
followed by an experimental evaluation through a pilot study to demonstrate its
effectiveness and usability. By advancing the knowledge graph engineering through this
LLM-based system, this thesis aims to significantly improve the efficiency and accuracy of
knowledge graph construction and maintenance processes, thereby enhancing the ability of
organizations to harness the full potential of their data for informed decision-making and
advanced analytics.