AI researcher Shuai Wang shows how computers can better understand and clean up huge webs of connected information.
His thesis focuses on a specific knowledge representation format known as knowledge graphs, where nodes represent entities and edges denote relations. Integrating knowledge graphs can result in richer resources but also lead to undesirable structures and even logical inconsistencies. Therefore, refinement methods that detect and correct such issues are essential. Scale matters. Problems that are easy for small knowledge graphs can become significantly more challenging at scale. Addressing these challenges requires data analysis, algorithm development, and rigorous evaluation. Wangs thesis investigates key issues in large, integrated knowledge graphs—such as identity, error sources, and knowledge evolution. Algorithms developed for analysis and refinement take advantage of, for example, graph theory and automated reasoning.
Improving LGBTQ+ knowledge connections
The study shows how computers can better understand and clean up huge webs of connected information, such as the ones that power search engines, digital libraries, or data systems in science and finance. By combining logical reasoning (so the computer can “think through” connections) with network analysis (so it can spot patterns and errors), the work makes these data networks more accurate and useful. Although working with such massive amounts of information is still difficult, the methods have already proven effective in real cases, such as improving LGBTQ+ knowledge collections. The tools and most datasets created are freely available so that others can build on them in future projects.
Improving data quality
The research improves how large, interconnected knowledge graphs (KGs) are built, refined, and maintained. By combining graph-theoretical and automated reasoning techniques, it enables the detection and correction of errors, outdated links, and concept drift at scale. For example, the proposed semi-automatic methods could help curators identify outdated LGBTQ+ terms. The results can contribute to improving data quality and data provenance, and sustainability in open knowledge systems—paving the way for trustworthy, evolving semantic infrastructures across disciplines.
Wang and his colleagues studied large integrated knowledge graphs. They took a data-centric approach. Different methods were used in this research. For example, graph theoretical methods were used for the detection of large connected components to get ready for refinement. Automated reasoning tools were used for resolving large, nested cycles. No simulation, field work, or laboratory experiments have been used.
More information on the thesis
