Knowledge Graph Construction and Applications
We build knowledge graphs from complex relational data in domains including molecules, finance, and documents, and use graph neural networks and language models for prediction, inference, fact checking, and discovery.
Knowledge Graph Construction and Applications
Members: 윤예진, 전예지, 이수형
Methods: Heterogeneous Graph Embedding, LLM, Agentic AI
We construct knowledge graphs from complex relational data in domains including molecules, finance, and documents, and use graph neural networks and language models for prediction, knowledge inference, fact checking, and discovery. Building knowledge graphs is a challenging research area: extracting entities and relationships from existing source databases and determining their inferred hierarchies is difficult because they can have different structures in a wide range of contexts. Concept hierarchies can be organized and interpreted differently depending on the perspective taken. We seek to use AI to build systems and structures of knowledge that domain experts have developed over many years, apply them to a range of fields, and challenge the limits of existing research. We explore opportunities and develop techniques ranging from the construction of stock graphs and fact verification to dialogue systems for specific tasks.
Funding: Ministry of Environment
Leader: JaeHwan Jung