Language as an Interface: Using LLMs for Knowledge Graph Exploration in the Social Sciences and Humanities

Knowledge graphs help improve LLM reasoning and factual grounding by providing structured, validated, and domain-specific knowledge. Conversely, LLMs also offer new opportunities for engaging with KGs, particularly in the area of exploration and question answering. They can process and generate natural language, making them promising interfaces for complex, structured datasets. Traditionally, user interaction with KGs has been mediated through graphical user interfaces (GUIs), which are typically generic, or through formal query languages like SPARQL, which require specialised knowledge and are inaccessible to many non-technical users. This has limited the broader adoption and utility of KGs, particularly in interdisciplinary contexts such as the social sciences and humanities (SSH).