Beyond SPARQL: Towards Conversation-based Access to Knowledge Graphs in GRAPHIA

SPARQL has long been the standard interface to knowledge graphs—but its technical complexity limits access for many audiences. This webinar introduced Quagga, an LLM-based agent developed within GRAPHIA that enables conversational exploration of knowledge graphs using natural language. We presented how the agent works, including methods to automatically generate metadata from any knowledge graph to support text-to-SPARQL translation. The session also highlighted the Quagga benchmark, an ongoing collaborative crowdsourcing effort to build a knowledge graph question-answering dataset focused on the social sciences and humanities. A recording of this webinar is avaliable here.