Skip to content

Overview

Author: Harshdeep Singh (Odoma)

This techncal note describes Quagga Agent, an LLM-powered conversational agent that enables researchers to explore musicological knowledge in the Culture Knowledge Graph (CKG) and beyond—without requiring any SPARQL expertise. The agent accepts natural-language research questions, autonomously identifies the relevant knowledge graph, retrieves and understands its schema, generates and executes SPARQL queries, and returns verified, structured results with full scientific transparency.

Quagga Agent was originally developed within the EU-funded GRAPHIA project (grant 101188018) as a general-purpose SPARQL agent for multiple knowledge graphs in the social sciences and humanities. For this challenge, we have indexed the Culture Knowledge Graph (CKG) and other music-related KGs in order to enable the exploration of a wide range of musicological research questions.

Contents

  1. Problem
  2. Architecture
  3. Approach
  4. Musicology
  5. User experience
  6. Transparency
  7. Limitations