14 February 2025, by Janis-Marie Paul

Photo: Janis-Marie Paul

We are excited to announce that Longquan Jiang, Junbo Huang, and Cedric Möller have been awarded the Best Student Paper for their research "Ontology-Guided Hybrid Prompt Learning for Generalization in Knowledge Graph Question Answering" at the 19th IEEE International Conference on Semantic Computing (ICSC 2025).

About IEEE ICSC

The IEEE International Conference on Semantic Computing (ICSC) focuses on the extraction, representation, integration, and application of semantics in various domains, including data, documents, processes, and AI systems. It brings together experts working on semantic analytics, knowledge integration, and intelligent interfaces, promoting interdisciplinary research at the intersection of AI, knowledge graphs, and computing.

Research Highlights: OntoSCPrompt

Their award-winning paper introduces OntoSCPrompt, a Large Language Model (LLM)-based approach for Knowledge Graph Question Answering (KGQA). Most existing KGQA systems are tailored to specific Knowledge Graphs (KGs) like Wikidata, DBpedia, or Freebase and struggle to generalize to unseen KGs due to their varying schemas and structures.

Key Contributions of OntoSCPrompt:

• Two-stage architecture: Separates semantic parsing from KG-dependent interactions
• Ontology-guided hybrid prompt learning: Incorporates KG ontologies into the prompt-learning process
• Efficient generalization: Achieves state-of-the-art (SOTA) performance without additional training on new KGs
• Domain adaptability: Successfully generalizes to unseen domain-specific KGs like DBLP-QuAD and CoyPu KG

By generating SPARQL query structures first and filling them with KG-specific data afterward, OntoSCPrompt minimizes the need for resource-intensive retraining, making it a scalable and flexible solution for KGQA.

Read the full paper here