Neurosymbolic AI and the Logic-Knowledge Graph Spectrum
KGC 2025
•
1h 23m
Kaushik Roy, University of South Carolina, Student
Neurosymbolic AI is a rapidly evolving and emerging field. In recent works, researchers have proposed several conceptualizations of neurosymbolic AI, yet the field still lacks a unified perspective on how different symbolic formalisms, ranging from traditional predicate logic to description logics and knowledge graph–based models, can effectively integrate with neural networks. This tutorial aims to clarify these options by comparing their relative merits in terms of interpretability, computational feasibility, and real-world scalability. We will show how purely logical systems (e.g., Prolog, Markov Logic), description logics (e.g., OWL-based reasoning), and knowledge graphs (e.g., RDF, property graphs) each contribute distinct advantages in neurosymbolic pipelines, especially in conjunction with large language models (LLMs). By systematically analyzing these approaches through case studies and code demonstrations, attendees will gain a deeper understanding of when, why, and how to integrate logic or KG-based techniques within modern AI applications.
Up Next in KGC 2025
-
Discover How the World's Best Knowled...
Discover how AI meets business knowledge at the “AI Knowledge, Rising Stars Night” on May 6, 2025, 1 PM, EST time.
Get inspired by what the best KnowHax teams have developed for three weeks in April this year, and how they supported several U.S. National Science Foundation's "Knowledge Networks"...
-
“Beyond Relational” Data Systems For ...
Relational database systems are the de facto systems to implement data-intensive applications. Yet, many classes of applications require different data models, query languages, and/or computational capabilities from their underlying data systems. This 90-minute tutorial will cover three classes o...
-
Building a Hybrid Knowledge Graph System
Giuseppe Futia, CSI - Piedmontese Consortium for Information Systems (Italy), Data Engineer
Knowledge Graphs (KGs) are essential for data integration and analytics, yet selecting the right technology stack remains challenging. While Label Property Graphs (LPG) enable the usage of powerful graph-...
