DAGMAR: A Knowledge-driven Framework for Assessing Supply Chain Resilience

Modern supply chains are increasingly exposed to systemic risks arising from geopolitical tensions, resource scarcity, and cascading disruptions. At the same time, relevant information for assessing such risks is fragmented across heterogeneous data sources with varying levels of reliability, timeliness, and structure which represents a risk for the availability of critical raw materials and essential goods. This paper presents DAGMAR, an AI and knowledge-driven framework for assessing supply chain risks by integrating validated statistical data, expert knowledge, and dynamically acquired (crawled) unstructured information within a technological framework. DAGMAR combines ontologies and knowledge graphs with neuro-symbolic AI techniques to enable the systematic comparison of AI-generated statements against trusted, curated data sources. The applicability of the framework is demonstrated through a use case on the analysis of phosphorus supply chains for Austria.

Keywords: Supply Chain, Retrieval Augmented Generation, Artificial Intelligence, Large Language Models, Ontologies, Knowledge Management.