Abstract Meaning Representation for Cross-Domain Knowledge Integration: A Semantic Framework for Explainable and Trustworthy AI

The explosive growth of data in medicine, embodied intelligent manipulation, wind-power generation, and surgical planning has created unprecedented opportunities for scientific discovery and intelligent decision support, while simultaneously exposing the limitations of traditional information management. Vast amounts of knowledge remain locked in unstructured or semi-structured texts, images, and sensor logs, where keyword search and shallow parsing fail to capture the causal and relational semantics required for reasoning. Abstract Meaning Representation (AMR) offers a principled solution by encoding the core meaning of natural language–who did what, to whom, when, where, why, and how–as a language-neutral, graph-based structure. This review introduces the conceptual foundations and evolution of AMR, outlines its distinctive advantages over surface-level representations, and examines its synergy with large language models for scalable, multilingual semantic parsing. Through case analyses in biomedicine, robotics, renewable energy, and surgical care, we demonstrate how AMR facilitates high-fidelity knowledge graphs, supports causal inference, and underpins explainable, cross-domain decision-support systems. We also discuss current limitations, including challenges in handling highly technical terminology and multimodal data, and propose future research directions to strengthen AMR’s role as a cornerstone of trustworthy artificial intelligence.