[3O1-GS-10v] AI application

10:15 AM - 10:30 AM JST(1:15 AM - 1:30 AM UTC)

[3O1-GS-10v-06] Semantic Twin: A Heterogeneous Multilayer Network Model for Enterprise Data IntelligenceA Case Study in Supply Chain Management

〇LOYS BELLEGUIE¹, Teruaki Hayashi², Takahiro Sanada¹ (1. Quollio Technologies, 2. School of Engineering, the University of Tokyo)

Keywords:

Semantic Twin,Multilayer Network,Knowledge Graph,Epistemic Intelligence

Enterprises accumulate vast siloed data, yet AI agents cannot reliably distinguish facts from data signals from inferences. Without explicit provenance and confidence, data remains information rather than justified knowledge—limiting AI trustworthiness.This work introduces the Semantic Twin—a semantic projection of enterprise reality elevating data intelligence to an epistemic level. It establishes a Single Source of Truth (SSoT) layer making explicit what counts as knowledge, its origin, and confidence. Building on metadata catalogs, it functions as a semantic control plane for AI reasoning.We formalize this as a Heterogeneous Multilayer Network: Schema layer (metadata), Flow layer (instances, lineage, data signals), and Business layer (entities, unstructured knowledge). Layers connect via identity links and a Semantic Bridge inferring associations through lexical, semantic, and topological methods—enabling cross-domain reasoning impossible in single-layer systems.We demonstrate through supplier disruption propagation in supply chain management, revealing how risks traverse semantic pathways and expose vulnerabilities invisible when dimensions are analyzed separately.