Home  | Publications | HDL+24a

Critical Path Identification in Supply Chain Knowledge Graphs With Large Language Models

MCML Authors

Link to Profile Volker Tresp

Volker Tresp

Prof. Dr.

Principal Investigator

Database Systems, Data Mining and AI

Abstract

In the ever-evolving landscape of global commerce, supply chain management (SCM) has gained increasing significance. An important task in SCM is to find critical supply chain paths for a target company because these paths often represent potential bottlenecks in supply networks and thus could be crucial to risk management. The mainstream solution to this task requires supply chain managers to manually review supply chain data to uncover critical paths, resulting in considerable human labor costs. To better study SCM, recent efforts have been made to construct supply chain knowledge graphs (KGs) that connect supply chain-related data from different sources, facilitating the identification of critical paths through KG reasoning. In this paper, we develop an automated approach for critical path identification (CPI) based on supply chain KGs. We encode supply chain KGs into text and use large language models (LLMs) for CPI. LLMs can not only analyze the topological KG information but also leverage their world knowledge for better path identification. We experiment with two popular LLMs, i.e., GPT-3.5 and GPT-4, and find that they are able to do CPI and meanwhile generate reasonable explanations.

inproceedings

ESWC 2024

Extended Semantic Web Conference. Hersonissos, Crete, Greece, May 26-30, 2024.

Authors

Y. Han • Z. Ding • Y. Liu • B. He • V. Tresp

Links

DOI

Research Area

 A3 | Computational Models

BibTeXKey: HDL+24a

Back to Top