Title: Formalising contextual expert knowledge for causal discovery in linked knowledge graphs about transformation processes: application to processing of bio-composites for food packaging

Authors: Melanie Munch; Patrice Buche; Hélène Angellier-Coussy; Cristina Manfredotti; Pierre-Henri Wuillemin

Addresses: Bordeaux University, Nouvelle-Aquitaine, France ' INRAE IATE, Montpellier University, Montpellier, France ' INRAE IATE, Montpellier University, Montpellier, France ' AgroParisTech, UMR MIA-Paris, Paris-Saclay, Paris, France ' Paris-Sorbonne Universities, Paris, France

Abstract: With numerous parameters and criteria to take into account, transformation processes are a challenge to model and reason about. This work can be eased thanks to knowledge graphs, which are a widespread practice for formalising knowledge associated with structured and specialised vocabulary about a given domain. They allow to draw semantic relations between concepts, and thus offer numerous tools for reasoning over complex queries. Yet, some of these queries in transformation processes might rely on an additional layer hard to transcribe: uncertainty. In this article, we present how knowledge graphs and probabilistic models can benefit each other for reasoning over transformation processes and address the necessity of formalising contextual expert knowledge for this combination. We then show how this can be used for (1) reverse engineering approaches and (2) linking knowledge bases, through a detailed example on the process of bio-composites for food packaging.

Keywords: knowledge graph; probabilistic model; expert knowledge; causality; linked open data.

DOI: 10.1504/IJMSO.2022.131129

International Journal of Metadata, Semantics and Ontologies, 2022 Vol.16 No.1, pp.1 - 15

Received: 28 May 2022
Accepted: 12 Sep 2022
Published online: 31 May 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article