Semantic Representation of Processes with Ontology Design Patterns
Abstract
The representation of workflows and processes is essential in materials science and engineering, where reproducibility depends on structured and semantically coherent process models. Although many ontologies exist for process modeling, they often remain complex and difficult to reuse. This work identifies and evaluates reusable Ontology Design Patterns (ODPs) embedded in existing ontologies relevant to scientific workflows and materials engineering. It defines three domain-specific requirements—process structure, data & resources, and project roles—and proposes a baseline method for automatic ODP extraction using semantic similarity and modular ontology techniques. All extracted ODPs and the extraction workflow are openly available at https://github.com/ISE-FIZKarlsruhe/odps4mse.
Full text
Semantic Representation of Processes with Ontology Design Patterns Ebrahim Norouzi, Sven Hertling, Jörg Waitelonis and Harald Sack FIZ Karlsruhe & KIT SeMatS 2025 @ ISWC, Nov 2nd, Nara, Japan The 2nd International Workshop on Semantic Materials Science
2 Agenda ⬥Motivation & Introduction ⬥ODP Extraction Approaches ⬥Evaluation & Results ⬥Limitations & Future Work
Motivation ⬥Existing MSE ontologies are large, heterogeneous, and domain-specific. ⬥Difficult to reuse, integrate, or extend across projects and disciplines. ⬥Lack of modular, reusable Ontologies. ⬥Repeated effort in designing similar ontology structures. ⬥High chance of modeling inconsistencies and redundant work. 3 Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan
Motivation – Why Ontology Design Patterns ⬥Provide ready-to-use, validated modeling solutions. ⬥Avoid common modeling mistakes. ⬥Reduce development effort and accelerate ontology design. ⬥Enable interoperability and semantic alignment across MSE workflows. 4 Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan
Ontology Design Patterns ⬥Reusable, modular solutions for recurring ontology modeling problems [1]. ⬥A pattern must represent a recurring and reusable structure across multiple ontologies. ⬥It should provide a balance between abstraction (broad applicability) and specificity (domain-specific applicability). Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 5 [1] Gangemi, A., & Presutti, V. (2009). Ontology design patterns. In Handbook on ontologies (pp. 221-243). Berlin, Heidelberg: Springer Berlin Heidelberg.
Ontologies Related to Process Modeling ⬥Foundations: BFO, PlansLite (DOLCE) ⬥General purpose: EXPO, EXACT2, SMART Protocols ⬦lacks the specificity required to model the technical details of MSE processes ⬥Engineering: OntoCAPE, PKO, BBO ⬦often focus on industrial or chemical workflows ⬥Workflow/MSE: PMDcore, GPO, WILD, P‑PLAN, M4I Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 6
“ Problem Statement Often existing ontologies in MSE are large, heterogeneous and difficult to reuse 7 Identify reusable Ontology Design Patterns (ODPs) for process modeling in Materials Science & Engineering (MSE)
Methodology: Expert Reviewed Pattern Extraction Workflow ⬥Survey & select relevant ontologies (materials, workflows, provenance). ⬥Extract coherent modules matching MSE requirements. ⬥Validate with domain experts (NFDI-MatWerk LOD-WG). Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 8
Requirements Definition ⬥Req 1 – Process Structure: Steps / subprocesses / execution order. ⬥Req 2 – Data & Resources: Inputs / outputs / parameters (temperature, pressure, atmosphere). ⬥Req 3 – Project Goals & Roles: Represent project stages, milestones, agents, roles. ⬥Expert Validation: 9 experts (NFDI-MatWerk LOD-WG). Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 9
Extracted patterns: WILD Ontology Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 16
Extracted patterns: PMDco Ontology ⬥Extracted ODPs: Process + Resource + Project ⬥Findings: ⬦Process ODP: captures interlinked nodes in process networks. ⬦Resource ODP: clear definition of objects and value objects as data carriers. ⬦Project ODP: connects processes to project identifiers → lifecycle traceability. Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 17
Extracted patterns: PMDco Ontology Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 18
Evaluation Results Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 19
Evaluation Results: Key Findings ⬥P-PLAN → highest F₁ (Process 0.59 | Resource 0.53) → most precise retrieval. ⬥PMDcore → only ontology covering all three ODPs (Process 0.36 | Resource 0.21 | Project 0.25). ⬥M4I → moderate for Resource/Project (≈ 0.26 & 0.19); weak Process (0.07). ⬥GPO → low scores due to sparse annotation (≈ 0.12 & 0.07). ⬥WILD → 0.00 → insufficient metadata for similarity matching). ⬥OPMW → solid Process (0.55) but weaker Resource (0.24). Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 20
Example: Ground Truth vs Extracted Pattern (P-PLAN) for Requirement 1 ⬥Correct relations between process components correctly identified. ⬥Key class IRIs not retrieved and not present in the extracted module. ⬥ROBOT-extracted pattern remains semantically consistent. Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 21 Ground Truth Terms Activity, Plan, Step, correspondsToStep, hasInputVar, hasOutputVar, isPreceededBy, isStepOfPlan Automatically Extracted Terms isStepOfPlan, correspondsToStep, isPreceededBy, hasOutputVar, hasInputVar, isOutputVarOf, isInputVarOf, isVariableOfPlan, correspondsToVariable
Example: Ground Truth Pattern (P-PLAN) for Requirement 1 Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 22
Example: Extracted Pattern (P-PLAN) for Requirement 1 Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 23
Accessing the ODP Library & Resources ⬥All identified Ontology Design Patterns (ODPs) are available online. ⬥You can browse each ontology, view extracted Process, Resource, and Project ODPs, and download them for reuse or extension. ⬥GitHub Repository (implementation + workflow) https://github.com/ISE-FIZKarlsruhe/odps4mse ⬥Online ODP Library (interactive access) Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 24
Limitations ⬥Annotation Bias ⬦Evaluation favors textual metadata quality → bias toward well-documented ontologies (e.g., P-PLAN, PMDcore). ⬥Requirement Formulation Sensitivity ⬦Retrieval quality depends on how requirements are phrased linguistically. ⬥Ontology Independence ⬦Workflow treats each ontology in isolation → shared or cross-ontology ODPs remain undetected. Ebrahim Norouzi et al., Semantic Representation of Processes with ODPs, SeMatS 2025 @ ISWC, Nov 2, Nara, Japan 25