[Paper Review] The landscape of ontologies in materials science and engineering: A survey and evaluation

Ebrahim Norouzi, Jörg Waitelonis+1 more|arXiv (Cornell University)|Aug 12, 2024

Semantic Web and Ontologies4 citations

TL;DR

This paper presents a comprehensive survey and evaluation of 60 ontologies in materials science and engineering (MSE), assessing their quality, reuse, structural complexity, and interoperability using standardized metrics. It identifies key strengths and weaknesses, offering domain experts actionable criteria for selecting suitable ontologies to enhance data integration and knowledge sharing in MSE research.

ABSTRACT

Ontologies are widely used in materials science to describe experiments, processes, material properties, and experimental and computational workflows. Numerous online platforms are available for accessing and sharing ontologies in Materials Science and Engineering (MSE). Additionally, several surveys of these ontologies have been conducted. However, these studies often lack comprehensive analysis and quality control metrics. This paper provides an overview of ontologies used in Materials Science and Engineering to assist domain experts in selecting the most suitable ontology for a given purpose. Sixty selected ontologies are analyzed and compared based on the requirements outlined in this paper. Statistical data on ontology reuse and key metrics are also presented. The evaluation results provide valuable insights into the strengths and weaknesses of the investigated MSE ontologies. This enables domain experts to select suitable ontologies and to incorporate relevant terms from existing resources.

Motivation & Objective

• To address the lack of comprehensive, quality-controlled surveys of ontologies in materials science and engineering (MSE).
• To identify and evaluate 60 MSE ontologies based on structural, reuse, and interoperability metrics to support informed selection by domain experts.
• To provide a systematic framework for ontology selection by defining quality-control criteria, including reusability, completeness, and alignment with external standards.
• To highlight gaps in current ontology practices, such as poor documentation, limited use of competency questions, and low adoption of ontology design patterns.
• To support the FAIRification of materials data by enabling better alignment, reusability, and integration of semantic resources in MSE workflows.

Proposed method

• A multi-phase methodology was employed, combining expert surveys from 13 industry-led pilot projects within the Platform Material Digital (PMD) to identify domain-specific ontology requirements.
• The study evaluated 94 ontologies, with 60 selected for in-depth analysis based on relevance, domain coverage, and availability of metadata.
• A set of 12 quality-control metrics was applied, including root and leaf cardinality, number of classes, external class references, and modularity to assess structural complexity and interoperability.
• The evaluation included analysis of ontology reuse through citation and dependency tracking, as well as assessment of namespace alignment and external references to measure integration with external standards.
• The study leveraged existing repositories such as BioPortal, MatPortal, and IndustryPortal to collect and validate ontology metadata and accessibility.
• All evaluation results were published online at https://ise-fizkarlsruhe.github.io/mseo.github.io/ for transparent, reproducible access.

Experimental results

Research questions

• RQ1 Which ontologies in materials science and engineering are most suitable for specific research applications, and what criteria should guide their selection?
• RQ2 How do existing MSE ontologies perform in terms of structural complexity, reusability, and interoperability with external standards?
• RQ3 To what extent are ontology design patterns, competency questions, and documentation practices adopted across the surveyed ontologies?
• RQ4 What is the level of reuse and integration of MSE ontologies with external vocabularies and domain-specific frameworks?
• RQ5 How can quality-control metrics be systematically applied to improve the FAIRness and usability of materials science ontologies?

Key findings

• The study evaluated 94 ontologies, with 60 undergoing in-depth analysis, revealing significant variation in structural complexity, with EMMO Crystallography and EMMO Microstructure showing high root and leaf cardinality, indicating broad and detailed foundational structures.
• CHAMEO and EMMO Microstructure demonstrated strong interoperability with 125 and 118 external class references, respectively, indicating high integration with external ontologies.
• PODO and PLDO had low external references (11 and 12), suggesting limited alignment with external standards and reduced interoperability.
• Only a minority of ontologies adopted ontology design patterns or included competency questions, highlighting a critical gap in best practices for ontology engineering.
• The analysis revealed that many ontologies suffer from inadequate documentation and metadata, impeding discoverability and usability despite available reuse potential.
• The study found that ontologies such as CSO and CDCO have simpler, more focused structures with fewer root classes, suggesting narrower domain scopes despite detailed depth in specific areas.