To connect fragmented materials research data systematically and create value, an ontological approach is no longer a choice—it is a necessity. However, when faced with the vastness of the materials domain, knowing where and how to begin can be daunting. Many research teams ambitiously attempt to implement complex OWL (Web Ontology Language) from the outset. Yet, they often find themselves hitting a wall of extensive design work and rigid logical structures, leading to exhaustion and, eventually, abandonment of the project.
SKOS (Simple Knowledge Organization System) provides a clear roadmap for our first steps. By stripping away complexity and focusing on connecting the core links of knowledge, SKOS is the optimal tool to jumpstart ontology construction.
In this post, we introduce a strategy for building a simple knowledge skeleton using SKOS and expanding it into a robust, domain-level ontology.
Why Focus on Simplicity in an Ontological Approach?
The danger lies in the “trap of technical perfection.” If we become too obsessed with building a flawless system, we risk losing sight of our primary goal: the actual connection of knowledge. It is time to seriously consider the most simple yet powerful way to begin.
The Trap of Perfect Design: Why Do We Burn Out Before Starting?
The materials domain possesses an extremely complex information hierarchy. This is due to the intricate correlations between data—ranging from microstructures to macroscopic properties, complex manufacturing processes, and countless combinations of chemical compositions. Most attempts to resolve this complexity involve building a perfect domain ontology aligned with a Top-level Ontology. However, such attempts inevitably face significant hurdles:
- The Learning Curve Barrier: Investigating existing complex ontology standards and aligning them with specific research data can take anywhere from several months to over a year.
- Expert Disengagement: Domain experts in materials science—the very people who must provide the knowledge—often struggle with the abstract concepts and complex logical structures of OWL. Consequently, the ontology project often devolves into an “IT-only” task, disconnected from the actual research field.
Ultimately, the pursuit of a perfect start often leads to a standstill, where a project fails to even launch.
SKOS: Start Light, Iterate Fast
To rapidly build a domain-level ontology and apply it to actual research fields, a “start light, iterate fast” strategy is highly effective. The optimal tool for this is SKOS (Simple Knowledge Organization System).
SKOS is a lightweight data model designed to systematically share knowledge classification systems and thesauri over the web. By providing an intuitive structure centered on concepts and labels rather than complex logical design, it allows domain experts to immediately participate in the essential work of organizing core terminology and establishing hierarchical structures—all without needing deep expertise in ontology engineering. This simplicity and flexibility serve as the fastest starting point for systematizing vast materials knowledge and provide a solid foundation for evolving into a sophisticated domain ontology in the future.
In particular, instead of the complex logical restrictions or axioms found in OWL, SKOS offers the following intuitive structure:
- Concept: Defines flexible units of knowledge (e.g., specific materials, processes).
- Label: Standardizes terminology through Preferred Labels (
prefLabel) and Alternative Labels (altLabel). - Relation: Establishes Hierarchical (
broader/narrower) and Associative (related) links.
Thanks to this simplicity, materials domain experts can focus immediately on the core task of organizing knowledge based on their expertise, without being hindered by complex coding or deep ontological theories.
A Powerful Foundation Built on a Simple Schema
Starting with SKOS does not mean the data lacks depth. On the contrary, by first building a flexible yet solid foundational schema, you can more easily create a powerful base for evolving into a complex knowledge model later.
The greatest advantage is Rapid Prototyping. Design work that would traditionally take months can be shortened to just a few weeks, allowing the core skeleton of knowledge to be established quickly. During this process, researchers can prioritize digitizing the essential materials knowledge held by domain experts rather than getting bogged down in complex logical structures.
Furthermore, SKOS offers excellent Incremental Scalability. Concepts defined early on can be “promoted” to OWL Classes or assigned more detailed Properties when higher analytical precision is required. Ultimately, the systematic connection of materials data begins not with wasting time on a grand blueprint, but with the rapid execution of linking real-world terms through SKOS.
The Ideal Strategy: Starting with the PSPP Framework
Knowledge structuring in materials science is challenging due to the complex causality between data. To solve this, adopting the PSPP (Process-Structure-Property-Performance) framework—the fundamental principle of materials science—as the SKOS ‘Concept Scheme’ is the most ideal strategy for the following reasons:
Perfect Alignment with Domain Experts’ Mental Models
The practical success of ontology construction depends on the active participation of domain experts (materials researchers) who know the field best. In this regard, the PSPP framework is a powerful starting point because it is the “common language” and “framework of thought” used across both academia and industry.
Researchers find it natural and intuitive to categorize their data into the four major “buckets” of Process, Structure, Property, and Performance. This low barrier to entry allows them to contribute to a high-quality knowledge base simply by placing their specific alloys or process variables within the PSPP hierarchy using SKOS relations, even without specialized knowledge in ontology engineering.
Standardizing Interoperability: Starting with SKOS
Materials research data is a field where fragmentation is severe; identical concepts are often expressed in different terms depending on the laboratory environment or the researcher’s subjectivity. In this context, a SKOS framework with PSPP as the top-level schema serves as a powerful standardization reference point that unifies scattered knowledge.
Specifically, by utilizing the labeling functions of SKOS, various process names or property units used in the field can be systematically organized into Preferred Labels (prefLabel) and Alternative Labels (altLabel). This allows synonyms or abbreviations used differently by each researcher to converge into a single standardized concept, enabling the rapid construction of a domain-standard glossary optimized for each stage of the PSPP framework.
Furthermore, this standardization work plays a decisive role in securing external data interoperability beyond internal organization. When linking with authoritative external materials databases or public knowledge bases in the future, the presence of clear PSPP categories and a refined terminology system will drastically increase the accuracy of data mapping. Consequently, building SKOS based on PSPP increases data reusability and serves as a core foundation for transforming fragmented information into valuable knowledge assets.
Why SKOS is Highly Practical for the Materials Domain
The diversity of data formats and the complexity of terminology systems in materials science are areas where the flexible and intuitive characteristics of SKOS are most effective. Its utility is particularly high in the following three aspects:
Precise Terminology Standardization and Multilingual Management (Labeling): In materials science, even the same substance or composition is called by various names—common names, chemical formulas, IUPAC names, etc.—depending on the research site, posing a major challenge for data integration. By establishing standard names through prefLabel and grouping scattered synonyms and abbreviations under a single concept via altLabel, SKOS provides a robust standardization environment where the system recognizes different terms as the same material.
Rapid Construction of a Flexible Taxonomy for Vast Materials Knowledge: The classification system for materials—spanning metals, ceramics, and polymers—is incredibly broad and complex. Using the broader and narrower relations of SKOS, one can establish an intuitive materials taxonomy without the need for the rigid and complex logical property definitions required by OWL. This simplicity significantly reduces the design burden in the early stages of knowledge structuring and plays a critical role in quickly converting domain experts’ knowledge into a data skeleton.
Securing Global Data Interoperability: Since SKOS is an international standard schema established by the W3C (World Wide Web Consortium), knowledge models built by individual labs or institutions do not remain isolated but can easily connect with the outside world. For example, it is highly compatible for mapping and linking with established global materials ontologies like MatOnto or professional chemical dictionaries, serving as a standard bridge to share and expand knowledge within the global materials data ecosystem.
Next Steps: Transitioning to Full-Scale Ontology Construction
Once the basic skeleton and terminology system of knowledge have been established through SKOS, the next step is to expand into a Domain Ontology (OWL) equipped with logical sophistication to add depth to the knowledge. This enhancement process goes beyond simple classification to complete the “semantic connection” between data, involving the following core tasks:
First, the Class & Property Definition stage takes place. Elements that were loosely defined as “concepts” in SKOS are promoted to OWL Classes, and the relationships between them are defined as clear Object Properties. For example, a relationship between “Material A” and “Process B” that was previously marked as simply “related” is now defined with a specific property such as “Material A isManufacturedBy Process B.” This creates a clear knowledge structure that machines can explicitly understand and reason upon.
Second, ensure knowledge rigor by setting Logical Constraints and Axioms. This involves logically defining parameters such as the temperature ranges or pressure conditions under which a specific process can be performed, or the limits of chemical composition that a certain material can possess. These constraints serve as criteria to automatically validate the integrity of incoming materials data, playing a crucial role in significantly enhancing overall data quality.
Third is the Instance Mapping (Population) work. This step involves linking actual data generated in laboratories, numerical values extracted from research papers, or simulation results onto the elaborately designed ontology structure. Through this process, the ontology—once an abstract model—is transformed into a “living” materials knowledge graph, becoming a tangible data asset that can be utilized for search and analysis in real-world research settings.
Finally, maximize the value of knowledge through the Application of Reasoning Rules. By connecting a reasoning engine to the constructed ontology, you can analyze relationships between data to automatically discover new correlations not explicitly entered or identify logical contradictions within the data. For example, by inferring expected properties based on the relationship between specific process conditions and microstructures, the system can provide intelligent research support that offers new insights beyond simple data retrieval.
Conclusion: Start Small, Grow Big
The success of building a materials knowledge graph does not depend on the perfection of the initial design, but on how flexibly it can expand to meet a changing research environment. SKOS is the optimal tool that guarantees this flexibility while helping to establish the core knowledge skeleton as quickly as possible.
Rather than getting lost in grandiose blueprints, focus on rapid execution—linking the terms scattered across your lab one by one through SKOS. This process of systematically organizing field knowledge will serve as the strongest foundation for evolving into a complex domain ontology in the future. This small start will be the most practical first step toward an era of future-oriented AI materials research driven by data-led innovation.