📌 Why is Materials Ontology essential for innovation in advanced materials development? In this post, we explore how to overcome data silos, teach the “language of materials” to AI, and build the foundation for intelligent knowledge graphs and interoperability.
With the rise of Materials AI and data-driven research, researchers worldwide are demanding vast amounts of data to drastically shorten the development cycle of new materials. While immense efforts are being made to produce and accumulate this data, we face a paradoxical reality: we are suffering from a “drought of usable knowledge amidst a flood of data.”
Why is this happening? The answer is simple: the data cannot talk to each other. In this post, we will examine the basics of Materials Ontology, the first step in granting intelligence to materials data, and discuss why it is absolutely necessary.
Beyond Standardization: Towards Ontology
To a non-expert, “data standardization” might simply mean aligning column names in an Excel sheet or unifying units (e.g., Celsius vs. Fahrenheit). While this is a fundamental mission and the starting point for data integration, authentic data-driven materials research requires a much more sophisticated and complex process.
Currently, standardization in the materials field focuses primarily on the “unification of terminology.” Because the types of elements, processes, and properties are so vast and complex, standardizing terms is recognized as a core task for building a data ecosystem.
In practice, even experts in the same field use different names for the same data points. For example, regarding the temperature in a heat treatment process, one researcher might record it as ‘Annealing Temp,’ another as ‘Heat Treatment Temp,’ and a third simply as ‘Temp.’ Consolidating these fragmented terms into a single “standardized terminology” is the most urgent task for data integration in the field.
However, to enable AI to interpret and reason with data independently, we must move beyond standardizing the “shape” of data and enter the stage of explaining its “meaning” and “context.”
- Terminology-centered Standardization: “Unify ‘Temp’ from Equipment A and ‘Temperature’ from Equipment B into a single term, and standardize the unit to ‘℃’.”
- The Need for Relational Definition in Ontology: “‘Temperature’ is an ‘attribute’ of the ‘Heat Treatment Process,’ which induces a ‘phase transition’ in the crystal structure, ultimately creating a causal link that changes the ‘strength (property)’ of the material.”
The word Ontology originates from the branch of philosophy that studies the nature of existence. When applied to modern information science, it refers to a “blueprint that defines the relationships between objects and concepts in a way that computers can understand through a logical system.”
Among various data-intensive fields like biotechnology, the adoption of ontology in materials science is being deeply considered. It goes beyond mere terminology to establish Semantic Links. While traditional standardization asks, “What should we record?”, ontology teaches the computer—the AI—the answer to: “How is this data contextually connected to other data?”
 💡 Technical Terminology Guide
- Semantic Link: Assigning ‘logical meaning’ to the relationship between data points. Instead of just listing ‘Lithium’ and ‘Battery,’ it defines the hierarchy and roles: “Lithium (Material) is a key element (Component) that constitutes the Cathode (Part) of a Battery (Product).”
Why Materials Ontology Now? – The Limits of Data Silos
Many companies and research institutes currently face a massive barrier known as Data Silos. A “silo” originally refers to a tall, vertical tower used to store grain. While the grain inside is plentiful, it is completely isolated from the neighboring tower. Materials research sites are no different; data is often isolated by department, laboratory, or even individual equipment, piling up without ever being interconnected.
This phenomenon goes beyond making data hard to find—it causes a severe lack of Interoperability. Interoperability is the state where different systems can not only exchange data but also accurately understand and utilize its meaning immediately, without the need for additional processing.
 💡 Technical Terminology Guide
- Interoperability: A state where different systems from various research institutes or companies can not only exchange data but also accurately understand and utilize its meaning immediately without additional interpretation or processing.
Consider a real-world research scenario: Lab A precisely recorded the electrode coating process variables (slurry viscosity, coating speed, etc.), and Lab B collected long-term battery life data using those same electrodes. A researcher wants to integrate these two datasets to analyze how “process conditions affect lifespan” using AI.
However, if the terminology and context defined by each lab differ, the AI hits a wall. To the AI, these are not “connected pieces of knowledge” but merely “unrelated clusters of numbers” with no inherent meaning. Consequently, before even running an AI model, researchers waste over 80% of their total time on data preprocessing—collecting scattered data, aligning terms, and filling in missing contexts. This is the single biggest obstacle to the “innovation in research speed” that AI is intended to provide.
Implementing ontology in materials data means breaking down these silo walls and granting data a “lingua franca” (common language) and “intelligence.” When relationships are logically defined through ontology, AI begins to grasp the hidden context of knowledge without human intervention.
For example, in an ontology-based system, an AI can perform high-level logical reasoning:
“Analyzing the dataset, I see that Element A stabilizes the lattice constant when combined with Element B within a specific crystal structure. Therefore, this combination will serve as a key mechanism for increasing thermal stability in high-temperature environments.”
In this way, ontology serves as a “bridge of knowledge,” allowing AI to go beyond simple data integration to understand the causality of materials science and suggest new discoveries.
A Blueprint for Teaching the ‘Language’ of Materials to AI
“To make AI understand materials, we must first teach it the language of materials.”
This is the essence of ontology. While we will introduce various cases and trends in future posts, the most notable current trend is the Knowledge Graph approach. Materials ontology provides the foundation for converting the vast knowledge system of materials science into a knowledge graph format that AI can explore and reason through.
To turn simple text or numerical clusters into an intelligent knowledge graph, ontology designs three core elements:
- Concept Definition: Clearly defining core terms used in the materials domain, such as ‘Crystal Structure,’ ‘Conductivity,’ and ‘Sintering Temperature.’ These become ‘Nodes’ in the knowledge graph.
- Relation Setting: Creating logical links such as “A is a sub-concept of B (Is-a)” or “Process C is the cause of Property D (Causes).” These act as the ‘Edges’ of the graph, teaching causality to the AI.
- Property Assignment: Setting the range of values, units, or metadata for each concept, ensuring data precision and reliability.
Ontology thus becomes the backbone of the neural network that grants ‘intelligence’ to materials data. Within a well-structured ontology, data evolves from a mere sequence of numbers into living knowledge that proves its own value and position.
Technical Guide to Ontology Terms
When implementing ontology in actual data models, standardized technical terms (primarily OWL and RDF specifications) are used:
- Class = Concept: A set of objects (e.g., ‘Metal’ and ‘Ceramic’ are classes).
- Object Property = Relation: Defines the relationship between classes (e.g., ‘Cathode Material’ [contains] ‘Lithium’).
- Datatype Property = Property: Defines the specific data values held by a class (e.g., ‘Density’ has a [Float] value).
- Instance / Individual: Refers to specific, individual data points belonging to a class (e.g., ‘NCM811’ is an instance of the ‘Cathode Material’ class).
Closing Thoughts: On the Philosophy of Materials Ontology
Ontology is not merely a tool for organizing data. It is an epistemological task—translating the insights and intuitions humanity has gathered over thousands of years of trial and error into the logic of the digital world. The philosophical question, “How do we perceive and define materials?” directly determines how AI understands materials. Ultimately, building an ontology is about transplanting the knowledge systems we have constructed into the “brain” of the computer, creating a grand blueprint where data forms its own context and reason for existence.