Workshop on Ontologies for Materials-Databases Interoperability (OMDI2021)

In the field of materials science, the increasing adoption of data-driven methods calls for a higher degree of standardisation. Open databases of materials properties have appeared during the last decade and are rapidly growing. This availability of data has opened up for entirely new ways to perform materials research and have created new challenges. A number of EU projects relate to interoperability and seamless data sharing in materials science, including BIG-MAP, OpenModel, DOME 4.0, MarketPlace, and INTERSECT. It is becoming increasingly clear that the standardisation of databases, and the field in general, would strongly benefit from more fundamental work on ontologies for materials databases.

Ontologies

Ontologies aim to define the basic terms and relations of a domain of interest, as well as the rules for combining these terms and relations. They standardise terminology in a domain of interest and are a basis for semantically enriching data, semantic search, integration of data from different data sources, and reasoning over the data. Using ontologies can alleviate challenges associated with the variety (data sources are heterogeneous regarding the type and nature of data they store), variability (data can be inconsistent), and veracity (not all data can be trusted). Further, they are proposed as an enabler making data FAIR, i.e., findable, accessible, interoperable, and reusable, with the purpose of enabling machines to automatically find and use the data, and individuals to easily reuse the data. One of the most successful examples of ontologies in science is the GeneOntology (GO) with upwards 20 thousand citations. It showcases the benefits of ontology usage in scientific research.

Importance for materials databases

An ontology for materials databases establishes semantic relations between common materials definitions and thus formally provides a directed graph with communal conceptual constructs. This formalisation is crucial for, e.g., data integration and semantic-based search. Combining unified access to materials databases with a shared understanding of materials properties and relations and the use of a common data format will be of major benefit to the materials science community and enable databases and software to work together in more informed and useful ways.

Existing ontologies in materials science

There are several existing and ongoing efforts in materials science-related taxonomies and ontologies. The EMMO effort of the EMMC is a well-known broad approach to an ontology for physics. With the latest materials science-related EU projects it is fast becoming a basis for other domain-specific efforts to be applied in interoperational software for high-throughput computing, machine learning/AI, and seamless data sharing. Other efforts are domain-specific ontologies often tied to specific materials databases, e.g., the MetaInfo structure of NOMAD. Other examples of materials-oriented ontologies include MatOnto, MatOWL, NanoParticle, and MMOY. There are also many well-established materials science glossaries, such as the Pauling file and Electronic Structure Library.

Community-driven standardization

For future use in standardisation and data integration, there appears to be a need to move to a more community-driven effort to design ontologies practically useful across databases, research software and AI-based approaches, e.g., machine learning. OPTIMADE has been highly successful as a community-driven effort to develop a common API for materials database data access with engagement from many major materials databases. The OMDI2021 workshop aims to leverage the part of that community interested in community-driven ontologies for materials-database interoperability and put it in contact with others already involved in ontology development for materials science, as well as researchers in semantic methods and ontologies.