A proper metadata content can tackle many of the issues addressed by the FAIR Principles. Search engines usually index metadata content, supporting findability. For accessibility, metadata should contain information about the license (who or which algorithms can access the data under what conditions) and the accessibility protocol. For interoperability and reusability, the metadata should contain information about the semantics in the data and metadata, and about data provenance. In addition, the metadata should comply with domain-relevant community standards.
As metadata play a central role in the FAIR principles and the GO FAIR metrics, GO FAIR recommends a layered metadata approach, composed of the following layers:
- Data Repository Metadata: The Data Repository Metadata layer contains information about the data repository that provides the data and its governing entity. This informs data users about which entity provides the requested metadata and data. In addition, the Data Repository Metadata layer provides information about the data repository service itself, such as API version, service license, etc.
- Catalogue Metadata: Data repositories may opt to organise their datasets in different groups. Each of these groups is represented by its own Catalogue Metadata.
- Dataset Metadata: The Dataset Metadata defines information such as the dataset’s name, description, authors/publishers, version, date of last modification, license, and rights. This is the central layer.
- Distribution Metadata: A distribution is a specific serialisation of the dataset. For instance, a given dataset may be made available in RDF, XML, and CSV, or as a relational database. The Distribution Metadata layer informs data users which serialisation formats are available for the dataset.
- Data Record Metadata: The Data Record Metadata layer contains metadata about the structure and semantics of the data. It is the most complex layer; it contains concepts that are represented in the data, their relationships, domain, range, and restrictions.