mlkgTag

mlkgTag is a metadata tag used within the ML Knowledge Graph (MLKG) ecosystem to annotate machine learning resources such as datasets, models, experiments, and pipelines. It provides a lightweight, machine-readable means to classify and filter resources by task, domain, data characteristics, provenance, and other attributes, enabling easier discovery and management in large ML workflows.

The tag is defined as part of the MLKG metadata model and is designed to be interoperable across formats such as JSON-LD, RDF, and YAML. A typical mlkgTag entry consists of the tag name and optional properties, which may include provenance, version, data schema, quality metrics, and access controls. In practice, mlkgTag is used to attach contextual information directly to resources, supporting queryable catalogs and automated reasoning about lineage and dependencies.

Origin and adoption: mlkgTag emerged with the MLKG standard as a means to address inconsistent tagging across diverse ML projects. It has been described in MLKG specification drafts and is supported by tooling that can serialize and deserialize tags within resource metadata. Adoption varies by project, with some teams applying mlkgTag as a core tagging mechanism and others using it alongside custom metadata fields.

Usage and interoperability: mlkgTag is commonly applied to datasets, models, experiments, and pipelines in metadata records. When paired with standard vocabularies and schema mappings, it facilitates cross-project search, governance, and reproducibility. Best practices emphasize clear tag naming, minimal duplication, and alignment with broader metadata schemas to maximize interoperability.

See also: metadata tagging, MLKG, data provenance, dataset curation.