Reference

Glossary

The vocabulary behind the Source Commons Framework — the concepts, standards, and record types used to document, evaluate, and reuse data sources.

Core concepts

Source Commons Framework (SCF): An open framework and directory for documenting, evaluating, and reusing data sources, tools, skills, and use cases under one interoperable metadata schema.
L1 — Portal: A catalogue or institutional access point that hosts many datasets, such as an open data portal or a national data catalogue.
L2 — Dataset: A specific dataset, file collection, API, or stable data access path that can be referenced and retrieved directly.
L3 — Fragmented source: A useful source that requires extraction, parsing, reconciliation, or cross-referencing before it can be reused as structured data.
Provenance: The documented origin of a record — its publisher, landing page, stable identifiers, and history — used to establish trust and reusability.

DCAT (Data Catalog Vocabulary): A W3C vocabulary for describing datasets and data services in catalogues. SCF uses DCAT fields such as landing page, access URL, distribution format, and theme.
DQV (Data Quality Vocabulary): A W3C vocabulary for expressing data quality. SCF evaluations use DQV dimensions, metrics, measurements, and quality annotations.
schema.org: A shared vocabulary for structured data on the web. SCF pages publish schema.org JSON-LD (Dataset, SoftwareApplication, HowTo) so search engines and AI assistants can understand records.
Wikidata QID: A stable identifier for an entity in Wikidata (for example Q4989906). SCF links sources and topics to Wikidata to enable cross-referencing and disambiguation.
Field prefixes: Column name prefixes that signal a field's vocabulary: dct_ (Dublin Core), dcat_ (DCAT), dqv_ (Data Quality Vocabulary), mcp_ (Model Context Protocol), github_ and hf_ (GitHub / Hugging Face source metadata).

Join key: A field that allows two sources to be linked, such as a shared identifier, code, or geographic key. SCF records join-key types, examples, granularity, and confidence.
Crosswalk: A documented mapping between identifiers or schemas across two sources, enabling records to be matched and merged.
Reconciliation: The process of matching records that refer to the same real-world entity across different sources, often using join keys and external references like Wikidata.
Access method: How a source is obtained — for example file download, API, open data portal, or repository — along with authentication and rate-limit details.
Stable identifier: A persistent identifier for a record or entity that does not change over time, making references durable and machine-resolvable.

MCP (Model Context Protocol): An open protocol that lets AI agents and applications connect to external tools, resources, and data sources through a standard interface.
MCP server: A concrete MCP endpoint recorded in SCF as a tool with the category 'MCP server', documenting transport, endpoint, authentication, and the tools, resources, and prompts it exposes.

Source evaluation: A quality assessment of a source using DQV — a dimension, metric, 1–5 rating, evaluator, confidence, annotation, and recommended action.
Use case: A documented reuse scenario combining one or more sources and tools, describing the question answered, workflow, join strategy, and outcome.
Skill: A reusable method, workflow, or capability package documented in YAML and Markdown, with inputs, outputs, tools, and references.
Tool: Software for accessing, validating, cleaning, reconciling, analysing, or publishing data — including MCP servers — described with category, license, and legal-risk guidance.