Patrick Colm Audley

I've started a new ontology project: GMEOW — the Global Metadata and Entity Ontology for the Web.

The design goal is a reasoning-centric, OWL 2 DL, gUFO-grounded super-vocabulary for modelling a person's or organization's digital existence.

This came out of a practical problem: once you start trying to build local-first agents over real personal or organizational memory, you quickly run into vocabulary fragmentation.

Contacts, email, documents, projects, notes, legal agreements, genealogy, publications, accounts, calendars, and social presence all have their own mature-but-isolated ways of describing the world. FOAF, REL, DOAP, GEDCOM, PROV-O, ORG, schema.org, vCard, Wikidata, and others all carry useful structure — but none of them gives you the whole shape.

GMEOW's approach is to mint canonical terms and align outward. So rather than rewriting source data, it creates a coherent upper layer where surface vocabularies can map into a common model. The ontology is grounded in gUFO, checked against OWL 2 DL constraints, and built around the idea that provenance, confidence, temporal validity, and coreference are first-class modelling concerns.

The project grows by slices. The first slice is entities + contacts. Each slice adds canonical terms, SSSOM alignment tables, fixtures, and coverage reporting so progress is measurable against real data rather than wishful thinking.

The toolchain is also part of the point: validate, reason, mappings, Wikidata checks, metadata, content negotiation, docs, build artifacts, and publishing support all need to run cleanly. Ontologies should have CI discipline too.

I'm particularly interested in feedback from people working with OWL/RDF, upper ontologies, ontology alignment, semantic-web publishing, personal data stores, provenance models, local-first AI, and agent memory.

Where have you seen personal or organizational knowledge graphs break because the vocabulary layer was too thin?

The repo is open: gmeow-ontology.