Semantic Source Evidence¶

3. Direct GPC and GDSN observations¶

3.1 GPC observations¶

The source baseline reports that the supplied GPC ontology contains 22,281 owl:Class resources across six levels:

The evidence establishes that GPC is more than a category tree. Each Brick may define Attribute Types and controlled Attribute Values. This creates a category-scoped faceting model suitable for Product Holon construction and SHACL generation.

The Cheese (Frozen) chain used throughout the examples is:

50000000 Food/Beverage → 50130000 Milk/Butter/Cream/Yogurts/Cheese/Eggs/Substitutes → 50131800 Cheese/Cheese Substitutes → 10000030 Cheese (Frozen).

3.2 GDSN observations¶

The source baseline reports:

• 30,630 RDF descriptions;
• approximately 3,946 core trade-item attributes;
• numerous retailer- and region-specific extension groups;
• 557 distinct code lists.

Examples of extension-group sizes included BestBuy, Australasian Healthcare, Carrefour and GS1 US Healthcare. The important architectural conclusion is that the source model includes both globally relevant product facts and recipient-specific or operational detail. A public Product Holon view therefore requires a deliberate allow-list.

The baseline identified code lists and attributes relevant to consumer and AI use cases, including:

• 218 allergen-type values;
• 1,080 nutrient-type values;
• 937 packaging-marked accreditation values;
• 292 claim-element values;
• ingredient statements;
• preparation instructions;
• recipes;
• recycled-content ratio;
• carbon-footprint and sustainability details;
• channel-specific marketing content.

These counts and examples are observations from the supplied source-derived ontology. They do not imply that the current reference shape exposes all of them.

3.3 Evidence for a bounded product unit¶

The source models are too large to use as per-product payloads. The evidence supports selecting a bounded subgraph based on:

1. product identifier;
2. GPC Brick and ancestor chain;
3. Brick-scoped Attribute Types and populated values;
4. populated GDSN facts selected by a profile;
5. product relationships and provenance.

This product-specific graph is the evidence basis for the Product Holon concept.

3.4 Value for shopping and agent use cases¶

The source observations support the following capabilities:

• category-appropriate faceted filtering using controlled GPC values;
• allergen and dietary filtering using structured codes and statements;
• certification and trust-signalling through controlled accreditation data;
• nutrition comparison beyond basic macros where downstream profiles support it;
• claim substantiation and regulatory context;
• sustainability and packaging information;
• localised connected-pack content.

The public interpretation layer remains Web Vocabulary/schema.org or another approved target profile rather than raw exposure of the entire source model.