Analysis: Slot URI mapping coverage and alignment with Fairscape/ROCrate

[cmungall]

Summary

An analysis of how D4D schema slots map to standard vocabularies, and how those mappings align (or conflict) with the Fairscape/ROCrate Pydantic models that already have a D4D conversion layer (fairscape_models/conversion/mapping/d4d.py).

D4D Slot URI Coverage

Of ~414 domain-specific attributes across the D4D modules, approximately 28% are mapped to standard vocabulary URIs via slot_uri:

Module	Mapped/Total	Coverage
Distribution	3/3	100%
Ethics	7/7	100%
Motivation	8/9	88%
Base (core)	41/65	63%
Maintenance	7/13	53%
Uses	4/10	40%
Variables	9/27	33%
Composition	20/62	32%
Preprocessing	6/22	27%
Collection	5/20	25%
Data Governance	6/39	15%
Evaluation Summary	0/122	0%
Human	0/14	0%

Vocabularies used in D4D

• dcterms: (Dublin Core) — 70 mappings (dominant)
• schema: (Schema.org) — 33
• dcat: (Data Catalog) — 9
• prov:, skos:, qudt:, DUO: — 1 each

Notable gaps

• Evaluation Summary (122 slots, 0 mapped) — the largest module with zero URI mappings
• Human (14 slots, 0 mapped) — human subjects data with no standard vocab links
• Data Governance (39 slots, 15%) — governance/consent terms could map to DUO/ODRL
• No mappings to DATS, OBI, IAO, or other biomedical metadata standards

Fairscape URI Approach

Fairscape uses JSON-LD with @vocab: "https://schema.org/" as default namespace plus evi: "https://w3id.org/EVI#" for extensions and rai: for responsible AI fields. Key mappings in Fairscape:

Fairscape field	Effective URI	Standard
name	schema:name	Schema.org
description	schema:description	Schema.org
@id	schema:identifier	Schema.org
dateCreated	schema:dateCreated	Schema.org
dateModified	schema:dateModified	Schema.org
contentUrl	schema:contentUrl	Schema.org
license	schema:license	Schema.org
author	schema:author	Schema.org
contentSize	schema:contentSize	Schema.org
evi:formats	https://w3id.org/EVI#formats	EVI (custom)
rai:dataUseCases	RAI namespace	Custom
rai:dataBiases	RAI namespace	Custom

D4D ↔ Fairscape Alignment at the URI Level

Both schemas use Schema.org for core metadata. Where they overlap:

D4D slot_uri	Fairscape JSON-LD	Alignment
schema:name	schema:name	✅ Exact
schema:description	schema:description	✅ Exact
schema:identifier	schema:identifier	✅ Exact
schema:license	schema:license	✅ Exact
schema:url	schema:url	✅ Exact
dcterms:created	schema:dateCreated	⚠️ Same concept, different vocab
dcterms:modified	schema:dateModified	⚠️ Same concept, different vocab
dcterms:creator	schema:author	⚠️ Same concept, different vocab
dcat:downloadURL	schema:contentUrl	⚠️ Same concept, different vocab

The core tension

D4D leans on Dublin Core (dcterms:) for provenance and dates, while Fairscape uses Schema.org for everything. Both are valid standard vocabularies, but this creates unnecessary mapping friction. For example:

• dcterms:created vs schema:dateCreated — semantically identical
• dcterms:creator vs schema:author — nearly identical
• dcat:downloadURL vs schema:contentUrl — same concept

Fairscape's existing D4D mapping

fairscape_models/conversion/mapping/d4d.py already contains a ROCRATE_TO_D4D_MAPPING dict that maps D4D field names to ROCrate/Fairscape source keys. However, this mapping operates at the field name level, not at the URI level. A formal URI-level alignment (e.g., via SSSOM) would:

1. Make the mapping vocabulary-aware and auditable
2. Capture the dcterms↔schema.org equivalences explicitly
3. Identify D4D slots that have no Fairscape equivalent (and vice versa)
4. Enable automated interoperability tooling

Proposed Next Steps

1. Add slot_uri mappings to the currently unmapped D4D modules, prioritizing Evaluation Summary (0%) and Human (0%)
2. Consider harmonizing the dcterms vs schema.org choice — or at minimum, add exact_mappings cross-references between them
3. Produce a formal SSSOM mapping between D4D slot URIs and Fairscape/ROCrate URIs
4. Consider adding mappings to domain-relevant standards: DUO (consent), OBI (assays), IAO (information artifacts)

[realmarcin]

Update

This analysis has been updated with significant new developments in #131:

Major Updates:

1. ✅ FAIRSCAPE Pydantic Models Integration - Runtime validation and type safety
2. ✅ SKOS Semantic Alignment - 88 formal D4D ↔ RO-Crate mappings (59% exact, 23% close, 11% related)
3. ✅ Reference Implementation - CM4AI full-ro-crate-metadata.json as canonical FAIRSCAPE example
4. ✅ Working Converter - D4D → FAIRSCAPE transformation with Pydantic validation
5. 📊 Updated Slot URI Coverage - 93.9% (31/33 slots) vs original 28% claim

Key Changes:

• Schema structure significantly consolidated since original analysis
• FAIRSCAPE models now integrated as git submodule
• Formal semantic mappings capture dcterms ↔ schema.org equivalences
• Converter tested and validated with VOICE D4D data

See #131 for complete updated analysis.