# Schema Modularization Architecture

**Version**: 0.2.1  
**Date**: 2025-11-09  
**Status**: ✅ Complete

## Overview

The Heritage Custodian schema has been modularized from a single 1,102-line YAML file into 7 focused, reusable modules. This architecture enables:

- **Independent usage**: Use only the modules you need
- **Clear separation of concerns**: Core entities vs. provenance vs. collections vs. bibliographic
- **Easier maintenance**: Changes isolated to specific modules
- **Better comprehension**: Each module is self-contained and focused
- **Future extensibility**: Easy to add country-specific or domain-specific modules
- **Standards integration**: SPAR Ontologies (FaBiO, CiTO, PRO, PSO, DoCO) for scholarly publications

## Module Structure

```
schemas/
├── heritage_custodian.yaml    # Main schema (74 lines, imports all modules)
├── enums.yaml                 # Enumerations (institution types, data tiers, etc.)
├── core.yaml                  # Core organizational classes
├── provenance.yaml            # Provenance tracking, GHCID history, change events
├── collections.yaml           # Collections, digital platforms, partnerships
├── dutch.yaml                 # Dutch-specific extensions
└── bibliographic.yaml         # NEW: SPAR Ontologies for scholarly publications
```

## Module Details

### 1. `enums.yaml` (8,671 bytes)

**Purpose**: Define all enumeration types used across the schema.

**Contents**:
- `InstitutionTypeEnum` - 13 heritage institution types (GALLERY, LIBRARY, ARCHIVE, MUSEUM, etc.)
- `OrganizationStatusEnum` - Operational status (ACTIVE, INACTIVE, MERGED, etc.)
- `DataSourceEnum` - Data provenance sources (ISIL_REGISTRY, WIKIDATA, CONVERSATION_NLP, etc.)
- `DataTierEnum` - Data quality tiers (TIER_1_AUTHORITATIVE → TIER_4_INFERRED)
- `MetadataStandardEnum` - Standards used (DUBLIN_CORE, MARC21, EAD, RIC_O, etc.)
- `DigitalPlatformTypeEnum` - Platform types (COLLECTION_MANAGEMENT, DIGITAL_REPOSITORY, etc.)
- `ChangeTypeEnum` - Organizational changes (FOUNDING, CLOSURE, MERGER, RELOCATION, etc.)

**Usage**:
```yaml
imports:
  - enums

classes:
  MyClass:
    slots:
      - institution_type  # Uses InstitutionTypeEnum
```

### 2. `core.yaml` (13,810 bytes)

**Purpose**: Core organizational classes and identification.

**Contents**:
- `HeritageCustodian` - Main heritage institution class
  - GHCID tracking (numeric, current, original, history)
  - Organizational metadata (name, type, status, description)
  - Temporal fields (founded_date, closed_date, prov_generated_at, prov_invalidated_at)
  - Hierarchies (parent_organization, sub_organizations)
  - TOOI naming (official_name, sorting_name, abbreviation)
- `Location` - Physical/virtual locations
  - Address fields (street, city, postal_code, region, country)
  - Geocoding (latitude, longitude, geonames_id)
  - Primary location flag
- `ContactInfo` - Contact information (email, phone, fax)
- `Identifier` - External identifiers (ISIL, VIAF, Wikidata, KvK)
- `OrganizationalUnit` - Departments and divisions within institutions

**Ontology Mappings**:
- `HeritageCustodian` → `org:Organization` + `prov:Entity`
- `Location` → `schema:Place`
- `ContactInfo` → `cpov:ContactPoint` + `schema:ContactPoint`
- `Identifier` → `dcterms:identifier`
- `OrganizationalUnit` → `org:OrganizationalUnit`

**Usage**:
```yaml
imports:
  - core

# Now you can use HeritageCustodian, Location, etc.
```

### 3. `provenance.yaml` (6,715 bytes)

**Purpose**: Data quality tracking, GHCID history, and organizational change events.

**Contents**:
- `Provenance` - Data provenance metadata
  - Source and tier (data_source, data_tier)
  - Extraction metadata (extraction_date, extraction_method, confidence_score)
  - Verification (verified_date, verified_by)
  - Source references (conversation_id, source_url)
- `GHCIDHistoryEntry` - Historical GHCID tracking
  - GHCID values (ghcid, ghcid_numeric)
  - Validity period (valid_from, valid_to)
  - Context (institution_name, location_city, location_country, reason)
- `ChangeEvent` - Organizational lifecycle events
  - Event metadata (event_id, change_type, event_date, event_description)
  - Affected entities (affected_organization, resulting_organization, related_organizations)
  - Documentation (source_documentation)

**Ontology Mappings**:
- `ChangeEvent` → `prov:Activity` + `tooi:Wijzigingsgebeurtenis`

**Usage**:
```yaml
imports:
  - provenance

classes:
  MyClass:
    slots:
      - provenance  # Uses Provenance class
      - change_history  # List of ChangeEvent
```

### 4. `collections.yaml` (4,834 bytes)

**Purpose**: Collections, digital platforms, and partnerships.

**Contents**:
- `Collection` - Heritage collections
  - Metadata (collection_id, collection_name, collection_description, collection_type)
  - Content (item_count, subjects, time_period_start, time_period_end)
  - Access (access_rights, catalog_url)
  - **NEW**: Custodian link (custodian slot → implements `rico:hasOrHadHolder`)
- `DigitalPlatform` - Digital systems and platforms
  - Platform info (platform_name, platform_type, platform_url)
  - Technical (vendor, implemented_standards)
- `Partnership` - Partnerships and network memberships
  - Partner info (partner_name, partner_id, partnership_type)
  - Temporal (start_date, end_date)

**Ontology Mappings**:
- `Collection` → `schema:Collection` + `rico:RecordSet`
- `DigitalPlatform` → `schema:SoftwareApplication`
- `Partnership` → `org:Membership`

**Key Feature**:
The `Collection.custodian` slot implements RiC-O's `rico:hasOrHadHolder` pattern, creating bidirectional relationships between collections and their custodial organizations.

**Usage**:
```yaml
imports:
  - collections

classes:
  MyClass:
    slots:
      - collections  # List of Collection objects
      - digital_platforms
      - partnerships
```

### 5. `dutch.yaml` (4,931 bytes)

**Purpose**: Dutch-specific extensions for heritage institutions.

**Contents**:
- `DutchHeritageCustodian` - Extends HeritageCustodian
  - Dutch identifiers (kvk_number, gemeente_code)
  - Regional info (provincie)
  - Networks (samenwerkingsverband)
  - Aggregation platforms (in_museum_register, in_rijkscollectie, in_collectie_nederland, in_archieven_nl)

**Ontology Mappings**:
- `DutchHeritageCustodian` → `HeritageCustodian` + `tooi:Overheidsorganisatie`

**Usage**:
```yaml
imports:
  - dutch

# Now you can use DutchHeritageCustodian
```

### 6. `bibliographic.yaml` (949 lines)

**Purpose**: SPAR Ontologies integration for scholarly publications about heritage institutions.

**Contents**:
- **9 Classes**:
  - `ScholarlyWork` - FRBR Work level (abstract intellectual creation)
  - `Publication` - FRBR Expression level (concrete realization with authors, dates)
  - `Journal` - Academic journals with ISSN, impact factors, CiteScore
  - `Conference` - Conference proceedings and papers
  - `Person` - Authors, editors, reviewers with ORCID support
  - `Organization` - Publishers, affiliations with ROR IDs
  - `Citation` - Citation relationships between publications (CiTO)
  - `DocumentSection` - Document structure (abstract, methods, results, discussion)
  - `PublishingRole` - Author/editor/reviewer roles (PRO)

- **5 Enumerations**:
  - `PublicationTypeEnum` - 13 types (journal article, conference paper, book, thesis, preprint, etc.)
  - `CitationTypeEnum` - 14 CiTO types (cites, citesAsAuthority, supports, refutes, critiques, etc.)
  - `OpenAccessStatusEnum` - 6 types (fully open, green OA, hybrid, closed, embargoed, bronze)
  - `PublishingStatusEnum` - 8 PSO states (submitted, under-review, accepted, published, retracted, etc.)
  - `DocumentSectionTypeEnum` - 9 DoCO types (abstract, introduction, methods, results, discussion, etc.)
  - `PublishingRoleTypeEnum` - 9 PRO roles (author, editor, reviewer, translator, corresponding-author, etc.)

**SPAR Ontologies Integrated**:
- **FaBiO** (FRBR-aligned Bibliographic Ontology) - Core bibliographic classes
- **CiTO** (Citation Typing Ontology) - Citation relationships
- **BiRO** (Bibliographic Reference Ontology) - Citation references
- **DoCO** (Document Components Ontology) - Document sections
- **PRO** (Publishing Roles Ontology) - Author/editor/reviewer roles
- **PSO** (Publishing Status Ontology) - Publishing workflow
- **DataCite** - DOI, ORCID identifiers

**Validation Patterns**:
```yaml
doi: "^10\\.\\d{4,9}/[-._;()/:A-Za-z0-9]+$"     # DOI format
issn: "^\\d{4}-\\d{3}[\\dX]$"                   # ISSN format
isbn: "^(97[89])?[0-9]{9}[0-9X]$"               # ISBN-10/13
orcid: "^\\d{4}-\\d{4}-\\d{4}-\\d{3}[0-9X]$"    # ORCID format
ror_id: "^https://ror\\.org/0[a-z0-9]{6}[0-9]{2}$"  # ROR ID
```

**Ontology Mappings**:
- `ScholarlyWork` → `fabio:Work` + `frbr:Work`
- `Publication` → `fabio:Expression` + `frbr:Expression`
- `Journal` → `fabio:Journal` + `bibo:Journal`
- `Conference` → `fabio:AcademicProceedings` + `bibo:Conference`
- `Person` → `foaf:Person`
- `Organization` → `schema:Organization`
- `Citation` → `cito:Citation` + `biro:BibliographicReference`
- `DocumentSection` → `doco:Section`
- `PublishingRole` → `pro:PublishingRole`

**Integration with HeritageCustodian**:
```yaml
# core.yaml now includes:
HeritageCustodian:
  slots:
    - publications  # Links to Publication class
      # Tracks scholarly works about/by this institution
```

**Usage**:
```yaml
imports:
  - bibliographic

# Example: Publication with citations
- publication_id: https://doi.org/10.1007/s00799-023-00123-4
  title: "Linked Open Data for Museum Collections: The Rijksmuseum Case"
  publication_type: JOURNAL_ARTICLE
  authors:
    - person_name: "John Smith"
      orcid: "0000-0002-1234-5678"
      affiliation:
        organization_name: "University of Amsterdam"
  published_in:
    journal_title: "International Journal on Digital Libraries"
    issn: "1432-5012"
    impact_factor: 2.5
  doi: "10.1007/s00799-023-00123-4"
  open_access_status: FULLY_OPEN_ACCESS
  publishing_status: PUBLISHED
  citations:
    - citing_work: https://doi.org/10.1007/s00799-023-00123-4
      cited_work: https://doi.org/10.1145/2970000.2970111
      citation_type: CITES_AS_AUTHORITY
      citation_context: "Following the approach of Smith et al. [42]..."
```

**FRBR Work/Expression Pattern**:
```yaml
# Abstract work
- work_id: https://w3id.org/heritage/work/glam-linked-data
  work_title: "Linked Open Data for Heritage Institutions"
  discipline:
    - Computer Science
    - Library and Information Science
  has_expression:
    # English journal article
    - publication_id: https://doi.org/10.1007/s00799-023-00123-4
      title: "Linked Open Data for Museum Collections"
      language: en
      publication_year: 2023
    
    # Dutch book chapter
    - publication_id: https://isbn.org/9789012345678
      title: "Gekoppelde Open Data voor Musea"
      language: nl
      publication_year: 2024
```

**Citation Graph Example (CiTO)**:
```yaml
# Paper A cites Paper B as authoritative source
- citation_id: https://w3id.org/heritage/citation/001
  citing_work: https://doi.org/10.1007/paper-a
  cited_work: https://doi.org/10.1145/paper-b
  citation_type: CITES_AS_AUTHORITY
  citation_intent: "Foundational methodology for LOD extraction"
  citation_context: "Following the approach of Jones et al. [15]..."
  page_number: "p. 7"

# Paper C refutes Paper D
- citation_id: https://w3id.org/heritage/citation/002
  citing_work: https://doi.org/10.1016/paper-c
  cited_work: https://doi.org/10.1108/paper-d
  citation_type: REFUTES
  citation_intent: "Challenge assumptions about metadata quality"
  citation_context: "Contrary to Smith's findings [23], we observe..."
```

**Document Structure (DoCO)**:
```yaml
- publication_id: https://doi.org/10.1007/example
  title: "Semantic Technologies for Heritage Institutions"
  document_sections:
    - section_id: https://doi.org/10.1007/example#abstract
      section_type: ABSTRACT
      section_content: "This paper presents..."
      section_order: 1
    
    - section_id: https://doi.org/10.1007/example#introduction
      section_type: INTRODUCTION
      section_content: "Heritage institutions face challenges..."
      section_order: 2
    
    - section_id: https://doi.org/10.1007/example#methods
      section_type: METHODS
      section_content: "We developed a LinkML schema..."
      section_order: 3
    
    - section_id: https://doi.org/10.1007/example#results
      section_type: RESULTS
      section_content: "Our approach successfully extracted..."
      section_order: 4
```

**Publishing Workflow (PSO)**:
```yaml
# Track publication status over time
- publication_id: https://doi.org/10.1007/example
  title: "Heritage Institution Metadata Analysis"
  publishing_status: PUBLISHED  # Current status
  
  # Status history (not in schema, but tracked externally):
  # 2024-01-15: SUBMITTED
  # 2024-02-01: UNDER_REVIEW
  # 2024-04-10: ACCEPTED
  # 2024-05-20: IN_PRESS
  # 2024-06-01: PUBLISHED
```

## Main Schema (`heritage_custodian.yaml`)

The main schema file now serves as an aggregator, importing all 7 modules:

```yaml
id: https://w3id.org/heritage/custodian
name: heritage-custodian
title: Heritage Custodian Schema
description: >-
  Comprehensive LinkML schema for modeling heritage institutions...

version: 0.2.1

# Prefixes (all ontologies used across modules)
prefixes:
  heritage: https://w3id.org/heritage/custodian/
  org: http://www.w3.org/ns/org#
  prov: http://www.w3.org/ns/prov#
  rico: https://www.ica.org/standards/RiC/ontology#
  fabio: http://purl.org/spar/fabio/
  cito: http://purl.org/spar/cito/
  datacite: http://purl.org/spar/datacite/
  # ... (30+ prefixes)

# Import all modules
imports:
  - linkml:types
  - enums
  - core
  - provenance
  - collections
  - dutch
  - bibliographic  # NEW: SPAR Ontologies for scholarly publications
```

## Using Modules

### Full Schema (All Modules)

```python
from linkml_runtime import SchemaView

sv = SchemaView('schemas/heritage_custodian.yaml')
# Access all 21 classes, 155+ slots, 12 enums
# Includes bibliographic module for scholarly publications
```

### Individual Modules

```python
# Just enumerations
sv = SchemaView('schemas/enums.yaml')

# Just core classes
sv = SchemaView('schemas/core.yaml')

# Just provenance tracking
sv = SchemaView('schemas/provenance.yaml')

# Just bibliographic (SPAR Ontologies)
sv = SchemaView('schemas/bibliographic.yaml')
```

### Selective Imports

Create a custom schema importing only what you need:

```yaml
id: https://example.org/my-schema
imports:
  - linkml:types
  - enums
  - core
  # Omit provenance, collections, dutch, bibliographic if not needed

classes:
  MyCustomClass:
    is_a: HeritageCustodian
    # Use core classes without Dutch-specific fields
```

### Domain-Specific Module Example (Bibliographic)

For scholarly publication tracking:

```yaml
id: https://example.org/publication-tracker
imports:
  - linkml:types
  - bibliographic  # Just SPAR Ontologies classes

# Track publications about heritage institutions
# without importing full organizational metadata
classes:
  SimplifiedPublication:
    is_a: Publication
    slots:
      - title
      - doi
      - authors
      - citations
```

## Benefits

### 1. **Independent Usage**
Use only the modules relevant to your use case. For example:
- Global dataset without Dutch institutions → omit `dutch.yaml`
- Simple institution list without provenance → omit `provenance.yaml`
- Just enumerations for validation → use `enums.yaml` alone

### 2. **Clearer Organization**
Each module has a single, focused purpose:
- **Enums**: Type definitions
- **Core**: Organizational structure
- **Provenance**: Data quality and history
- **Collections**: Holdings and platforms
- **Dutch**: Country-specific extensions

### 3. **Easier Maintenance**
Changes are isolated to relevant modules:
- Add new institution type → edit `enums.yaml`
- Add new core field → edit `core.yaml`
- Add new country module → create `brazil.yaml` (doesn't affect existing modules)

### 4. **Better IDE Support**
Smaller files load faster and are easier to navigate in IDEs.

### 5. **Future Extensibility**
Easy to add country-specific and domain-specific modules following established patterns:

**Country-Specific Modules** (following `dutch.yaml` pattern):
```
schemas/
├── heritage_custodian.yaml
├── enums.yaml
├── core.yaml
├── provenance.yaml
├── collections.yaml
├── dutch.yaml
├── brazil.yaml      # New: Brazilian-specific extensions (CNPJ, Ibram registry)
├── vietnam.yaml     # New: Vietnamese-specific extensions
└── japan.yaml       # New: Japanese-specific extensions (ISIL-JP, NACSIS)
```

**Domain-Specific Modules** (following `bibliographic.yaml` pattern):
```
schemas/
├── heritage_custodian.yaml
├── bibliographic.yaml  # SPAR Ontologies for scholarly publications
├── museum_objects.yaml # Future: CIDOC-CRM for museum object cataloging
├── archival_desc.yaml  # Future: RiC-O for archival description
└── conservation.yaml   # Future: Conservation event tracking
```

## Validation

### LinkML Validation

All modules validate successfully with LinkML runtime:

```bash
$ python3 -c "from linkml_runtime import SchemaView; sv = SchemaView('schemas/heritage_custodian.yaml'); print(f'✓ {len(list(sv.all_classes()))} classes, {len(list(sv.all_slots()))} slots, {len(list(sv.all_enums()))} enums')"
✓ 21 classes, 190 slots, 13 enums
```

### Backward Compatibility

All existing parsers and tests pass without modification:

```bash
$ pytest tests/parsers/ -q
53 passed in 0.40s  ✓
```

### Python Models

Generated Pydantic models work correctly:

```python
from glam_extractor.models import DutchHeritageCustodian, Provenance, Publication
# All models functional, no breaking changes
```

## Migration Guide

### For Schema Developers

**Before (monolithic)**:
- Edit 1,102-line `heritage_custodian.yaml`
- Hard to find relevant sections
- Risk of breaking unrelated code

**After (modular)**:
- Edit focused module (e.g., `collections.yaml` for collection fields)
- Clear module scope
- Changes isolated

### For Schema Users

**No changes required!** The main schema still works exactly as before:

```python
from linkml_runtime import SchemaView
sv = SchemaView('schemas/heritage_custodian.yaml')
# All classes, slots, enums available as before
```

### For Extension Developers

**New pattern**: Create country/domain-specific modules:

```yaml
# schemas/my_extension.yaml
id: https://example.org/my-extension
imports:
  - core  # Import base classes

classes:
  MyCustomCustodian:
    is_a: HeritageCustodian
    slots:
      - my_custom_field
```

## Metrics

| Metric | Before (v0.1) | After (v0.2.1) | Change |
|--------|---------------|----------------|--------|
| Main schema size | 1,102 lines | 73 lines | **-93%** |
| Total schema size | 1,102 lines | 2,436 lines* | +121% (comprehensive) |
| Modules | 1 | 7 | **Focused separation** |
| Classes | 12 | 21 | **+9 bibliographic classes** |
| Slots | 104 | 190 | **+86 bibliographic slots** |
| Enums | 7 | 13 | **+6 bibliographic enums** |
| Test coverage | 53 tests | 53 tests | ✓ 100% pass rate |
| Ontologies integrated | 5 | **12** | **+7 SPAR ontologies** |

*Includes extensive documentation, examples, and SPAR Ontologies integration

### Module Breakdown

| Module | Lines | Classes | Key Features |
|--------|-------|---------|--------------|
| `enums.yaml` | 257 | 0 | 13 enumerations (institution types, data tiers, citation types, etc.) |
| `core.yaml` | 610 | 5 | HeritageCustodian, Location, ContactInfo, Identifier, OrganizationalUnit |
| `provenance.yaml` | 237 | 3 | Provenance, GHCIDHistoryEntry, ChangeEvent |
| `collections.yaml` | 194 | 3 | Collection, DigitalPlatform, Partnership |
| `dutch.yaml` | 117 | 1 | DutchHeritageCustodian (extends HeritageCustodian) |
| `bibliographic.yaml` | 948 | 9 | ScholarlyWork, Publication, Journal, Conference, Person, Organization, Citation, DocumentSection, PublishingRole |
| `heritage_custodian.yaml` | 73 | 0 | Main aggregator (imports all modules) |
| **TOTAL** | **2,436** | **21** | Comprehensive heritage institution modeling |

## References

- **Schema Files**: `/schemas/`
- **Documentation**: `/docs/`
- **Tests**: `/tests/parsers/`
- **Models**: `/src/glam_extractor/models.py`

---

**Last Updated**: 2025-11-09  
**Schema Version**: 0.2.1