Semantic Enhancement Guide

Overview

Semantic enhancement in SDCStudio connects your data to universal ontologies and knowledge bases, improving AI understanding and enabling true data interoperability. This guide explains how to work with ontologies and semantic linking.

What is Semantic Enhancement?

The Problem

Traditional data models use local, ambiguous terms: - "Status" could mean order status, account status, or health status - "Weight" could be in kg, lbs, or other units - "Date" could be creation, modification, or expiration

The Solution

Semantic enhancement links your data to universal ontologies: - "Status" → SNOMED: 263490005 (clinical status) - "Weight" → UCUM: kg (kilogram) - "Date" → schema.org: dateCreated

Benefits: - ✅ Universal understanding - ✅ Better AI suggestions - ✅ Data interoperability - ✅ Automated validation - ✅ Standards compliance

Working with Ontologies

Uploading Ontologies

  1. Navigate to Ontologies:
  2. Go to Settings → Preferences → Click "Manage Ontologies" link
  3. Or navigate directly to /app/ontologies
  4. Click "Upload Ontology" button
  5. Select File:
  6. Drag and drop your file
  7. Or click to browse
  8. Supported formats:
    • Turtle (.ttl)
    • OWL (.owl)
    • RDF (.rdf)
    • N3 (.n3)
    • N-Triples (.nt)
  9. Maximum size: 10MB
  10. Configure Required Fields:
  11. Namespace abbreviation (max 15 characters)
  12. Namespace URI
  13. Optional Configuration:
  14. Description
  15. Organization
  16. Access Level (private, team, corporate, public)
  17. Upload: Processing begins automatically

Supported Ontologies

Healthcare: - SNOMED CT: Clinical terminology - LOINC: Laboratory codes - ICD-10: Diagnosis codes - RxNorm: Medication codes

General: - schema.org: Web semantics - Dublin Core: Metadata - FOAF: People and organizations

Domain-Specific: - GeoNames: Geographic data - Industry ontologies - Custom ontologies

Managing Ontologies

Access the Ontologies page (Settings → Preferences → "Manage Ontologies" or /app/ontologies):

View Ontologies: - See all uploaded ontologies in a table - Check processing status (in triplestore or error) - Filter by status and access level - Search by name or abbreviation - Pagination for large lists

Ontology Information: - Namespace abbreviation and URI - Description and organization - Access level (private, team, corporate, public) - File size and upload date - Triple count (if successfully processed)

Actions: - View Details: Click on an ontology to see full details - Download: Download the original ontology file - Reprocess: Retry processing if errors occurred - Delete: Remove unused ontologies (with permission check)

Semantic Linking

Automatic Linking

When ontologies are uploaded, the AI automatically:

  1. Analyzes Components: Reviews component names and descriptions
  2. Searches Ontologies: Finds matching concepts
  3. Suggests Links: Proposes semantic connections
  4. Ranks Matches: Orders by relevance

Example:

Component: blood_pressure_systolic
AI Suggestion: SNOMED CT 271649006 (Systolic blood pressure)
Confidence: 95%

Manual Linking

Add semantic links manually:

  1. Edit Component
  2. Click "Add Semantic Link"
  3. Search Ontologies:
  4. Enter search term
  5. Browse results
  6. View concept details
  7. Select Match:
  8. Choose best match
  9. Review definition
  10. Confirm selection
  11. Save: Link is created

Predicate-Object Pairs

Define semantic relationships:

Common Predicates: - rdf:type: The type of thing - rdfs:subClassOf: Subclass relationship - skos:exactMatch: Exact equivalent - skos:closeMatch: Close equivalent - dct:subject: Subject/topic

Example:

Component: customer_email
Predicate: rdf:type
Object: schema:email

RAG Integration

How RAG Works

RAG (Retrieval-Augmented Generation) enhances AI with your knowledge:

1. Your Data → Uploaded to SDCStudio
2. Knowledge Base → Populated with domain info
3. AI Query → Searches knowledge base
4. Enhanced Response → Context-aware suggestions

Knowledge Base

What's Included: - Uploaded ontologies - Industry best practices - SDC4 standards - Component examples - Validation patterns

How It Helps: - Better type suggestions - Appropriate validation rules - Standards compliance - Domain-specific understanding

Using RAG

Automatic Integration: - No configuration needed - Works with all AI processing - Improves with more data

See Results: - Better component suggestions - More accurate validation - Improved semantic links - Standards-aligned models

Best Practices

Choose Relevant Ontologies

Match Your Domain: - Healthcare → SNOMED CT, LOINC - E-commerce → GoodRelations, schema.org - Geographic → GeoNames - General → schema.org, Dublin Core

Don't Over-Load: - Upload only relevant ontologies - Too many can slow search - Focus on quality over quantity

Core Components First: - Link most important components - Focus on shared concepts - Enable interoperability

Use Standard Predicates: - Stick to common predicates - Follow ontology guidelines - Ensure consistency

Review Periodically: - Check links still valid - Update as ontologies evolve - Remove broken links

Document Decisions: - Note why link was chosen - Explain alternatives considered - Record confidence level

Examples

Healthcare Example

Component: Patient Blood Glucose Semantic Links:

rdf:type → LOINC:2339-0 (Glucose [Mass/volume] in Blood)
rdfs:subClassOf → SNOMED:33747003 (Glucose measurement)
skos:exactMatch → UCUM:mg/dL (milligrams per deciliter)

Benefits: - Interoperable with EHR systems - Standards-compliant validation - Clear measurement units - Universal understanding

E-commerce Example

Component: Product Price Semantic Links:

rdf:type → schema:Price
dct:subject → GoodRelations:UnitPriceSpecification
skos:closeMatch → currency:USD

Benefits: - Web-friendly semantics - Clear currency specification - Search engine optimization - API interoperability

Troubleshooting

Ontology Upload Fails

Check File Format: - Must be TTL, OWL, RDF, N3, or NT format - Verify proper syntax - Ensure valid namespace

File Size: - Maximum file size is 10MB - Large ontologies may timeout during processing - Consider splitting into smaller files - Contact support for large uploads or increased limits

Required Fields: - Namespace abbreviation is required (max 15 characters) - Namespace URI is required - Both must be unique across your ontologies

Can't Find Matching Concept

Broaden Search: - Try different terms - Use synonyms - Search multiple ontologies

Manual Entry: - Add custom concept - Document source - Link to external reference

Verify Links: - Ensure links are valid - Check concept definitions match - Review predicate usage

Use More Ontologies: - Add domain-specific ontologies - Increase knowledge base coverage - Link more components

Next Steps

Getting Help

Ready to enhance your data? Upload your first ontology and start linking!