Dublin Core application profiles in context Thomas Baker 22 October 2009 Knowledge Organization Systems: Managing to the Future A joint CENDI/NKOS Workshop National Agricultural Library Beltsville, MD

2 RDF – a grammar for Web links Resource Propertyy Resource Propertyy Resource Propertyy Resource Propertyy Literal (descriptive text or numerical data) “Introduction to RDF” in one slide…! “Property” means “is related to”.

Interoperability Levels for Dublin Core metadata  1: Informal interoperability Shared vocabularies defined in natural language  2: Semantic interoperability Shared vocabularies based on formal semantics  3: Description Set syntactic interoperability Shared formal vocabularies in exchangeable records  4: Description Set Profile Interoperability Shared formal vocabularies and constraints in records Shared (natural-language) definitions Shared formal-semantic model Shared model for “records” Shared validatable constraints

Open- and closed-world Shared (natural-language) definitions Data in silos. “Intra-operability” within silos. Shared formal-semantic model “Open-world” data. Shared model for “records” Open-world data captured in manageable records. Shared constraints Open-world data optimized for specific environments.

Supporting technologies Shared (natural-language) definitions Closed systems. Proprietary systems. Web of APIs. DC-XML/2003 and other early DCMI specs. Shared formal-semantic model Linked data. RDF data. Extracted triples. DC-RDF. DC-HTML. RDFa! Shared model for “records” DCMI Abstract Model. DC-DS-XML. SPARQL Named Graphs. Shared constraints DCMI Description Set Profile. SPARQL Query Patterns.

Deployed base Shared (natural-language) definitions Shared formal-semantic model Shared “records” Shared constraints

Rate of growth Shared (natural-language) definitions Shared formal-semantic model Shared “records” Shared constraints

Which level do you require? Shared (natural-language) definitions Pro: Easier to deploy. Validatable records. Contra: Closed-world. Interoperability by (thousands of) ad-hoc agreements. Shared formal-semantic model Pro: Easier to integrate and migrate data. Contra: Harder to design, less tools. Shared model for “records” Pro: Provenance. Trust. Contra: Lack of mature, deployed models. Shared constraints Pro: Validation. Quality. Contra: It is “constraining”…

Level-1 apps interoperate with shared or mapped schemas Schema ASchema BSchema C same as mapped to

Good level-2 Application Profiles create good triples Profile AProfile BProfile C

Good triples can be merged coherently Profile AProfile BProfile C

Applications come and go… Profile AProfile BProfile C

The data remains

Data quality is independent of profiles used to create it SPARQL Endpoint Queries