Metadata definition and use with Hydra
Unconference session notes. Attendance:
- Anusha Ranganathan
- Chris Colvard
- Hector Correa
- Declan Fleming
- Corey Harper
- Mark Matienzo
- Julie Hardesty
- Joe Atzberger
- Tom Johnson
- Karen
- Sharon Farnel
- Matt C.
- Thomas Brittnacher
- Wendy Hagenmeier
- many others
Structural Metadata
Tom Johnson: Oregon digital did compound objects. Problem with track-list and similar structures is that order is hard in RDF. Hierarchy is easy, order is hard. Look at W3C media ontology as starting place including non-RDF.
Declan: use object literal. Pull it out when needed.
Tom: "Each of the ordered list ontologies currently being worked on are horrible in their own ways"
Existing ordered list structures:
- RDF List is "linked list", as in low level programming.
- RDF Sequence is very similar
- Ordered List Ontology is wholly separate
- ORE
- Collections
Declan: We want to agree here how we are going to do it. We use list now, but it is almost impossible to build a form from.
Julie H.: Sequence seemed like a better option for our experiments
Sequence is scarcely used elsewhere and not in Hydra anywhere.
Possible models for useful expressions: JSON-LD and turtle.
Declan: current approach is List, but considering using JSON as object literals
Anusha et al.: Mixing and matching ontologies/properties is recommended
Anusha: Do not develop the ontology/data model on the fly during development. Know your model first, even if you do end up changing.
Declan: Be practical, not purist.
Intellectual relationship vs. deduplication with digitial copies of the same born digital report. Do you really want to do FRBR-like modeling of digitial manifestations
Mark M.: DPLA potentially interested in contributing infrastructure for hosting collaborative metadata models.
Tom: Oregon digital considered metadataregistry.org, but it forced too much skos, was inflexible
Corey: "Vocabulary Hosting: A modest proposal" came up in dublin core community years ago. This group may now have the use case anticipated there and then.
Tom: "List is the bastard child of the whole spec".
Declan:
Tom: propose doing a ruby implementation of Ordered List Ontology as next step