UNEDITED NOTES
facilitator: Rob Sanderson
Things to Things
Dynamic
- provenance
- annotations as nanopub
Degrees of sameness
Outside our domain?
- no need to be perfect
Event stream
Use case – authorities
Deleted resources – tombstones
Entities joining/splitting
Alternate identities
Services/tools
- certainty and risk aversion
- more data = better, = more uses
(new group)
Sameness?
Reconcile at work level
Context – what we use the data for!
Reconcile most discriminant attributes
Similarity pocky for sameness
- certainty?
- risk aversion
- discovery vs. inference/data management
- schema:samas – avoiding the alt itoces
Indexing & uncertainty
- perfect as enemy of good enough
- sameAs external link as vote for URI
- distinctness different for different types of entity
special cases are special: Bible, Mark Twain
(new group)
Services like ref. cleaning
- identify auth. we trust
- e.g., VIAF
- map between authorities
- devil in details for mapping
- mechanism to address errors in auth!
- disambiguate
"Authority" – what do we really mean?
- Wikipedia/media wiki as authority
- Open URI resolver?
- endorsement of approach by adopters
- divide and conquer easy bits
- start with under-served domains
- cheaper going forward
- market will decide on URIs to use
- vocab/facet reconciliation
- places & dates, concepts with multilingual
(new group)
Ontology alignment
Global auth & global contexts
Dynamic entities : join/split
Pseudonym – Europe vs. US
Local assertion = local URI
Pre-existing identifiers
Link all the things
Subjects broader issue than people
- sameness degree?
- certainty?
Tools to produce the data?
- languages and character sets
- different to strings to things
- service to publish alignments
- provenance important for trust