SNAP:DRGN & CIDOC-CRM conversation
Friday, May 15, 2015: King’s College London
Attending: Gabriel Bodard (GB), Arianna Ciula (AC), Øyvind Eide (OE), Faith Lawrence (FL), Christian-Emil Ore (CEO), Paul Rissen (PR), Valeria Vitale (VV), Hafed Walda (HW).
Apologies: John Bradley, Steve Stead.
We have three main topics of discussion:
- personal relationships (the SNAP “bond” ontology);
- co-references (and the inferences that derive from them);
- SNAP use of ontologies, and mapping to CRM.
1. Personal relationships/bonds
- CEO: The CRM defines several types of relationship (e.g. event, group, unilateral, family)
- GB: SNAP will eventually need to cover many more than just family/household relationships, as we currently have in the ontology
- OE: Is the aim to map all SNAP relationships to the CRM ontology, so we can always represent SNAP in CRM?
- FL: gave a summary of snap ontology
- (digression on equivalences between [lawd|crm|foaf|snap|etc.]:Person; is this a union set, rather than a single overlapping definition?)
- dual-classes of relationships: serious/casual/legally recognized; social contracts; intimate; household; foster/adopted/inlaw/claimed;
- gender (and other assumptions) not modelled in top-level classes, but could (maybe should) be?
- CEO: we can model all of these with group relationships, and then type them. Maybe we should make a recommendation for extending CRM with SNAP classes?
- GB: can we model non-family relationships (at the top-level)?
- [Not currently.]
- CEO: We should probably leave events out of the typology for now.
- VV: How can “ContractualRelationship” and “Relationship” be siblings?
- (GB: rename “Rel” ~~> “QualifierRel”)
- GB: then let’s just list all the new non-family relationships we need, and worry about grouping them (or not) later.
- OE: Suggest a follow-up meeting on CRM-INF (with Steve Stead, Dominic Oldman and Hugh Cayless).
- PR: BBC programmes ontology defines relationships by membership to groups
- could build on snap bonds for new relationships; useful for scholarly/journalistic claims, inference etc.
2. Co-references–both unambiguous and suggested (and inference)
- OE: distinguishing explicit and implicit co-reference:
- implicit co-reference is what all scholarly systems do, x=y.
- explicit co-reference requires attributing statement to someone
- a negative co-reference has no common target
- do co-references need a target?
- CEO: CRM doesn’t model identity, per se; two entities with different ids are therefore two different entities. CRM can’t express conflicting/untrue information.
- AC: it’s the expressing of conflicting opinions that is the problem there, if you want to keep both.
- HW: is identity a combination of person and context?
- CEO: They have different identifiers/are in different data-spaces.
- OE: SNAP needs to make explicit co-reference statements (with “belief system” over the top). Do we want fuzzy reliability values on them? (GB: no!)
3. SNAP ontology(ies)
- AC: lawd:hasAttestation/hasCitation == crm:isReferredToBy ?
- AC: lawd:nameAttestation == crm:appellation ?
- AC: CRM-inf might be more useful for Scenario 3 (unambiguous and unproblematic co-reference) than Scenario 4 (scholarly commentary about co-references/relationship), because CRM is a bit deterministic.