Repository navigation
How should NeoWiki data be projected to RDF? (NativeRdfProjection.md) #999
Replies: 5 comments 4 replies
|
Fr Q6 I have a link but it's on ECHOES' space and accessible only if you become a member: https://univtoursfr.sharepoint.com/:w:/r/sites/ECHOES2/Documents%20partages/WP%2007%20Digital%20Commons%20Knowledge-Base/ECHOES_WP07_Deliverables/240624_ECHOES_WP07_D7.1_The_Digital_Commons/URI%20policy%20for%20ECHOES.docx?d=w589eedb3a8a445e4a23cf74eaf36172e&csf=1&web=1&e=oHxasl I could give you a snapshot if you would like it, but can't commit to updating you live in the future! |
Candidate library to consider when we implement the mapping system's RDF emission (and the import/parse side in #306): hardf ( Why it fits:
Caveats: effectively single-maintainer and still 0.x, with an irregular history (active again 2024–2026). Mitigated by its small size + MIT license (forkable/vendorable) and the W3C conformance suite. Worth pinning a version and wrapping it behind one adapter rather than leaking its array term-shapes across the code. No decision implied — recording for when the RDF I/O work starts. The |
|
Doc reorganization heads-up (#920, merged today):
The issue description was updated to match: renamed links fixed (the ADR 19 link had also been broken from the start) and terminology aligned ("base mapping" → "native projection", "mapping layer" → "ontology mapping"). The open questions here (Q3, Q6, Q7, Q9) are unchanged and this remains the place to discuss them; ontology-mapping feedback is best directed at the new doc.
|
|
Commenting Q1: Property predicate scope — Flat namespace ($base/prop/Name) rather than schema-scoped ($base/prop/Person/Name). This is more natural for RDF and enables cross-schema queries (?x neo-prop:Name "Berlin" returns all types). The tradeoff: NeoWiki doesn't enforce that same-named properties across schemas are semantically equivalent, so flat namespace implies an equivalence that may not hold. In practice this is mitigated by filtering on rdf:type in queries, and the ontology mapping handles ontology alignment independently of this choice. Not sure how to respond to this one. Does defining the predicate not involve defining a domain and a range for it? This would solve it and make it reusable? Or in NeoWiki data structure no property is reusable anywhere? Q2: Standard vocabulary in the native projection — Keep it minimal (rdf:type, rdfs:label, dcterms:created/modified). Further alignment (e.g., foaf:name) happens in the ontology mapping. I can't imagine a native projection. RDF presupposes an ontology. So at best you could do rdf:type and then the type of your scheme and then maybe the system is automatically making a label somewhere so you can push the value of that to the rdfs:label. Nothing else comes for free. Need to map it and transform it if the NeoWiki schemas don't have an ontology. Q4: CIDOC-CRM alignment — This is an ontology-mapping concern, not a native-projection concern. The native projection represents NeoWiki's data model; CIDOC-CRM event expansion (generating intermediate event nodes like E12_Production) happens in the ontology mapping. Feedback from KMA based on Wikibase experience described three approaches (basic, event-centric, qualifier-based) — all of which are expressible through mappings. I guess this depends on how people build their model. If they make a shortcut predicate like Person->born_at->Place where CIDOC CRM would say crm:E21_Person->crm:P98i_was_born → crm:E67_Birth → crm:P7_took_place_at → crm:E53_Place then one would need to expand (or contract) on export/import. On the other hand the use might create a schema set that includes: Person Birth Place And then they could write Person->had_birth_event->Birth Birth->took_place_at->E53_Place and then the mapping wouldn't need expansions as nothing is implicit. Depends on the modelling done with NeoWikis' schema mechanism by the data modeller. Q5: Named graph conventions — Resolved. No CH conventions exist for named graph usage. Per-page named graphs work for operational purposes (efficient sync, provenance, deletion). No common conventions. Many conventions exist. Still open If one can have the direct and the indirect representation, that seems handy (the wikibase reification is essentially a named graph) Q6: Base URI conventions. There is a URI policy being discussed within ECCCH. We'd like to align with it. Does anyone have a pointer to the current state of that discussion? n.b.: we would be outputting to many different schemas in principle so URI policy probably has to be set by project / neowiki intanace (like something that can be setup, probably in the mapper tool) Q7: URI design for properties. Property Names can contain spaces (e.g., "Founded at", "Has author"). Two conventions have been suggested: (a) CamelCase — hasAuthor, foundedAt — following the common RDF convention, or (b) underscores — Has_author, Founded_at — more readable for non-developer users. Both are viable; this is a convention choice, not an architectural one. Either is fine Q9: Ordering of multi-valued properties. NeoWiki stores multi-valued properties as ordered arrays; RDF triples are unordered. We propose accepting the ordering loss. Are there use cases where ordering truly matters (e.g., pages in a book)? This can and should be modelled as explicit data in the model. If these are important they should be recoverable in the rdf export as data |
|
Question to: NativeRDF projection.md What is relationType defining beyond prop direct or not direct - or what else is there? just relationType is somehow confusing with the word Type being overused in semantic contexts in general. Follow up on the comments to Qs listed by @Habennin in his comments above: Q1 - I also don’t fully understand this, we should stick to standards, so domain / range should be the way to go; I think we need a fully fledged OWL ontology based on an existing standard (which then maps to some other standards). And this is why I don't see why we're separating mapping away from the data model (see also my Q8 here), i think these things have to be considered alongside. Like is neoWiki (and it's data model) something other than just a use-friendly mask that sits on top of a standard ontology (e.g. CIDOC-CRM, or something else defined in OWL and existing somewhere published in a standard serialisation) that's selected by the users via a dropdown or upload button? Maybe it is something more, but if yes, what is it? And if we can just implement a standard, fully-fledge ontology (a T-box essentially), then maybe we need help from @Habennin to design that? Q4 - Doing the full paths of CIDOC, following the existing ‘strawman’ reification seems to be the most clear way forward at least from user’ experience perspective. (see also my Q6 here) |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Discussion venue for
docs/planning/NativeRdfProjection.md— the native RDF projection: NeoWiki's own vocabulary, plus the per-store infrastructure (IRIs, per-page named graphs, sync) shared by all projections.Post here; substantive input from other channels (chats, calls, email) gets distilled into this thread with attribution. The doc remains the authoritative, updated statement of where the design stands — this thread is where it gets shaped. Closed when the doc graduates to an ADR.
Converted from issue #723: earlier comments below reference the Q-numbering and framing now maintained in the doc's Open Questions (the previous body is in this post's edit history).
All reactions