Re: KR 101 - Spatial Ontology Baseline – conceptual alignment and pointers

I just attended a webinar of the ITU AI for Good (AI4G) series of webinars
where Rose Yu did an excellent job at explaining how she at UC San Diego
Department of Computer Science and Engineering  is developing foundation
models in an AI4G session on "Grounding foundation models for automated
climate science".

State-of-the-art work that has direct bearing on how we are trying to
reconcile different concept ontologies for KR.

https://www.youtube.com/live/7hvxUVDrDec?si=A_J1sGm06CgSg_WS

Also visit www.roseyu.com.

Milton Ponson
Rainbow Warriors Core Foundation
CIAMSD Institute-ICT4D Program
+2977459312
PO Box 1154, Oranjestad
Aruba, Dutch Caribbean

On Sun, Nov 16, 2025, 12:18 Daniel Ramos <capitain_jack@yahoo.com> wrote:

> Paola, all,
>
> Thank you for the pointer to Bateman & Farrar’s Spatial Ontology Baseline
> (SFB/TR8 D2). I’ve gone through it and here is a compact conceptual
> synthesis plus some pointers to open vocabularies/standards, and how this
> connects to what I’m doing with spatial KR in K3D.
>
> 1. What D2 is doing conceptually
>
> D2 sits in a sequence (D1–D4) and focuses specifically on space as an
> ontological category. In my reading, the key points are:
>
> Space as an ontological category
> Space is treated explicitly, not just as coordinates: we need to
> characterize spatial entities (regions, objects, features), their parts,
> boundaries, and the relations between them.
>
> Foundational tools: mereology, topology, geometry, mereotopology
> The report reviews how parthood, connection, boundary, overlap, interior,
> etc. can be axiomatized and combined:
>
> mereology → part/whole;
> topology → continuity, connectedness;
> mereotopology → parthood + connection (e.g., Cohn/Lehmann);
> geometry → metric and shape.
> Region‑based vs object‑based representations
> A central theme is region‑based calculi (e.g., RCC) vs object‑based
> accounts, and how region connection, contact, overlap, containment and
> vagueness (egg‑and‑yolk examples) should be represented.
>
> Survey across upper/domain ontologies
> It walks through spatial aspects of major ontologies:
>
> SUMO‑space: SpatialThing, Region, SelfConnectedObject, ShapeAttribute,
> SpatialRelation taxonomies.
> OpenCyc‑space: spatial parts, spatial relations, path/path systems.
> DOLCE‑space: physical objects vs features, spatial dependence, qualities.
> BFO‑space: SNAP/SPAN, SpatialRegion, Site, niches, layers, granularity.
> It then connects these to qualitative spatial representation, GIS,
> cognitive semantics and spatial language.
> Conclusions and recommendations
> The final sections stress:
>
> the importance of observer viewpoint;
> bridging formal spatial ontologies with spatial language and GIS;
> modular foundations (upper ontologies) and reusable submodules;
> and the need for a generic framework with clear categories and mapping
> between layers.
> So, to paraphrase: D2 is about how to carve up space in ontology using
> region‑based and object‑based tools, how existing upper ontologies (SUMO,
> DOLCE, BFO, etc.) do it, and how that should inform practical modeling.
>
> 2. Pointers to open vocabularies and standards (spatial KR)
>
> D2 itself points to several “upper” and spatial ontologies that are still
> relevant:
>
> SUMO (Suggested Upper Merged Ontology) – open upper ontology with spatial
> categories (SpatialThing, Region, ShapeAttribute, SpatialRelation).
> DOLCE / DOLCE+ – foundational ontology from LOA, with clear treatment of
> physical objects, features, and spatial dependence.
> BFO (Basic Formal Ontology) – widely used upper ontology (SNAP/SPAN,
> SpatialRegion, Site, niches).
> Region Connection Calculus (RCC) – qualitative spatial calculus for
> regions and their relations (RCC‑5, RCC‑8).
> GIS & OGC/ISO standards – e.g. ISO 19107 for geometry/topology; GeoSPARQL
> for spatial relations in RDF; OWL Time for temporal aspects.
> On the web/AI side, I would add:
>
> GeoSPARQL: OGC standard for geographical features and topological
> relations in RDF.
> OWL Time: temporal ontology often combined with spatial KR.
> PROV‑O: for provenance of spatial/temporal artifacts.
> These provide the “upper” and mid‑level vocabularies into which more
> specialized domain ontologies can be grounded.
>
> 3. How this connects to what I’m doing (and where K3D sits)
>
> K3D is not trying to replace any of that; it is trying to instantiate it
> in a spatial, inspectable substrate:
>
> Houses and Rooms as domains of discourse
> Each House is a bounded domain (in the sense Milton, and D2’s “domain
> ontologies,” use the term); Rooms partition that domain further.
> Conceptually, these can be grounded in upper ontologies like DOLCE/BFO
> (e.g., PhysicalRegion, Site, SpatialRegion) and specialized out with
> SUMO/DOLCE/BFO categories as needed.
>
> Nodes and Galaxies as spatial representations of concepts and relations
> Nodes are the things we place in space; Galaxies are the embedding spaces
> where distances encode similarity, but the categories (object vs region,
> path, feature, etc.) can be taken from or aligned with the ontologies D2
> surveys.
>
> Garden and Museum as structuring devices
> The Knowledge Garden is a fractal, tree‑like visualization of ontologies
> (roots/branches/leaves) that can be populated with DOLCE/BFO/SUMO
> categories; the Museum uses “portal cubes” to hold large, archived
> ontologies or GIS‑like datasets consistent with the layered approach D2
> advocates.
>
> I’ve documented the visual/structural encoding here:
>
> docs/SPATIAL_KR_VISUAL_ENCODING.md – how domains, concepts, relations,
> modalities and time are encoded as shapes, rays, trees and cubes in K3D’s
> Galaxy, Garden and Museum, independent of any particular GPU backend.
>
> In other words, I see D2 as a conceptual baseline for how we talk about
> space in ontologies (upper and domain), and K3D as one concrete way to host
> and visualize those ontologies (and their relations, and their
> temporal/adequacy status) in a shared 3D substrate for humans and AI.
>
> I agree entirely that domain ontologies must be grounded in upper
> ontologies; I’m not arguing against that. My contribution is to provide a
> spatial KR framework where that grounding and those relations become
> visible and navigable, while still being compatible with the ontological
> work D2 summarizes.
>
> Best regards,
> Daniel
>