[x3d-public] X3D Semantic Web Working Group minutes, 17 JAN 2019: structural and conceptual semantics

[Brutzman, Donald (Don) (CIV)]

[corrected title - please ignore prior version]

===========================================================================
Minutes of the 17JAN 2019 teleconference for X3D Semantic Web Working Group

Participants 2 (Athanasios and Jakub)

We think we have to specify the difference between structural and conceptual semantic objects, predicates or properties?

If something is structural or conceptual has to do what it describes. If an object or  predicate  represents a relation obviously produced by the geometry or 3D structure, things that the a “machine” use to deal with, it might considered as structural. On the other hand if the objects describes something closer to the human mind, to the way that the human understands and describes the structure then it might be considered as conceptual.

In general we consider conceptual semantic as domain specific.

The question is, if there is a possibility to find out a basic common set of conceptual descriptors objects, predicates and properties that may be considered cross-domain. Some basic objects that might be essential parts of most of the domain specific ontologies.

Is it possible to find a minimum set of semantic objects and predicates that can be used to describe the way that the human mind understands and interprets the scenes of an X3D no matter if it is a Humanoid model or a architectural design or a biological structure.
We have to consider that we deal with is a Graphic Scene with vertices, faces, textures, animations…..

So is there any minimum set of descriptors (objects, predicates and properties) that may give some kind of human mind centric basic conceptual description? Like- how the models interact to each other, is there any abstract characterisation of the scene or models, ……

Of course these minimum objects may be extended further by domain specific objects….

According to the description above, structural semantics could be understood as ontological (RDF, RDFS, OWL and rule-based) description of 3D content that is an equivalent to existing 3D formats and languages, e.g., X3D, FBX, XML3D, etc. Of course, such structural descriptions are 3D- and animation- oriented, and they describe the content in a form that enables its visualization by 3D browsers. A good example is the 3D Ontology (http://3dontology.org/) or sth. similar described a few years ago in one of our papers (https://annals-csis.org/proceedings/2013/pliks/416.pdf). Take a look at example listings - these are semantic counterparts to X3D.

Conceptual semantics, which is considered understandable to average users or experts in different domains, in general may cover very different classes, properties and objects that we cannot address or predict in advance. However, some common classes, properties and objects could be probably related to general features of objects that may affect their 3D geometry, structure, presentation and animation. I mean of such general features as physical state (solid, liquid, gas), position and orientation (on the left/right side, behind, in front of), structure (single or composed object), interactivity, behavior (walking, running, flying object). We have included some of such very general concepts in papers (http://semantic3d.org/wp-content/uploads/2017/03/BIS2013-Flotynski-Walczak.pdf and https://www.sciencedirect.com/science/article/pii/S1877050914010990). All the other concepts, which are in fact the vast majority, we considered domain-specific and made them extensions of general objects, which are finally mapped to structural semantic objects.

On the other hand, the meaning of conceptual semantics in the context of 3D varies depending on what we want to do with the content and on the particular use cases that may be very different in general. In particular, it may differ for existing 3D content that we semantically describe for querying, searching, exploring and indexing, and new 3D content that we create using conceptual semantics. If we semantically describe existing 3D content, general conceptual terms may be sufficient in some applications (e.g., related to location: left/right side, in front of/behind), but in others they may be different for different domains (e.g., depending on the micro/macro scale of the 3D worlds we describe). If we semantically create new 3D content, conceptual semantics must always be exactly specified using structural descriptions (maybe also semantic). For instance, we must exactly know what it means left/right side - how far away from the primary object, how rotated in relation to it and so on. It must be specified somewhere explicitly or implicitly. Such specification may be different for every domain and application thus cannot be done in advance.

To summarize, I think that we could start with a limited set of general conceptual classes, properties and objects that are common to different domains, and further, provide methods of extending such objects (e.g., by inheritance) to any possible class, property and object, which may be completely different across diverse applications and domains.

===========================================================================

all the best, Don
-- 
Don Brutzman  Naval Postgraduate School, Code USW/Br       brutzman at nps.edu
Watkins 270,  MOVES Institute, Monterey CA 93943-5000 USA   +1.831.656.2149
X3D graphics, virtual worlds, navy robotics http://faculty.nps.edu/brutzman