Notes on Upper Ontology Facets

Much disagreement surrounds upper ontologies.  This tries to bypass it using several simple facets, which major ontologists often use as discriminants in larger classification plans.

If you have questions on what is meant, or suggestion on changes, please write me.  And certainly, add extensions and/or comments to your namespace and documentation at the next level down.  These facet models are meant only for gross, top-level classifications. 

---

NATURE

This table show what the first two discriminants mean to me, but concepts from programming  languages muddy the ontological waters.  I trust the discriminants and the phrases in the cells more than the bolded upper case words, but some concise label seems needed, so I add them with this warning -- the meaning of "class" in OWL seems more like "set", and in Javascript it gets replaced with "prototype" - hence the shading of colors.

	Definite (specific)	Indefinite (typical)
Individual (singular)	INSTANCE One particular  example	PROTOTYPE One typical example of a class or set
Collective (multiple)	SET A number of  specific examples	CLASS All examples close enough to count

Each thing described above could also be either Real or Imaginary to a SAID publisher.  If flagged as the latter, the actual existence of the thing seems questionable or subjective at best, and may often be entirely fictional, such as a character or plot from a story.

---

REALM

The "universe" in which a thing exists is fairly easier to categorize, as simple tests exist.  If it has a location in space, it is Concrete, else it is Information.(always recorded in some concrete media, yet distinct).  If it has a human author/designer, it is Artificial, else it is Natural.  As a rule of thumb, if something existed 50,000 years ago, it is the latter. 

	Concrete	Information
Natural	PHYSICAL Spacetime, Matter, Organisms, Landmarks	SIGNALLING Senses, Genes, Instinct, Emotions, Desires
Artificial	SOCIETAL Materials, Transport, Infrastructure, Cities	DESCRIPTIVE Models, Morals, Language, Law, Art, Ownership

Within each realm exist things of Basic types - Composites and their Characteristics.  These fit O-O modeling paradigms, and also fit naturally into triples languages, so most technical readers will find them familiar.  Each realm may also hold Mediating types, which are disjoint but related, as the next section outlines.

---

MEDIATION

Each related type is effectively a design pattern.  It extends a basic type by relating it to other basic types using a triples structure, then imposes related, mediating constraints.  The pattern types below add considerable expressivity (like macros), without violating any basic modeling-language assumptions:

Association - a typed N-ary relation, optionally limited by imposing scopes
Iff a Composite is marked as mediating, it behaves and validates not as that Composite, but as an Association in which that composite plays a central role.   This pattern does a good job expanding it into a whole-part model.

Scope - an  Topic model saying when a Characteristic applies
Iff a Characteristic is marked as mediating, it behaves and validates not as that Characteristic, but as a Scoping topic able to declare (with rules) that the Characteristic is unknown and/or inapplicable to a given query.  This pattern can effectively make parts and properties conditional

Mediating types in PSAID interpretations have hyphenated realm-type modifiers.  On the index page, they are separately searched using the realm option with the '+' sign.  Logically, each implies that a comparable basic type must also exist, but (so far) no related checks or auto-creation code is supported here.

---

TYPE

This is a classic type tree, as shallow and binary as I could arrange.  It really splits Basic types into two disjoint trees, depending (respectively) on whether they refer to aspects of other things or can stand independently as topics in their own right.  NOUNS can name both these types, often leading to confusion.  These test may help

Characteristics of things exist ONLY as parts or properties of something else.  Different sorts are fairly intuitive, and easiest to subtype by the nature of their descriptive values.  In OWL and RDF-Schema all are sub-types of Property.  Topic Maps use the same split with a different terminology, and do a much better job modeling names (adopted here).

Composites divide up along other lines, based on persistence.  Entities are treated as long lived, and divide further into count nouns or mass nouns - under simple linguistic tests.  Their association within individual Occurrents, however, is modeled as transient, and even in ongoing processes, specific participants may come and go.

Many edge cases exist, but by using the other facets of each SAID, they usually can be clarified within each realm.  Here are some modeling suggestions on handling tough cases:

Collectives such as herds or juries work out cleanly as the Object playing the whole in an Occurrent whole-part association among its member parts,  which may be modeled as sets

A Substance (a weird special case) has normal instances only in some region, quantity or other countable Object composedOf it.  Sets of substances typically collect their subtypes

If you have other suggestions to contribute, please publish them and/or let me know.  Thanks: