My Deal

by bauer on November 11, 2021


“Blank nodes can’t be predicates.” - WWW REFERENCE

“Every perspective exists only in interdependence with something else.” - Nagarjuna

“Every book its reader.” - S. R. Ranganathan

“Shush!” - Madame Jocasta Nu, Chief Librarian, Jedi Archives

What is Ontology?

If you ask a conscientious practitioner in the field “What is Ontology?” - you’ll get a thoughtful answer.

“Ontology creates more robust, trustworthy definitions of concepts - those relevant to some particular aspect of an overall problem - using labels, categories, descriptions, references, properties and values, arranged in a categorical hierarchy and published in a form consumable by both humans and computers around the world which together facilitate the creation of better solutions to that problem.”

It’s hard to argue with such a robust definition for Ontology; particularly as this is manifested in the Semantic Web itself. It’s also easy to overlook this as a self-fulfilling definition; one updated to reflect the transmogrification of Ontology from its original conceptualization to one aligned with the discipline formerly known as Library Science. Going back to the future to its original conceptualization, Ontology can have a shorter (admittedly more ethereal) definition.

“Ontology determines existence.”

Taking some liberty but staying true to its Aristolelianistic roots, Ontology, at its core, is a discipline dedicated to answering one abiding question independent of topic:


Does some “thing” exist or not? Is there a multiverse? Is this a new species? Is they a qualified applicant? Sufficiently answering question of existence requires reflection upon the question itself. Better answers require more articulate questions.

As Nagarjuna observes, nothing exists absolutely, but only in relationship to something else, so, to start, all questions should be posed as questions of relative existence. Is there a multiverse that contains our universe? Can this bug be classified as a new species of a genus? Can this person’s skills be used for this job?

Upon further reflection you find the answers depend not just on the specific relationships but upon the types of relationships themselves. The questions themselves can be classified. Is this an instance of that class? - Categorical. Is this a part of that whole? - Compositional. Is this used for/by/with that? - Utilitarian. It is the characteristics of the types of relationships that best answers questions of relative existence. Ultimately all “meaning” is determined by relationships.

So, What?

At the end of the day we do want to produce robust, trustworthy answers - definitions, if you will, so you could argue we will end up comfortably next to the conscientious practitioner anyway. The contention is that you don’t get the richest, deepest, most useful representation without first defining a more robust relationship model. Moreover, by emphasizing only categorical relationships, an inordinate amount of time is spent around the publication details of what is best an approximation.

Now we can spend a lot of time exploring the depth of the differences between these two definitions, starting in Ancient Greece, passing through Imperial Rome, diverting to Enlightenment Europe, and even connecting through 1930’s Ubayavedhanthapuram. It’s more efficient to just start in the middle of the 20th century and start a (proxy) war.

Semantic Web* or Property Graphs?

Starting from first principles, let’s question the question - what kind of question is this?

If you’re a Taxonomist, you’d categorize this as basically a technical question. Something along the lines of which of these frameworks, in what context, given which constraints, facilitates the articulation of which terms, to help solve which customer problems; thereby exceeding which measurable goals, in what timeframe, in order to get you a new jet ski. All well and good.

If you’re an Ontologist, you might categorize this as an existential question, noting that the canonical answer invariably assumes its postulation as a disjunction - it’s one OR the other. Why isn’t the question posed as a conjunction - Semantic Web AND Property Graphs? It’s only then you might appreciate the potentially co-dependent, if not coordinated, forces at work, prejudicing the answer towards one choice in particular.

If you’re an Acolyte, say a proselyte of an orthodoxy stratifying the hierarchies of Academic and Standards bodies, you might consider this question as, in fact, heretical; a question, which shouldn’t be asked in the first place. To the veritable zealot, there are no Closed Worlds, only Open; there is only one Framework, one Way, one Truth, one Light - the Light Side - and that’s the Semantic Web.

To quote Madame Jocasta Nu: “if an item does not appear” in the Jedi Library (ne Archive) then “It does not exist!”

To which the ecclesiastic would add the corollary:

If it ain’t a Jedi Library, it ain’t a Library.

To them, it’s essentially sacrilege to ask this question. This orthodoxy arose from the commensalism derived from the discipline now named Library and Information Sciences and its absorption of the technology that is the Semantic Web. How’d this happen?

Let’s scuttle down memory lane.


Which came first?

A lot of people believe that the Semantic Web came first, springing fully formed, as an Athena, from the mind of ne Tim Berners-Lee, now, Sir Tim Berners-Lee, Knight Commander of the Most Excellent Order of the British Empire, whereas Property Graphs were invented by a couple of Swedes in a basement in Stockholm.

Despite repeated denials, reality did not begin with the introduction of Microsoft Explorer in 1998. Instead of wandering around the planet, we’ll use a geographic proxy to explore the origins of our proxy war. And as have so many other things been resolved, we’ll do so in the streets of Pittsburgh, let’s say Craig Street, halfway between two Pittsburghs, specifically two Pittsburgh Universities.

The Nobel Prize Winners

We start in the 1950’s at Carnegie Mellon University, where the framework for Property Graphs was laid down by a couple of guys in a cover band named Simon and Newell on a track called “Schema Representation.” Fundamental to Schema Representation are the Relationships, starting with the Instance and Is-A Classification Relationships, which together support Inheritance, and encompassing many other types of Relationships including Composition (Part-Of, Parts) and Utilization (Used-By, Uses) as well Temporal (Before, During, After), and Positional (On-Site, Remote).

By the 80’s “Schema (aka Knowledge) Representation” was manifested in the first commercial AI companies for building Expert Systems for solving planning, scheduling, diagnosing, and provisioning problems. This rich representation of Relationships provided the conceptual framework for an Ontologist to develop Ontologies that not only informed but were imbued within these Intelligent Systems. The Ontologies were part of the models, not a reference for the model. This synergy came from the creativity of both human and machine.

The Librarians

Meanwhile, just cross Panther Hollow, while waiting on the first Vax Computer for the University of Pittsburgh, the Library and Information Sciences department there and across the country received the sacrament of Broader/Narrower for Classification in 1985 to complement Subject Headings and Thesauri.

Agreement was reached on the necessity to define semantics for the seminal type of relationship, Classification. And the semantics were defined and articulated very well. Excellent. After that, though, it got pretty fuzzy. Consensus was reached on a set of catch-all matching relationships, namely Exact, Close, and the ultimate catch basin Related.

One type of Relationship to rule them all? Layered on top of the innovative use of syndetic from the 1940’s by the Library of Congress, this became enough to finally exorcize the uncontrolled from within hallowed halls. Controlled Vocabularies and Thesauri, where authorized terms, those chosen by a Curia, compliant with stricture, could finally be ordained.

If only there was some computer-sciency thing to dump all these print documents into.

The White Knight

In the mid-90’s Sir Tim re-conceptualized the World Wide into the Semantic Web. The stage was set for a little syncretic syndetic synergy. Categorical representational Standards framework? Meet categorical representational Academic discipline. Didn’t take long to tell these two kids to get a room and engage in a little commensal co-evolution.

As the Fig Leaf, so the Fig Wasp.

The adolescent technology seeking advocacy met a senescent discipline needing currency. As with all May-December relationships, all that was needed was a little compromise. Despite pleading with Sir Tim to treat Relationships as first class citizens, to actualize the Semantic part of the Semantic Web, the call was made to forgo this most essential aspect of implementing a knowledge representation framework. Meaning - derived from relationships - was deprived of the Semantic Web from inception, blithely accommodated by the discipline with its facelift - the Library and Information Sciences.

Blank Nodes can’t be Predicates.

C’mon! So What?

Am sure the magniloquence of this article alone is enough to put off people to it. Liberties taken with events surely put off others; insinuations enough for a few to call for the excommunication of the author. Most critical would be a challenge from the Machine Learning community stating that the reference library that is the Semantic Web is sufficient to current purposes, that any need for more expressive representation is obviated by today’s neural modeling, that the sophistication of natural language processing technologies is sufficient, that no real value can be added from more structured grammars enabled by conversational modeling through deeper graph representation.

Or to put it more simply: “Dude. Expert Systems are dead.”

All of this may be true but there’s one over-riding truth to add to all of this that’s undeniable.

An entire generation has been graduated as Ontologists but trained as Librarians. The bulk of their training has been in controlling the publication process rather than on actually publishing. Far more time spent on managing sharing rather than on having something to actually share. Authentication and authorization has taken precedence over discovery and development.

It ain’t Jedi. It’s Sith.

The Take-Away

The answer to “Semantic Web OR Property Graphs” is “Property Graphs AND Semantic Web”.

The best way forward for developing meaningful knowledge solutions is to start with the richest knowledge representation framework, Property Graphs, and develop your knowledge base there, publishing through a Semantic Web representation as needed.

Training the next generation as Ontologists is left as an exercise to the Professors.

{ 0 comments… add one now }

There are no comments yet...

Kick things off by filling out the form below ↓