Think like a Graph

I like graphs. You won’t need to dig too deeply into your language education to diagram the simple preceding sentence: subject, verb, object. Speaking is an expression of thinking. If you are a student of the psychologist Lev Vygotsky, you may believe our thinking is shaped by our language. Whether thought shapes language, language shapes thought, or both, we are in the midst of an artificial intelligence zeitgeist attempting to reflect, or replicate, human thinking in machines.?

Taxonomy and ontology practitioners who create models to reflect thought do so using foundational technologies and standards like the W3C standard Resource Description Framework (RDF ). These standards support the creation and interoperable exchange of graphs. Graphs are the expression of thinking codified in technology and can be a major shift in perspective for organizations.

Triples and Graphs

RDF allows for the naming of subjects, objects, and the relationships between them. These triples are effectively a sentence (in SVO languages, of which there are many): subject, predicate, object. Taxonomists defining concepts as subjects and objects and creating named relationships between them, all defined by the rules of an ontology, are essentially writing sentences expressing the truths of the business. They are building domain models, one simple sentence at a time. Eventually, all of these interconnected sentences become the book of knowledge about your organization, its assets, its processes, etc.

The great thing about a graph is its flexibility. It is possible to add, subtract, and change subjects, objects, and predicates (verbs) because each object in the graph has an immutable URI. If on Monday the organization decides it wants to call it the “Resource Description Framework” and has standards body “World Wide Web Consortium” and then on Tuesday call it the “RDF” has standards organization “W3C”, that’s possible without breaking the graph. You may break your users, but that’s a governance best practices issue. Not only can you change the object names, you can also change their relationships to each other. The concept “Resource Description Framework” can change from being the preferred label to the alternative label and vice-versa for the concept “RDF”.

Think like a Graph

Hierarchical thinking is pretty common in a lot of companies. Organizational charts are hierarchical as are navigational menus. Changing the thinking from strictly hierarchical to graph can be very challenging. Even more challenging is getting buy-in for a graph-based semantic layer and keeping this layer an unbiased source of truth across multiple domains, taxonomies, and consuming systems and use cases. In particular, there are significant risks involved in using taxonomies and ontologies as foundational data for artificial intelligence. While ontologies represent the business in relatively simple human thought, the complexity of graph structures and the unintentional ability to introduce bias is as complicated as the thoughts that graph was built to represent.

On October 23rd, I’ll be tackling some of these issues in my presentation Layers, Multi-Purpose Taxonomies, and the Ghost in the Machine at the Henry Stewart Conference “Semantic Data 2024: Taxonomy, Ontology, and Knowledge Graphs ” in New York City.

Consider attending this event and use the code Ahren100 to receive $100 off the registration price when you register using this link: https://na.eventscloud.com/ereg/newreg.php?eventid=788866& .