Back to Clarity

COGNISWITCH ARCHITECTURE SERIES • VOL. 1

Structural Intelligence

A comparative visualization demonstrating the shift from hierarchical categorization to semantic connectivity in enterprise knowledge graphs.

Taxonomy

Hierarchical • Rigid • Categorical

Folders & Files paradigm. Items exist in only one location. Relationships are strictly parent-child.

Ontology

Semantic • Connected • Multi-dimensional

RELATIONSHIPS

can lead to
risk factor for
monitored by
diagnosed by
commonly co-occurs with

From Text to Truth: The Grounding Process

See how unstructured enterprise data (like a PDF or email) is ingested and “grounded” into the Ontology to create structure and context.

CLINICAL_PROTOCOL_2024.pdf
UNSTRUCTURED

Standard Care Protocol: Metabolic Syndrome

ID: CP-MET-09 • REV: 4.2 • DEPT: INTERNAL MED

Clinical Overview:
Patients presenting with uncontrolled Type 2 Diabetes frequently exhibit comorbid Hypertension. This combination creates a compound risk profile that requires integrated management.

Risk Progression:
Long-term unmanaged hypertension is a primary precursor that can lead to Heart Failure. Early intervention is critical to prevent structural cardiac changes.

Diagnostic Action Plan

  • Monitor HbA1c levels quarterly to assess glycemic control.
  • Regularly screen for Obesity markers (BMI > 30).
  • If patient reports palpitations, order ECG to screen for potential Atrial Fibrillation.
KNOWLEDGE_GRAPH.json
STRUCTURED

RELATIONSHIPS

can lead to
risk factor for
monitored by
diagnosed by
commonly co-occurs with

HOVER OVER TEXT TO SEE CONNECTIONS

What's an Ontology? (And Why Should You Care?)

Everyone's suddenly talking about “context graphs” and “semantic layers.” Underneath all that hype? An idea from the 1990s called an ontology.

The Confusion Problem

You're in a meeting. Someone says “we need better knowledge management.” Another person mentions “context graphs.” A third drops “semantic layer.” A vendor promises “ontology-driven AI.”

Are these the same thing? Different? Complementary?

Nobody in the room is quite sure. But everyone nods. Let's fix that.

Start With What You Already Know: The Taxonomy

A taxonomy is a classification system. It puts things into boxes.

Example: The Library

  • 600s → Technology
  • 610s → Medicine
  • 616 → Diseases
  • 616.1 → Cardiovascular diseases

Clean. Hierarchical. A tree structure where everything has one parent.

In healthcare, ICD codes work the same way:

  • Chapter IX — Diseases of the circulatory system
  • I10-I16 — Hypertensive diseases
  • I10 — Essential hypertension

Taxonomies answer one question: “What category does this belong to?”

Now Add Relationships: The Ontology

An ontology takes the taxonomy and adds connections between concepts.

Back to the library. The Dewey Decimal System tells you where the book lives on the shelf. But a library catalog is richer:

    Catalog Metadata
  • → This book is written by Author X
  • → This author also wrote these other books
  • → This book is related to that book
  • → This edition supersedes the previous edition
  • → This book is required reading for Course Y

Same books. But now you can navigate between them in meaningful ways.

In Healthcare: SNOMED CT

It doesn't just classify “Type 2 Diabetes” — it captures:

is a
Metabolic disease
treated by
Metformin, Insulin
commonly co-occurs with
Hypertension, Obesity
requires monitoring via
HbA1c test

Ontologies answer: “How is this thing connected to other things?”

The E-Commerce Example

You shop on Amazon. Here's taxonomy vs. ontology in action:

Taxonomy (Category Tree)

Electronics
└─ Computers
   └─ Laptops
      └─ Gaming Laptops

Helps you browse.

Ontology (Relationship Graph)

  • →This laptop is compatible with these accessories
  • →Customers who bought this also bought that keyboard
  • →This model is the successor to last year's model

Makes recommendations actually useful.

Why This Matters for AI

Large Language Models (LLMs) like GPT-4 or Gemini are trained on text. They learn patterns in language. But they don't inherently understand relationships between concepts in your specific domain.

Without Ontology

“Can a patient on Warfarin take Aspirin?”

The LLM guesses based on probability patterns. Maybe it gets it right. Maybe it hallucinates.

Grounding

With Ontology

  • Fact: Warfarin is_a Blood Thinner
  • Fact: Aspirin is_a Blood Thinner
  • Rule: Blood Thinner contraindicates Blood Thinner

The ontology gives the AI structured reasoning paths.