In the vast universe of e-commerce platforms, product discovery is often reduced to text-matching or crude filters. At Shopiro, we chose a different path. We built CORTEX — our Computational Ontology for Relational Tracking and Explainability — a modular reasoning engine that turns product data into real-time semantic intelligence.

Unlike traditional faceted search or keyword parsing, CORTEX understands what a product is, how it relates to others, and why a buyer might care.

Why CORTEX Exists

Online marketplaces handle millions of listings, each rich in unstructured or loosely structured data. CORTEX's role is to:

  • Extract filterable dimensions from listings
  • Normalize values across international units (e.g. PSI to kPa, inches to mm)
  • Build compatibility rules between products (e.g. “will this battery fit my drone?”)
  • Allow prompt-driven expansion of ontologies using LLMs

All of this occurs behind the scenes, before a user ever sees a filter or suggestion — within milliseconds.

The Brain: Attribute Definitions & Property Types

At the heart of CORTEX are two key constructs:

Attribute Definitions

These describe individual data points — like max_speed_kph, has_bluetooth, or battery_capacity_mah. Each attribute includes:

  • Human-readable label and description
  • Data type (float, string, enum, etc.)
  • SI normalization unit (optional)
  • Validation rules and regexes

User Property Types

These define kinds of products (e.g. "cordless_drill", "quad_core_cpu") and include:

  • Schema of required/optional attributes
  • Parent/child product relationships
  • Match logic (e.g. >=, !=, or tolerance-based compatibility)

How Filters Are Built

Whenever listings are viewed, CORTEX can:

  • Aggregate their ontology_cache fields
  • Infer filter types (enum, float, string)
  • Normalize units when requested
  • Include attribute metadata (label, unit, enum list, etc.)
  • Sort values by popularity or alphabetically

These dynamic filters reflect real listings, not static category trees. And because they are schema-aware, we can prevent filter bloat by collapsing or pruning weak signals.

Compatibility Logic

CORTEX supports parent-child product matching. For instance:

"Is this CPU compatible with the motherboard I bought last week?"

CORTEX checks:

  • Shared attribute codes across types
  • Normalized values and operators (e.g. RAM speed must be >= 2400MHz)
  • Optional explanation output (e.g. "Mismatch: PCIe slot not present")

This logic is abstracted so it can run on any listing-property pair — even user-defined setups.

A Foundation, Not a Feature

CORTEX is not a plugin or a UX layer — it is foundational. Its outputs inform:

  • Search ranking
  • Compatibility badges
  • Filter interfaces
  • Recommendation logic

In short: CORTEX is how Shopiro thinks. Every product, every query, every match — grounded in ontological logic.

https://shopiro.ca: The new smart way to shop online™