Voyado Elevate

Search

Understanding ontology in Elevate

To deliver accurate search results and relevant recommendations, Elevate needs to truly understand your products. That understanding is powered by ontology, a structured, retail-specific knowledge model that goes far beyond keyword matching.

This article explains what ontology is, the problems it solves, how it works, and why it is a key differentiator in Elevate.

The problem with keyword matching

Without a deeper understanding of products, search engines rely heavily on exact keyword matching. This can lead to:

  • Missed results when customer wording differs from your product feed
    For example, different word forms (such as jacket vs jackets) or different words for the same concept (such as pants vs trousers) may not match, even though they refer to the same product
  • Poor result precision, with loosely related products appearing too often
    For example, a search for dress might return dress shirts, even though the shopper is likely looking for dresses, not shirts.
  • Heavy reliance on manual tagging and synonym management

For example, if a customer searches for pants and your product data uses the term trousers, a keyword-based system may fail to return relevant results.

Ontology solves this by understanding that pants and trousers refer to the same product type, allowing Elevate to return the correct products even when the wording differs, while also distinguishing between closely related but different concepts.

What is ontology?

In computer science, ontology is a formal, explicit specification of a shared conceptualization - a structured framework defining concepts, properties, and relationships within a specific domain. In Elevate - the ontology is the structured knowledge model built specifically for retail. It defines product types, attributes, relationships, and concepts in a way that reflects how humans understand products.

Instead of treating product data as plain text, ontology places each product into a network of connected concepts. For example:

  • A jacket can be linked to the product type “jacket”, the occasion “outdoor”, and the activity “hiking”
  • A lamp can be linked to the room “bedroom”, the usage “indoor”, and the product type “table lamp”
  • A sweater can be linked to the material “wool”, the season “winter”, and related products like wool detergent

This allows Elevate to match customer intent beyond exact keyword matching.

Ontology vs. synonyms

It is important to distinguish between ontology and traditional synonyms.

Synonyms are typically hard-coded word replacements. For example, jeans equals denim. Synonym lists grow over time and become difficult to maintain, especially across languages and markets. Ontology goes deeper. It does not just replace words. It models conceptual understanding. For example:

  • Denim can be understood as a material.
  • Jeans can be understood as a product type.
  • A denim jacket can be matched correctly even if the customer searches using different wording.

In the Elevate application, you can still define manual synonyms for specific business needs. However, ontology provides a broader and more structured foundation for understanding products.

How it works

When you send product data to Elevate, it is analyzed and enriched before being connected to the ontology. This involves several parts, including product analysis, classification, and the mapping of concepts and relationships.

Product analysis

Using AI and large language models (LLMs), Elevate analyses the product feed and interprets what the product actually is, not just what text appears in the feed. Many search solutions rely primarily on keyword matching or generic AI models. Elevate combines AI with a retail-specific ontology, purpose-built for commerce use cases.

Classification

The product is mapped to the correct product type, properties, and concepts in ontology. For example, a lamp may be classified as:

  • Product type: table lamp
  • Color: yellow
  • Power source: electric
  • Room association: living room

Relationships

The relationship layer is what allows Elevate to understand how different concepts connect to each other within the product catalog.

Instead of treating each attribute as isolated, the ontology defines how concepts relate in a hierarchy.

For example:

  • A night lamp is understood as a type of lamp, and will be found when searching for “lamp”
  • A product made of aluminium can be matched when searching for metal
  • A running jacket can be understood as a type of jacket

This allows Elevate to match broader or more general searches with more specific products, ensuring that relevant results are returned even when the wording differs.

Benefits of using a retail-specific ontology

The Elevate ontology is built specifically for retail. This means it is designed to understand products the way shoppers and merchandisers do — not just as data, but as meaningful concepts.

Because of this:

  • It understands retail-specific concepts such as product types, materials, usage, and seasons
  • It connects products through shared concepts, making it easier to match search intent and support discovery
  • It enables more relevant recommendations by understanding how products relate to each other in real-world contexts

Because the ontology is tailored for retail, Elevate can apply AI in a focused and efficient way. Instead of relying on generic AI to interpret everything from scratch, Elevate uses AI to analyze product data and place it into a structured, retail-specific framework.

How the ontology impacts search and recommendations

Ontology supports both search and recommendations, but in slightly different ways.

For search, ontology helps Elevate understand what a customer is looking for. It connects different words and concepts to the correct product types and attributes, so relevant products can be found even when the wording differs.

For recommendations, ontology helps Elevate understand how products relate to each other. It connects products through shared properties, usage, or context, making it possible to suggest complementary or related items.

Ontology supports several key capabilities across Elevate, enabling both search and product recommendations to work effectively.

For search, ontology helps match analyzed queries to the correct product types and properties, even when customer wording differs from the product feed. This ensures that shoppers can find relevant products without needing to use exact terms.

For facets and filters, ontology groups attributes in a logical way. This makes filtering clearer and easier to use.

For recommendations, ontology connects products through shared properties and concepts, allowing Elevate to suggest relevant complementary or alternative products.

For personalization, ontology links products to broader concepts such as occasions or usage, supporting more contextual and tailored shopping journeys.

What you can control

The ontology itself is maintained and continuously developed by Voyado as part of Elevate’s core intelligence. This ensures that product understanding stays accurate, consistent, and up to date across all customers and markets.

This also means that you do not need to manually manage how products are classified or connected. Elevate handles this automatically, using its product intelligence to interpret and structure your data in a way that would be difficult to maintain manually at scale.

As a retailer, your role is to guide and refine how Elevate works for your specific business needs. You can influence the outcome by :

 

Article last updated