When Do You Need a Taxonomy vs. an Ontology? Insights from Marjorie Hlava on Transforming Knowledge Management

In the evolving landscape of knowledge management—especially in the age of artificial intelligence (AI)—organizations are under increasing pressure to structure, retrieve and leverage their data more effectively. But a common question arises: Do you need a taxonomy, an ontology, or both?

In a recent KMWorld article, Marjorie Hlava, Chief Science Officer of Access Innovations and a leading voice in the field of information science, provided much-needed clarity on this topic. The piece gathered perspectives from four experts, each shedding light on the foundational frameworks that help organizations transform raw information into actionable knowledge.

According to Hlava, taxonomies are ideal when the primary goal is organizing data for human use. These hierarchical structures classify information into categories and subcategories, making it easier to navigate, find and apply content in a meaningful way.

“Taxonomies are sufficient if an organization is trying to organize data for knowledge management for other people,” Hlava explained. In other words, when your focus is on labeling and grouping data for content discovery, training materials, or internal knowledge portals, a well-structured taxonomy often does the job.

However, when your organization needs to move beyond basic categorization—to uncover and connect deeper relationships between entities—an ontology becomes essential.

Ontologies, Hlava noted, are designed to model complex relationships between concepts and their attributes. They create a structured framework that allows machines and humans alike to interpret the meaning and context of data. “It behooves organizations to implement ontologies in standards-based environments such as RDF and semantic knowledge graphs,” she said. Technologies like OWL (Web Ontology Language) make it possible to enrich these frameworks further, providing the backbone for smart applications, recommendation engines, and data interoperability.

This distinction becomes especially important in AI-powered environments where data-driven automation, recommendation systems, or decision support systems require nuanced understanding of how different concepts relate to each other.

One of the most valuable takeaways from Hlava’s insights is the idea that taxonomies and ontologies are not mutually exclusive. There’s a gradient between taxonomies, thesauri and ontologies, with considerable overlap in their application.

For instance, a controlled vocabulary or domain-specific thesaurus can serve as a stepping stone for organizations just beginning their journey toward semantic enrichment. These resources can evolve over time—first supporting traditional classification, and then expanding into more complex ontological structures.

Additionally, OWL isn’t just for academic or technical audiences. As Hlava pointed out, it’s especially useful for fleshing out conceptual data models, defining schema, and supporting triple stores used in many advanced query engines. In this way, ontologies become foundational tools for implementing AI and semantic search strategies.

For those looking to dive deeper into the nuances of taxonomies and their role in knowledge management, Hlava’s Taxobook series offers a comprehensive, three-volume exploration of taxonomy theory, practice, and application. It’s an essential resource for information professionals, data scientists and anyone tasked with making sense of ever-expanding information assets.

As knowledge ecosystems become increasingly intelligent and interconnected, understanding when to deploy a taxonomy vs. an ontology—or how to blend the two—will be key to unlocking the true value of your information assets.

Let your data speak more clearly. Whether for people or for machines, the right structure makes all the difference.

Melody K. Smith

Data Harmony is an award-winning semantic suite that leverages explainable AI.

Sponsored by Data Harmony, harmonizing knowledge for a better search experience.