Semantic Networks and Ontologies

Semantic networks and ontologies represent knowledge as structured collections of concepts and the relations between them, organizing what is known about a domain into a graph or formal vocabulary.

Definition

A semantic network represents knowledge as a graph whose nodes denote concepts or objects and whose edges denote relations such as is-a and part-of; an ontology is a formal, explicit specification of the concepts, relations, and constraints shared within a domain.

Scope

This topic covers structured (network- and frame-based) knowledge representation: semantic networks of nodes and labeled links, frames and slots, inheritance hierarchies with default values, and the modern notion of an ontology as a shared, formal specification of a domain's concepts and relationships. It addresses how categories, objects, and taxonomies are modeled and how inheritance and identity are handled. The formal logic underlying ontologies is detailed in the description-logics topic, and learned embeddings of entities belong to the machine-learning subfield.

Core questions

• How are concepts, objects, and their relationships represented as nodes and links?
• How does inheritance let subcategories acquire and override the properties of their parents?
• What makes an ontology a shared and reusable specification of a domain?
• How are categories, identity, and part-whole relations modeled consistently?

Key concepts

• semantic network
• nodes and labeled relations
• is-a and part-of relations
• frames and slots
• inheritance and default values
• taxonomy and categories
• ontology
• knowledge graph

Key theories

Inheritance and taxonomic structure

Organizing concepts in is-a hierarchies lets properties be stated once at a general level and inherited by specializations, with subclasses able to add or override default values, giving a compact and intuitive representation of categorical knowledge.

Frames as structured knowledge units

Minsky's frames represent stereotyped situations or objects as bundles of slots with default fillers, modeling how prior expectations structure understanding and providing a precursor to object-oriented and ontological representations.

Ontologies as shared conceptualizations

An ontology makes the concepts and relations of a domain explicit and agreed-upon so that different systems and people can share and reuse knowledge, with design principles emphasizing clarity, coherence, and minimal commitment.

Clinical relevance

Ontologies and semantic networks underpin the Semantic Web, biomedical vocabularies and knowledge graphs, product catalogs, and enterprise knowledge management, providing the shared vocabularies that let heterogeneous systems exchange and reason over data.

History

Semantic networks originated with Quillian's work on semantic memory in the 1960s, and Minsky's 1974 frames memo gave structured representation a major impetus. Brachman's critiques of informal links led toward formal description logics, and Gruber's 1990s work helped establish ontologies as engineered artifacts, culminating in Semantic Web standards.

Key figures

• Marvin Minsky
• Ross Quillian
• Ronald J. Brachman
• Thomas R. Gruber
• John F. Sowa

Seminal works

• minsky1974
• gruber1995
• brachman2004

Frequently asked questions

What is the difference between a semantic network and an ontology?

A semantic network is a graphical representation of concepts and relations, historically often informal. An ontology is a formal, explicit specification of a shared conceptualization, with defined semantics and constraints, that is meant to be reused across applications; modern ontologies typically have a logical foundation.

How does inheritance work in these representations?

Concepts are arranged in is-a hierarchies so that a specialized concept inherits the properties of its more general parents. Defaults can be overridden at lower levels, which is convenient but introduces the need to handle exceptions carefully, a concern that connects to nonmonotonic reasoning.