Semantic Networks & Ontologies in AI

A semantic network represents knowledge as a web of connected concepts, so an AI system can look at how things relate to each other, not just what they are on their own.

What Is a Semantic Network?

A semantic network is a graph-based way to represent knowledge. Concepts are drawn as nodes, and the relationships between them are drawn as labeled links. Following the links lets a system answer questions and infer things that were never stated directly.

Nodes and Relationships

Every semantic network is built from just two building blocks.

• Nodes — the concepts or objects, like Car or Vehicle.
• Links (edges) — labeled relationships between two nodes, such as is-a, has-part, has-property, or capable-of.

The most common link is is-a, since it lets a node inherit whatever is true of the node it points to.

Example Semantic Network

Here's a small semantic network about vehicles, using is-a, has-part, and has-property links.

Example: Building and Querying a Network

edges = [
    ("Car",       "is-a",         "Vehicle"),
    ("Truck",     "is-a",         "Vehicle"),
    ("Vehicle",   "has-property", "can-move"),
    ("Car",       "has-part",     "Engine"),
    ("Car",       "has-part",     "Wheel"),
    ("Engine",    "has-property", "burns-fuel"),
   ("ElectricCar","is-a",        "Car"),
   ("ElectricCar","has-property","zero-emissions"),
]
def relationships_of(node):
    return [(rel, obj) for subj, rel, obj in edges if subj == node]

def all_ancestors(node):
    ancestors = []
    current = node
    while True:
        parent = next((obj for subj, rel, obj in edges if subj == current and rel == "is-a"), None)
        if not parent:
            break
        ancestors.append(parent)
        current = parent
    return ancestors

print("Relationships of 'Car':", relationships_of("Car"))
print("Relationships of 'ElectricCar':", relationships_of("ElectricCar"))
print("'ElectricCar' is-a chain:", all_ancestors("ElectricCar"))

Output

Relationships of 'Car': [('is-a', 'Vehicle'), ('has-part', 'Engine'), ('has-part', 'Wheel')]
Relationships of 'ElectricCar': [('is-a', 'Car'), ('has-property', 'zero-emissions')]
'ElectricCar' is-a chain: ['Car', 'Vehicle']

Explanation:

ElectricCar only directly states one property (zero-emissions), but its is-a chain shows it's also a Car and a Vehicle — so it inherits has-part Engine and has-property can-move too, without those facts being repeated on ElectricCar itself.

Ontologies and Knowledge Graphs

Semantic networks scale into two related ideas once a domain gets bigger or more formal.

What Is an Ontology?

An ontology is a formal, agreed-upon set of concepts, categories, and relationships for a specific domain — it defines exactly what kinds of nodes and links are allowed, so different systems can share knowledge without ambiguity.

Knowledge Graphs

A knowledge graph is a semantic network built at a much larger scale, often powered by an ontology's rules, and often pulling in real-world data. Search engines use knowledge graphs to connect facts about people, places, and things across the web.

Quick Comparison