AI Notes
Master semantic networks. Nodes, links, inheritance. AI 2024.
What Are Semantic Networks?
A semantic network is a knowledge representation scheme that uses a directed graph to represent concepts and their relationships. Nodes represent entities, concepts, or events, while labeled edges represent the relationships between them. This intuitive visual representation mirrors how humans often think about knowledge—as a web of interconnected ideas.
Semantic networks were among the earliest AI knowledge representation formalisms, introduced by Ross Quillian in 1968 for modeling human semantic memory. They remain relevant today as the conceptual ancestor of modern knowledge graphs (Google Knowledge Graph, Wikidata), ontologies (WordNet), and graph databases (Neo4j).
Structure and Components
Basic Elements
NODES (Concepts/Entities)
● Dog
● Animal
● Mammal
● Fido (instance)
● Running (action)
EDGES (Relationships)
→ is-a (taxonomy/classification)
→ has-part (composition)
→ has-property (attributes)
→ can-do (capabilities)
→ instance-of (specific example)
Example: Animal Kingdom Network
| ┌─────│ Animal │──── can | move |
| │ └────┬────┘ has | metabolism |
| │ │ Mammal │──── has | warm_blood |
| │ └────┬────┘ has | fur |
| └─────│ Dog │──── can | bark |
| └───┬───┘ has | tail |
| instance │ eats | meat |
| │ Fido │──── color | brown |
| └───────┘ owner | John |
Inheritance in Semantic Networks
The power of semantic networks comes from inheritance traversal:
| Query | "Can Fido move?" |
| 1. Check Fido node | no direct "can: move" edge |
| 2. Follow instance-of | Dog |
| 3. Check Dog node | no direct "can: move" edge |
| 4. Follow is-a | Mammal → Animal |
| 5. Check Animal node | found! "can: move" |
| Answer | Yes (inherited from Animal) |
| Query | "Does Fido have warm blood?" |
| 1. Fido | Dog → Mammal → found "has: warm_blood" |
| Answer | Yes (inherited from Mammal) |
Inheritance with Exceptions
| │ Bird │── can | fly |
| │ Penguin │── can | fly = NO (override!) |
| └───────────────┘ can | swim = YES |
| Query | "Can a penguin fly?" |
| 1. Penguin | check local properties → can:fly = NO |
| Answer | No (exception to Bird's default) |
Types of Relationships
Taxonomic (Classification)
| is-a | Subclass relationship |
| instance-of | Individual membership |
| Distinction | "Dog is-a Animal" (class level) |
Part-Whole (Meronymy)
Associative Relationships
| causes | Virus causes Disease |
| lives-in | Bear lives-in Forest |
| used-for | Hammer used-for Nailing |
| made-of | Table made-of Wood |
Reasoning in Semantic Networks
Spreading Activation
When a node is activated, activation spreads to neighboring nodes along edges, with decay:
| "FIRE" + "RED" | both activate "danger" |
| Strongest path: FIRE | danger←RED |
| Association | fire and red are related through danger |
This models human associative memory: "What's related to both fire and red? → danger, warning."
Marker Passing
| Query | "How are Einstein and Princeton related?" |
| Pass marker from Einstein: | physicist, German, E=mc², Princeton |
| Pass marker from Princeton: | university, New Jersey, Einstein |
| Markers meet at | Einstein worked-at Princeton |
| Answer | "Einstein worked at Princeton University" |
Semantic Networks vs. Other Representations
| Feature | Semantic Nets | Frames | Logic | KGs |
|---|---|---|---|---|
| Visual | Yes (graph) | Tables | Formulas | Yes (graph) |
| Inheritance | Yes | Yes | Derived | Yes |
| Formality | Low | Medium | High | Medium-High |
| Reasoning | Traversal | Slot-fill | Inference | Query+ML |
| Scale | Small-medium | Medium | Any | Massive |
| Modern use | Conceptual | Limited | Theorem proving | Industry standard |
Problems with Semantic Networks
The IS-A Ambiguity
Lack of Formal Semantics
Quantification Problems
"Every student reads some book" →
How to represent in a network?
∀x(Student(x) → ∃y(Book(y) ∧ Reads(x,y)))
Nodes are specific things — how to show "every" and "some"?
Solution: Conceptual graphs (Sowa) add explicit quantifier nodes
Modern Evolution: Knowledge Graphs
Semantic networks evolved into today's knowledge graphs:
Traditional Semantic Network (1970s)
Small scale, academic, no formal semantics
Modern Knowledge Graph
Billions of triples (subject-predicate-object)
Formal ontology (OWL/RDFS schemas)
Embedding-based reasoning (TransE, DistMult)
Industry deployment (Google, Amazon, Microsoft)
RDF Triple format
(Einstein, birthPlace, Ulm)
(Ulm, country, Germany)
(Einstein, field, Physics)
Interview Questions
Q: What is the difference between a semantic network and a knowledge graph? A: A knowledge graph is a modern, large-scale evolution of semantic networks with formal semantics (RDF/OWL), standardized query languages (SPARQL), and scalable storage. Semantic networks are the historical concept; knowledge graphs are the industrial implementation with billions of facts and machine learning integration.
Q: How does inheritance work in semantic networks? A: Properties are inherited by traversing is-a links upward. When querying a node's property, if not found locally, the system follows is-a edges to parent classes until the property is found. Local values override inherited ones (exceptions). This enables compact representation—shared properties stated once at the appropriate abstraction level.
Q: What are the limitations of semantic networks for AI reasoning? A: Three main limitations: (1) No formal semantics—edges lack precise logical meaning, (2) Cannot easily represent quantifiers, negation, or disjunction, (3) Inheritance anomalies with multiple inheritance paths. These led to the development of description logics and formal ontology languages that provide rigorous semantics.
Exam Focus
Revise definitions, diagrams, examples, and short-answer points for Semantic Networks - Knowledge Representation.
Interview Use
Prepare one clear explanation, one practical example, and one common mistake for this Artificial Intelligence topic.
Search Terms
artificial-intelligence, artificial intelligence, artificial, intelligence, knowledge, representation, semantic, networks
Related Artificial Intelligence Topics