Source: http://www.jfsowa.com/pubs/semnet.htm 1 Semantic Networks John F. Sowa A semantic network or net is a graphic notation for representing knowledge in patterns of interconnected nodes and arcs. Computer implementations of semantic networks were first developed for artificial intelligence and machine translation, but earlier versions have long been used in philosophy, psychology, and linguistics. What is common to all semantic networks is a declarative graphic representation that can be used either to represent knowledge or to support automated systems for reasoning about knowledge. Some versions are highly informal, but other versions are formally defined systems of logic. Following are six of the most common kinds of semantic networks, each of which is discussed in detail in one section of this article. 1. Definitional networks emphasize the subtype or is-a relation between a concept type and a newly defined subtype. The resulting network, also called a generalization or subsumption hierarchy, supports the rule of inheritance for copying properties defined for a supertype to all of its subtypes. Since definitions are true by definition, the information in these networks is often assumed to be necessarily true. 2. Assertional networks are designed to assert propositions. Unlike definitional networks, the information in an assertional network is assumed to be contingently true, unless it is explicitly marked with a modal operator. Some assertional netwoks have been proposed as models of the conceptual structures underlying natural language semantics. 3. Implicational networks use implication as the primary relation for connecting nodes. They may be used to represent patterns of beliefs, causality, or inferences. 4. Executable networks include some mechanism, such as marker passing or attached procedures, which can perform inferences, pass messages, or search for patterns and associations. 5. Learning networks build or extend their representations by acquiring knowledge from examples. The new knowledge may change the old network by adding and deleting nodes and arcs or by modifying numerical values, called weights, associated with the nodes and arcs. 6. Hybrid networks combine two or more of the previous techniques, either in a single network or in separate, but closely interacting networks. Some of the networks have been explicitly designed to implement hypotheses about human cognitive mechanisms, while others have been designed primarily for computer efficiency. Sometimes, computational reasons may lead to the same conclusions as psychological evidence. The distinction between definitional and assertional networks, for example, has a close parallel to Tulving's (1972) distinction between semantic memory and episodic memory.Source: http://www.jfsowa.com/pubs/semnet.htm 2 Network notations and linear notations are both capable of expressing equivalent information, but certain representational mechanisms are better suited to one form or the other. Since the boundary lines are vague, it is impossible to give necessary and sufficient conditions that include all semantic networks while excluding other systems that are not usually called semantic networks. Section 7 of this article discusses the syntactic mechanisms used to express information in network notations and compares them to the corresponding mechanisms used in linear notations. 1. Definitional Networks The oldest known semantic network was drawn in the 3rd century AD by the Greek philosopher Porphyry in his commentary on Aristotle's categories. Porphyry used it to illustrate Aristotle's method of defining categories by specifying the genus or general type and the differentiae that distinguish different subtypes of the same supertype. Figure 1 shows a version of the Tree of Porphyry, as it was drawn by the logician Peter of Spain (1329). It illustrates the categories under Substance, which is called the supreme genus or the most general category. Figure 1. Tree of Porphyry, as drawn by Peter of Spain (1329)Source: http://www.jfsowa.com/pubs/semnet.htm 3 Despite its age, the Tree of Porphyry represents the common core of all modern hierarchies that are used for defining concept types. In Figure 1, the genus Substance with the differentia material is Body and with the differentia immaterial is Spirit. The modern rule of inheritance is a special case of the Aristotelian syllogisms, which specify the conditions for inheriting properties from supertypes to subtypes: LivingThing inherits material Substance from Body and adds the differentia animate; Human inherits sensitive animate material Substance and adds the differentia rational. Aristotle, Porphyry, and the medieveal logicians also distinguished the categories or universals from the individual instances or particulars, which are listed at the bottom of Figure 1. Aristotle's methods of definition and reasoning are still used in artificial intelligence, object-oriented programming languages, and every dictionary from the earliest days to the present. The first implementations of semantic networks were used to define concept types and patterns of relations for machine translation systems. Silvio Ceccato (1961) developed correlational nets, which were based on 56 different relations, including subtype, instance, part-whole, case relations, kinship relations, and various kinds of attributes. He used the correlations as patterns for guiding a parser and resolving syntactic ambiguities. Margaret Masterman's system at Cambridge University (1961) was the first to be called a semantic network. She developed a list of 100 primitive concept types, such as Folk, Stuff, Thing, Do, and Be. In terms of those primitives, her group defined a conceptual dictionary of 15,000 entries. She organized the concept types into a lattice, which permits inheritance from multiple supertypes. The basic principles and even many of the primitive concepts have survived in more recent systems of preference semantics (Wilks & Fass 1992). Among current systems, the description logics include the features of the Tree of Porphyry as a minimum, but they may also add various extensions. They are derived from an approach proposed by Woods (1975) and implemented by Brachman (1979) in a system called Knowledge Language One (KL-ONE). As an example, Figure 2 shows a KL-ONE network that defines the concepts Truck and TrailerTruck as subtypes of Vehicle.Source: