Semantic Network

A semantic network is one way to formally represent human knowledge as a web of interconnected concepts and relations. Typically, the network is made up of labeled nodes, with directed, labeled links between the nodes. The labels can be ordinary words, concepts, or codes that make up the semantic aspect of the network. The term semantic network dates back to early work on artificial intelligence and automated translation in the 1960s, but the same idea was used much earlier by philosophers, psychologists, and linguists for classification, logic reasoning, and grammar. Today, semantic networks are but one of many methods underlying the development of geospatial ontologies, that is, formalized specifications of geographic knowledge.

Since a semantic network describes and stores meaning by linking words together using specified relationships, it works very much like a normal dictionary, where any word is described with the help of other words and how it is related to other words, for example, by giving synonyms, antonyms, functional relationships, and so forth. The American Heritage Dictionary, for example, defines the term moraine as “An accumulation of boulders, stones, or other debris carried and deposited by a glacier.” What makes a semantic network special is the way these descriptions are made explicit and follow a specific syntax: labeled nodes and links in a network structure. This lends itself well for a graphical notation and visualization, which provides an intuitive interface similar to a mind map. In this way, a semantic network integrates ideas [Page 385]from formal logics, lexical semantics, and mathematical graph theory. As such, it can be used not only to describe definitions and assertions about a body of knowledge but also to reason with and act upon information to support system automation.

One type of semantic network is mostly declarative in that it focuses on type-subtype relations. We could, for example, link the two words boulder and moraine with the relation “is part of.” Such networks typically form hierarchical or taxonomical structures that describe the knowledge embedded in a classification system. Another common type of semantic network is more process oriented in that it contains assertions about how nodes interact. For example, the two words glacier and moraine linked with the relation “deposits” provides some knowledge about the process creating a moraine. Other types of semantic networks are also common, as are combinations of these. Combining the two examples above, we have a body of knowledge not only about glaciological deposits, but we may also through logic and reasoning say that a glacier deposits boulders, a fact not explicitly coded in the network. Initially, semantic networks were constructed as isolated islands of knowledge, but recent developments of the World Wide Web have pointed at the potential of a “semantic web,” built on the notion of knowledge developed as a distributed system where agreed-on standards to a large degree follow the original ideas of a semantic network.