Semantic network: definition, classification and application

A semantic network is an information model of a particular domain. The network is constructed by the type of an oriented graph, in which vertices are objects of a given domain, and edges or arcs are relations between objects. The role of objects can be concepts, properties, events, operations or processes, and the number of possible types of network relationships depends solely on its creator, based on the goals. Thus, any semantic network is one of the most effective methods of representing knowledge. In addition, it is most conveniently perceived by the human brain form.

For the first time, the phrase "semantic network" was used in 1968-1969 in the works of the American psychologist Quillian, who studied semantic memory. Later this term began to be actively applied in mathematics, computer science and linguistics. Today, the presentation of information in the form of semantic networks is common in many areas of knowledge, there are a great many of them, and they all differ from each other in the number of types of relations, the number of arguments and the size.

If the number of relationship types is taken as the basis for the classification, then each semantic network can be referred to either a homogeneous or a non-homogeneous type. For homogeneous networks, there is only one type of relationship (arrows). Heterogeneous networks have two or more types of relations. For practical purposes, the most interesting is the heterogeneous semantic network. An example of such a network is the structure of the well-known Internet encyclopedia Wikipedia.

According to the number of arguments, semantic networks are divided into binary and n-ary. Binary models demonstrate relations only between two objects (concepts) and are the simplest and most convenient for graphical representation. Models of n-ary type, most often, are encountered in practice and are necessary for mapping the connection between several objects.

Finally, in size, any semantic network can be attributed either to small models that solve specific tasks, or to networks of an industrial scale that serve as the basis for creating systems, or to a group of global models.

If we talk about the areas in which semantic networks have received the greatest distribution to date, this is certainly computer science. It is in this area of knowledge that the concept of a network is used most often. Typically, computer semantic networks are used to implement machine translation of languages. In this case, the translation process is performed in two stages. On the first, the source text is translated into a so-called intermediate form, that is, its representation in the form of a semantic network, and in the second stage the intermediate form obtained is translated into the desired language.

To effectively implement networks today, specially designed network languages such as Net and Simer + Mir are used. In addition, expert systems are widespread that use semantic models as a language of visual representation of knowledge: Casnet, Prospector, Torus.