Semantic network

A semantic network is a formal model of concepts and their relations (relations). It is used in computer science in the field of artificial intelligence for knowledge representation. Occasionally, one also speaks of a knowledge network. Usually, a semantic network is represented by a generalized graph. The nodes of the graph while the terms dar. relations between the concepts are realized by the edges of the graph. What relationships are allowed, is defined very differently in different models, but most of the relationship types is inherent in a cognitive aspect. Semantic networks have been proposed in the early 1960s by the linguist Ross Quillian as a form of representation, semantic knowledge. Thesauri, taxonomies and word networks are forms of semantic networks with limited set of relations.

Lexical- semantic relations

A ( usually binary ) relation between two graph nodes can be, among others:

• Hierarchical relations Inheritance relation: for example, is ' dog ' is a narrower term of ' mammal ', ' mammal ' is a narrower term for ' animal '. This relation is transitive and asymmetric. Some semantic networks allow multiple inheritance. So ' banana ' as a term can be modeled both of ' Tropical ' and of ' plantation plant '. For the Term preamble also the expression hypernym is used for the term minor term, the expression hyponym.
• Instance relation: this relation connects individuals with classes; For example, Bello is an instance of the Dog class. This relationship is asymmetrical and, in contrast to OA Subset relationship is not transitive. Counter-example: Bello is an instance of the Dog class. Dog is an instance of the class species. Bello as a dog person but is not an instance of the class species.
• Parti Tive relation ( meronymy ): eg dog's fur is a part of him. The purpose converse relation is called Holonymie, eg containing a pea soup peas, a lot is made of elements, a window containing the material glass. This relation is asymmetric. Also not necessarily true that A is the Holonym to B if B is the meronym of A. The transitivity of this generic relation is not always given. Roger Chaffin has shown that subclasses of the partitive relation, for example, the element group relation, quite transitive are ( ref: Chaffin, pp. 273-278 ).

The psychological reality of such semantic networks can be studied, for example, with the help of association techniques and sentence verification tasks.

Problems of modeling

Polysemy and homonymy play in the modeling of semantic networks a subordinate role, as it comes to relationships between concepts. A polysemous (or homonymous ) lexeme is associated with two or more terms or lies in the lexical range of values ​​of two or more terms. However, there is one in practice often difficult question of how many and is assigned to which terms a lexeme.

A much bigger problem for the modeling of semantic networks are lexical gaps dar. These are terms that have no simple lexical characters can be assigned as a value in a natural language. A well-known example is the term ' no longer thirsty '.

Modern representatives of the semantic networks

Current knowledge representation methods that are based on semantic networks, are developed by Stuart C. Shapiro Semantic Network Processing System ( SNePS ) and the MultiNet paradigm of multi-layered extended semantic networks of Hermann Helbig. For both approaches, there are also tools to support knowledge acquisition and processing. MultiNet is particularly focused on the semantic representation of natural language knowledge and is used in various applications of Natural Language Processing.

History

In the winter semester 1789/90 described Johann Friedrich Flatt in his Tübingen lecture on empirical psychology - which was attended by Hegel, Schelling, and Hölderlin - a network model of memory to explain associative activations during retrieval of memory contents. As early as 1900, by psychologists, such as Gustav Aschaffenburg, studies carried out as concepts are linked together in our brain. It was found that certain words the same associations cause in most people, for example, white-black or mother - father. By association, it is possible to combine terms on the semantic level with each other, ie, that words for a reason related to each other. Words and meanings are stored in the mental lexicon neither alphabetical nor completely disorganized, but like a net. The meaning of a word is represented in such a network by nodes and by neighborhood relations to other content or terms. Such networks can be obtained in the approach of the associations just mentioned.

This concept of associations and neighborhood relations tries to pick up the Semantic Web. This may be possible that semantically related but syntactically completely different terms are found in the semantic network. The figure shows such a mental association structure which forms the basic structure for a semantic network.

The WWW as a semantic network

According to Tim Berners -Lee, a semantic network ( → semantic web ) is to be stretched over the organized as a hypertext portion of the Internet (WWW). The contents of the resources that make up this hypertext, will be described with metadata. This resource description is done using a Resource Description Framework (RDF). As a modeling language, the Web Ontology Language (OWL ) to be used. The goal is that the terms that are used in the metadata are provided with well-defined and thus machine interpretable meanings. This would for example allow content-based information retrieval. This means that not all the terms are recognized globally in a complex ontology, but to create a rather loose network of decentralized specialized ontologies.

Possible formats for the representation of semantic networks are RDF, RDF Schema, OWL and XML Topic Map.