Catalogue of Artificial Intelligence Techniques
Aliases: Semantic Nets
Keywords: combinatorial explosion, nodes
Categories: Knowledge Representation
Author(s): Karen Sp\"arck Jones
Semantic nets are a principle for the large scale organisation of knowledge emphasising the multiple associations of individual concepts. Concepts, objects, entities, etc. are represented as nodes in a linked graph, and relationships between these are represented as labelled arcs. The range of possible network structure types is very wide (see Findler). Semantic nets should properly be based on definitions of the net structure, i.e., the syntax and semantics of nodes and links and of configurations of these, and of net operations, i.e., the syntax and semantics of node-node transitions, but too frequently are not. Nets have been found an attractive descriptive device, but genuine algorithmic exploitation of nets based, e.g., on the general idea of Marker-passing for selective reading or writing at nodes, is comparatively rare (formal graph theory is rarely seriously invoked in Artificial Intelligence). The emphasis on concept association introduces difficulties in representing any partitioning or grouping of net elements, for example to represent quantified propositions, clusters of similar entities, etc. (but see Partitioned Net), and network searching conspicuously manifests the general AI problem of the combinatorial explosion.
- Lehmann, F., ed., Semantic Networks in Artificial Intelligence, Pergamon Press, Oxford, 1992.