Automatic Pattern Recognition Applied to Semantic Problems

The idea of computer pattern recognition as applied to semantics has the same basic principles as the theory of the perception, which represents a sufficiently generalized model for that part of the brain whose function is classification, insofar as classification presupposes similarity between elements (or groups of elements) of the same class and difference between elements (or groups of elements) of different classes. In this respect automatic semantic pattern recognition (ASPR) consists in the following. An unknown text is analyzed by computer. At the first stage of the text treatment its subject matter is identified at the word-form level. The second stage results in indicative abstracting at the phrase (word-group) level. The theory of semantic pattern recognition offers a number of concepts with their structural and functional particularities. First of all, these are concepts of 1) semantic space, as a certain semantic zone of the language; 2) semantic regions, as certain semantic subzones of this space; 3) distinguishing factors, those semantic elements in whose terms the semantic space is being formed; 4) pattern alphabet-a list of the region names and finally the concept of 5) semantic pattern itself. The semantic pattern is regarded to be the name of a particular region of the semantic space wherein information structure on phenomena of the outer world is reflected. This information structure is represented as an ordered set consisting of objects (as extra-linguistic reality) and its relations. Any information on the phenomena of objective reality could be rendered by means of combinations of objects together with their relations. Any finite set of objects combined with the relations between them produces situation. So, properly speaking, the semantic pattern is a situation mould, sui generis, transferred to the na