ALGORITHM FOR SEMANTIC TEXT ANALYSIS BY MEANS OF BASIC SEMANTIC TEMPLATES WITH DELETION

Abstract

The systems of automatic text processing have become more and more important due to the constant growth of textual data. One of the main issues arising in such systems is a problem of semantic analysis. The paper deals with an algorithm for finding semantic dependencies by means of basic semantic templates with deletion. While working with the Drools expert system (and PHREAK algorithm for fast pattern matching) we have developed and implemented a semantic analyzer for construction of semantic dependencies between parts of a sentence. During the semantic analysis we add some text parts to the priority queue according to the rules described in the semantic templates, and then at each iteration of the sentence being analyzed we drop some segment of the analyzed text which has the highest priority in the queue. To determine the priority in this queue two values are used: the priority of semantic relationship group and word position. The proposed algorithm is implemented in Java. We have prepared 2160 rules using Drools expert system. The software implementation of the proposed algorithm has shown its applicability for the systems of automatic text processing. Testing results have proved that suggested algorithm of semantic analysis without Drools expert system operates 6-8 times slower, on the average. We use proposed semantic analyzer as a composite module to intellectual question-answering system.

Keywords

• semantic dependencies
• semantic analyzer
• semantic templates