ORGANIZATION OF A TEXT AUTOMATIC SUMMARIZATION SYSTEM BASED ON FUZZY HIERARCHICAL NEURAL NETWORK

Abstract

The approach to the construction of the automatic text summarization (ATS) system based on hierarchical fuzzy neural network (HFNN), which is oriented on structural and non-structural features for extraction at different levels of the text, taking into account connectivity initial text of the document, which allows consider not only statistical information, but also organizational and syntax-semantic structure of the text, is proposed. The mathematical apparatus HFNN and its learning algorithm for solving ATS task are considered. By means of Matlab the experiment for HFNN approbation and also verifying its effectiveness is done. It is shown that the quality of the summary, obtained using the proposed method is higher compared to the summary, obtained by conventional statistical methods.

Keywords

• автоматичне реферування тексту, ієрархічна нечітка нейронна мережа, ознаки екстрагування