Способи виявлення логічних зв’язків між частинами текстових документів

Abstract

Метою роботи є створення формального опису способів виявлення логічних зв’язків між частинами тексів. Одним із засобів лінгвістичного аналізу текстової інформації та екстракції знань є логіко-лінгвістичні моделі. Довільний текст можна представити у вигляді масиву логіко-лінгвістичних моделей речень природної мови, логічні операції та структурні компоненти яких вказують на типи логічних зв’язків. Пропонується для відображення структури тексту використовувати формальний опис п’яти абстрактних моделей, що в подальшому дають можливість визначити тематику тексту.

The objection of work is creation a formal specification of methods of identifying logical connections between parts of texts. One of the means of text information linguistic analysis and knowledge extraction is logic and linguistic model. Any text can be presented as array of logic and linguistic models of natural language sentences, where logic operations and structural components indicate the types of logical relationships. There are two levels while constructing the text. The first level reflects the grammar links between the natural language sentences, and the second one is in content relationships between fragments of text. Exactly the construction of the second level is proposed to be used for formal description of five abstract models. For that purpose, every natural language sentence transforms into logic and linguistic model, which is an atomic predicate for the simple sentence and, correspondingly, set of atomic predicates for the complex sentence. The first abstract model is based on the consistent deployment of information when the object of the preceding sentence becomes subject for the next sentence. For the second abstract model the key point is usage of the same subject in a few natural language sentences, and within these sentences subject is represented in different interpretations. The third model is based on the fact that the subjects of logic and linguistic models of the next natural language sentences are hyperons or conversives in relation to the subject of the first sentence of text fragment. For the fourth abstract model the subject of the first sentence is partly the subject for all the next sentences. In the fifth model the subjects of the next sentences can be both objects and matters of the first sentence and its subject. The result of the construction of such abstract models are polyhedrons that subsequently used to determine the texts’ subject (topic).