Building a model for resolving referential relations in a multilingual system

Abstract

This paper considers an approach to resolving referential relations when extracting information from a text. The proposed approach is an attempt to integrate the multifactorial model of the activation coefficient with the approach to resolving the referential ambiguity of the text when replenishing the ontology. The found objects are compared based on an assessment of the proximity of attributes and relationships of objects. An ontological interpretation of relations and measures of similarity of attributes based on a multifactorial model is proposed. This model is distinguished by the fact that it makes it possible to introduce the concepts of "rhetorical distance", "linear distance", "animation", "distance between paragraphs", and "syntactic and semantic role of the antecedent". A multifactorial model is proposed, which is a necessary and sufficient component for the purpose of explaining the measure of similarity of referents for choosing the best applicant. The counting system and its modification were revealed by trial and error; the work was carried out until the selected numerical weights began to explain all the available material. The current study also examines the factors of choice of reference devices that make it possible to work with complex sentences and texts. Moreover, examples of finding a measure of proximity in a multilingual system for the Kazakh, Russian, and English languages are offered. For the current paper, texts in the Russian, English, and Kazakh languages were used as a source for practical tasks. The texts were selected using news articles on the Internet sites where translations into other languages, including those named above, were offered. The authors of this study have done massive practical work, which confirms the correctness of the thesis they are considering