Computer Analysis of Text Tonality Based on the JSM Method

Abstract

The necessity of taking into account the coefficient of emotionality of lexemes to assess the emotionality of the text, which should be taken into account in voice control systems, is substantiated. The reduction of the JSM method was made, which made it possible to apply it to solving the problem of estimating the emotional load of the text. The core of the program module has been formed as part of a thesaurus classifier trained on collections of expert texts. The analysis of the quality of the proposed method of estimating the tonality of the text is presented. The results of the calculated values of the efficiency metrics showed that the use of a manual tonal dictionary allows to obtain a more accurate assessment of the text tonality and significantly reduce the operating time, since, compared to the automatic dictionary, the manual dictionary does not contain neutral words. The dependence of the accuracy of determining the tonality on the conflict resolution function used to classify texts by tonality and the number of training texts was discovered. The problem of limited emotive space was discovered: some of the words did not fall or partly fell under the binary classification. The algorithm for determining the tonality of the morphological composition of the text on the basis of the JSM method using tonal dictionaries and classification allows you to determine the tonality of the text on a binary scale.