Место издания:Ульяновский государственный технический университет Ульяновск
Первая страница:71
Последняя страница:76
Аннотация:This article is devoted to sentiment analysis of the texts written in a natural language, which contain elements of a person's appearance. The provided experiment was carried out only for texts in a Russian language. For the study, an information model of the appearance of a person was developed and described in the frame representation language notation. At first, person's references and elements of their appearance are extracted from the text in the form of separate unigrams. The next step is to teach certain models of sentiment analysis on a training data. The neural network, the naïve Bayes and the support vectors machine were used in this work. The accuracy of the correct determination of the sentimental class was chosen as efficiency criteria of these methods.