Детальная информация

The main purpose of our research is to improve the algorithm for analyzing the tonality of texts containing feminitives. In the course of our work, we had the task of selecting corpus materials to test one of the systems for evaluating the tonality of texts containing feminitives, and if the evaluation algorithm works inaccurately, retrain it to obtain more correct results. As a hypothesis, it was suggested that a significant part of the feminitives describing womens professional activities are used more often with negative connotations in various contexts. The practical part of the study was carried out in several stages. First, we analyzed the proposals of the NKRJ newspaper half-corps using the Dostoevsky library. The results obtained at this stage were compared with the assessment of the emotional coloring of the sentences, which was obtained during a psycholinguistic experiment to determine the assessment. 39 students of the SPbPU took part in the survey. At the third stage, the algorithm for evaluating proposals was improved, for which the SpaCy library was used. To work, we first selected comments on Yandex Zen and conducted an analysis to train the model. We also used a dictionary of evaluative words.  Expressions of the Russian language - Rusentilex. Then we re-analyzed the proposals in order to verify the operation of the system. For this purpose, the same proposals were used that we analyzed at the first stage, and the students in the process of completing the survey. The result turned out to be much more accurate and more consistent with the answers received when evaluating the tonality of sentences by native speakers.