Text Mining of Semantic Keywords Associated with a Target Text Feature

Author: Dzhemadinov Eskender Rustemovich, 2nd year undergraduate student at IAMUP 17–1 group at HSE, Perm. The topic of the master's thesis is the automatic identification of “semantic elements” of tests that affect their target characteristics. In this paper, we propose an approach to identifying such elements in texts programmatically and then building on their basis predictive models of machine learning. For analysis, data were taken from the Google competition to identify the toxicity of comments and machine learning models were built. The paper shows that The current study includes 2 chapters: 1. Theoretical aspects of the topic under study, a description of the main models for the presentation of textual content and brief theoretical information about the machine learning models used. 2. Description of the proposed methodology and the practical implementation of machine learning models using real data as an example, social network comments with an assessment of their accuracy. The number of pages - 42, the number of images - 14, the number of tables - 3.