Algorithm for Hierarchical Evaluation of the Entity of Direct Opinion in the Task of Sentiment Analysis of Complicated Text Content From Social Media

Internet messengers and chat platforms, including those utilized by transport companies, taxi services, energy providers, and similar entities, have long been established as customary channels for customers to interact with service providers and provide feedback.In this context, the main task is the analysis of the semantic and sentiment components of the textual content produced in them. A feature of chat messages is a relatively small number of words, the use of jargon and specific symbols, emojis, etc., which creates a serious problem for extracting the semantic component for both humans and artificial intelligence. The authors call such a text complicated.The authors propose an algorithm for evaluating the entity, which is the subject of discussion in a chat message, and their features, based on the use of dependency parsing. Its quality is assessed on real data, where both humans and artificial intelligence based on a generative model act as experts.Implementing such an approach, specifically through the utilization of AI-powered chatbots or managers, will enhance the quality of service experienced by customers of power generating companies and similar services.