APPROACHES TO IMPLEMENTING EMOTIONAL ARTIFICIAL INTELLIGENCE IN TEXT ANALYSIS TASKS

In the era of digital transformation and the exponential growth of text data generated on the Internet, analyzing the emotional sentiment of texts is becoming particularly relevant. Emotional Artificial Intelligence (EAI), aimed at recognizing, interpreting, and simulating human emotions, is finding increasingly widespread application in natural language processing. This article provides an overview and comparative analysis of the main approaches to implementing EAI in text analysis tasks, commonly known as sentiment analysis or tonality analysis. Three key groups of methods are considered: lexiconbased approaches relying on sentiment dictionaries; classical machine learning methods requiring manual feature engineering; and modern deep learning methods, including recurrent and transformer neural networks, which demonstrate stateoftheart results. The advantages and disadvantages of each approach are analyzed in terms of accuracy, interpretability, computational complexity, and data requirements. Special attention is paid to the evolution of methods, from simple dictionarybased techniques to complex deep neural network architectures, and their applicability to various types of text data and tasks. The conclusion discusses current challenges and promising directions for the development of EAI in text analysis.