Особливості обробки текстових даних соціальних мереж при аналізі емоційної тональності

Abstract

Are being considered the specific features of processing textual data from social networks for sentiment analysis. Social media comments, particularly on YouTube, reflect users’ opinions and emotional reactions to digital content. However, such texts often contain slang, abbreviations, emojis, and spelling errors, which complicate automatic analysis. Therefore, text preprocessing plays an important role and includes data cleaning, normalization, and tokenization. Special attention is given to emojis that may carry additional emotional meaning. Modern Natural Language Processing methods and transformer models such as BERT improve the accuracy of sentiment classification. The use of the YouTube API allows automatic collection of comments for analysis. The obtained results can be visualized to better understand audience reactions.