Based on RuNewsCorp: Improving Accuracy in Long Text Classification

Text classification is an extremely important task in the field of Natural Language Processing (NLP), and sentiment analysis is a subfield of text classification that identifies and categorizes the emotional tendencies within text. This is crucial for market research, brand monitoring, and customer service. However, current sentiment analysis is predominantly based on short English reviews posted on Twitter. There is a scarcity of sentiment analysis tasks for long Russian texts. This paper introduces the construction of a sentiment classifier for Russian sports news, based on a large language model-BERT, using fine-tuning methods on the self-built dataset Ru_News_Corp_Sen. Furthermore, the paper explores and compares various methods for improving the accuracy of sentiment classification in long texts.