Research on computer recognition of sentiment expression in Russian satirical literature based on an improved transformers algorithm

This study explores an enhanced Transformer-based algorithm for the automated recognition of sentiment expression in Russian satirical literature, a genre known for its use of subtle linguistic devices such as irony, sarcasm, and metaphor. These devices often pose significant challenges for conventional natural language processing (NLP) models, which struggle to capture such nuanced expressions. To address these challenges, we propose an improved Transformer architecture that incorporates domain-specific pre-training and fine-tuning on a carefully curated dataset of Russian satirical texts. Key innovations include a novel attention mechanism designed to better capture the complex contextual features of satire, as well as an advanced sentiment classification module optimized for detecting sentiment polarity in satirical contexts. Experimental results demonstrate that the proposed model outperforms state-of-the-art baselines in both sentiment polarity detection and classification accuracy, showcasing its effectiveness in literary sentiment analysis. This research underscores the potential of computational methods to enhance our understanding of literary texts, particularly in complex genres like satire, and offers new tools for sentiment analysis and text mining.