Enhancing emotion analysis through valence-arousal modeling with user and AI evaluations

With the advancement of text mining and natural language processing technologies, sentiment analysis has found widespread application across various fields. However, current research often emphasizes binary or multi-class classifications, which fail to capture the full spectrum of human emotions. To address this issue, the valence-arousal (VA) model has been proposed but encounters challenges such as data imbalance and subjective labeling. This study presents a novel approach that integrates large language models with user evaluations and employs relevant augmented generation techniques to enhance data quality and consistency. In addition, the VA data is visualized to assess its utility in multidimensional sentiment analysis. Future research will focus on expanding the dataset and conducting in-depth analyses to further validate the proposed approach.