Weibo sentiment analysis has gained prominence, particularly during the COVID-19 pandemic, as a means to monitor public emotions and detect emerging mental health trends. However, challenges arise from Weibo’s informal language, nuanced expressions, and stylistic features unique to social media, which complicate the accurate interpretation of sentiments. Existing models often fall short, relying on text-based methods that inadequately capture the rich emotional texture of Weibo posts, and are constrained by single loss functions that limit emotional depth. To address these limitations, we propose a novel framework incorporating a sentiment graph and self-supervised learning. Our approach introduces a “sentiment graph” that leverages both word-to-post and post-to-post relational connections, allowing the model to capture fine-grained sentiment cues and context-dependent meanings. Enhanced by a gated mechanism within the graph, our model selectively filters emotional signals based on intensity and relevance, improving its sensitivity to subtle variations such as sarcasm. Additionally, a self-supervised objective enables the model to generalize beyond labeled data, capturing latent emotional structures within the graph. Through this integration of sentiment graph and self-supervised learning, our approach advances Weibo sentiment analysis, offering a robust method for understanding the complex emotional landscape of social media.

In this paper, we aim to improve sentiment analysis on weibo, a vital platform for sentiment and opinion expression. Current sentiment analysis methods struggle with the unique challenges of Weibo posts, which often contain informal language, sarcasm, and lack of context, making it difficult to capture complex emotions. Additionally, existing text-based models fail to account for visual cues and rely heavily on predefined emotion categories, limiting their ability to detect nuanced sentiments. To address these issues, we propose a novel multimodal approach that integrates both textual and visual elements, leveraging a ppretrainedext-to-image model to generate images from posts. Integrated with a self-supervised contrastive learning objective, this approach empowers our model to extract knowledge from unlabeled data and achieve enhanced comprehension of nuanced emotional cues. Our method enhances sentiment analysis on microblogs by introducing a framework that captures both text and synthesized visual cues, enabling deeper sentiment interpretation. By incorporating self-supervised learning, we significantly improve the detection of complex emotions, particularly those related to mental health, offering a more comprehensive solution for sentiment analysis on social platforms.