<p>Waste sorting is crucial in reducing greenhouse gas emissions and minimizing resource waste. Social media platforms, with their global reach and real-time communication, serve as effective tools for disseminating waste sorting information (WSI). However, the impact of sentiment on the dissemination of WSI has not been thoroughly explored. This study, grounded in information ecology theory and the elaboration likelihood model, proposes a dual-path analytical framework to investigate the interaction between sentiment and information, user, and environmental attributes in shaping forwarding behavior. Utilizing a dataset of 443,358 WSI from Sina Weibo, sentiment is identified using a BERT-based classification model enhanced with emoji cues to detect positive, neutral, and negative expressions. Topic types were performed through a human-AI collaborative approach that combines BERTopic with ChatGPT-4o, guided by the theory of planned behavior, to enhance semantic coherence and theoretical alignment. The findings show that neutral sentiment is more likely to be shared than either positive or negative sentiment. The sentiment effect weakens with greater information richness. High-activity and high-authority users are less likely to share negative content, whereas media and organizational accounts are more likely to spread it. Additionally, topic types modulate sentiment effects, with neutral sentiment performing best in public participation topics, and positive sentiment being more effective in knowledge and policy content. This study highlights the need for precision-driven emotional communication strategies that combine multimodal content design with differentiated user engagement, leveraging sentiment-channel-topic alignment to enhance the reach and resonance of WSI.</p>