To address data lag, subjectivity, and challenges in tracking emotional-spatiotemporal dynamics in traditional smart volunteer service evaluation, this study proposes a quality assessment system integrating spatiotemporal sentiment analysis. The system employs a self-attention mechanism for deep fusion of multimodal data (text, images, and speech) and develops a cross-modal spatiotemporal analysis method using graph convolutional networks. Implemented on a cloud platform, it enables real-time service monitoring and dynamic evaluation. Testing achieved an 89.0 F1-score and 48.0% accuracy on CMU-MOSI dataset. Ablation experiments on Yelp datasets showed Hit Rates of 44%/23% (k = 10) and 74%/48% (k = 50), with NDCG values reaching 28%/13% and 38%/18%, respectively. The system demonstrated superior sentiment analysis precision and assessment reliability, offering an intelligent solution for optimizing volunteer service management and decision-making. This advancement promotes the evolution of intelligent, precise evaluation frameworks in volunteer services.