Addressing the research gap in comprehensive scenic area quality evaluation is crucial in today’s evolving tourism industry. Existing approaches rely heavily on sentiment analysis, capturing general visitor sentiment but failing to reflect fine-grained aspects. To address this limitation, we introduce the first large-scale, multi-view dataset for Multi-view Scenic Areas Quality Evaluation (MSAQE), consisting of 291,714 comments labeled across eight specific quality aspects: business management, excursions, hygiene, post and telecommunications, tourism transportation, travel safety, travel shopping, as well as resources and environmental protection. Meanwhile, we propose an innovative data-driven framework integrating Reference-Based Sentiment Analysis (RBSA) and Global and Local Ensemble Encoding (GLEE) based multi-label classification. RBSA employs a fine-tuned ALBERT model to generate sentiment scores for comments, followed by similarity calculations with a reference comment set, resulting in more accurate sentiment evaluations. Since sentiment analysis alone cannot fully assess quality, we integrate GLEE-based multi-label classification to evaluate eight specific quality aspects more comprehensively. Experimental results confirm our framework’s superiority over existing sentiment analysis and multi-label classification benchmarks. The dataset and code are available at https://github.com/Yajie-good/MSAQE .