Perception Evaluation of Coastal Landscapes Based on Multi-Source Fusion Strategy

With the widespread adoption of social media platforms like Weibo and Rednote, vast amounts of user-generated content have accumulated on these platforms. Analyzing the sentiment orientation of such texts holds significant practical value and utility for evaluating scenic areas. This study utilizes social media text data as its source to analyze visitor sentiment toward Xuejia Island Scenic Area. A research framework that integrates multi-source sentiment dictionaries, multi-level rule engines, and machine learning for model training was constructed. Sentiment analysis was performed using fastText, which was trained on multi-source data, with comparative evaluations conducted on the resulting models. Experimental results show the fastText-trained model achieves recall, precision, and F1-score of 0.9187, 0.9166, and 0.9176, respectively. This framework enables relatively accurate sentiment analysis discrimination, markedly boosting classification performance. The framework yielded 2,384 positive texts, 1,649 neutral texts (emotion-neutral), and 700 negative texts.