基于属性级情感分析的公园生态系统文化服务感知特征 研究

A Study on Perceptual Characteristics of Cultural EcosystemServices in Parks Based on Aspect-based Sentiment Analysis

关键词:  风景园林  生态系统文化服务  城市公园  属性级情感分析  大语言模型

Key words:  landscape architecture  cultural ecosystem services  urban park  aspect-based sentiment analysis  large language model

摘要:

城市公园是公众获取生态系统文化服务的主要来源，识别公众对于城市公园生态系统文化服务感知特征，对于提升公园的文化服务供给能力与公众幸福 感具有重要意义。然而，当前关于公众对生态系统文化服务感知特征的研究仅局限于服务类型层面，未能探究与实际建设紧密联系的要素感知特征。基于此，从 大众点评、携程旅行获取评论文本数据，运用大语言模型进行属性级情感分析，量化了公众对于不同类型公园的生态系统文化服务感知特征与差异。在此基础上 进一步探索了公众对于不同类型公园中生态系统文化服务感知要素的偏好与差异。结果表明：1)基于大语言模型的属性级情感分析能够更为准确地量化公众对于 生态系统文化服务感知要素的情感倾向；2)公众对于不同公园类型的生态系统文化服务感知要素偏好存在显著差异与共性；3)除物质要素外，行为活动、精神感 受等非物质要素也是公众偏好的重点。研究结果为成都市公园城市建设、提高公园服务品质提供了建议

Abstract:

Urban parks are among the most important spatial settings through which residents experience cultural ecosystem services (CES), including recreation, aesthetic appreciation, social interaction, spiritual enrichment, and other non-material benefits derived from human-nature relationships. Accurately identifying how the public perceives these services is therefore essential for improving the cultural service provision of urban parks, enhancing park quality, and promoting public well-being. Existing studies, however, have generally examined public perceptions at the broad level of CES categories. Although this approach helps reveal which service types receive greater public attention, it provides limited insight into the specific perceptual elements that shape these experiences, and that can be directly connected to park planning, design, construction, and management. As a result, how particular material settings, behavioral activities, and psychological experiences influence public evaluations of CES remains insufficiently understood. To address this limitation, this study develops an element-oriented framework for analyzing public perceptions of CES in urban parks. Online review texts were collected from Dianping and Ctrip, two major platforms on which visitors record their experiences and evaluations of destinations and recreational spaces. A large language model was employed to conduct aspect-level sentiment analysis of the review corpus. Rather than assigning a single overall sentiment label to each review, the method identifies the specific CES-related elements mentioned in a text and determines the sentiment orientation associated with each element. This procedure makes it possible to distinguish multiple objects of perception and multiple emotional attitudes expressed within the same review. On this basis, the study quantifies the characteristics and differences of public perceptions of CES across different types of urban parks and further investigates the similarities and variations in public preferences for CES perceptual elements among these park types. The analytical focus is thus shifted from the general question of which CES categories are perceived to the more practice-oriented question of which specific elements contribute to positive or negative public experiences. The results show, first, that aspect-level sentiment analysis based on large language models can more accurately quantify public sentiment tendencies toward perceptual elements of ecosystem cultural services. Second, public preferences for CES perceptual elements exhibit both significant differences and notable commonalities among different park types. Different types of parks provide distinctive environmental settings, opportunities for activities, and experiential contexts, which are reflected in the elements emphasized by visitors and in the emotional evaluations associated with them. At the same time, several recurrent concerns appear across park types, indicating that certain dimensions of park experiences are widely valued by the public. Third, public preferences are not confined to material elements such as physical environments, landscape features, and facilities. Behavioral activities and non-material dimensions, including opportunities for participation, interaction, relaxation, emotional experience, and spiritual fulfillment, also constitute important components of public preferences for CES. By integrating social media review texts, large language models, and aspectlevel sentiment analysis, this study provides a more fine-grained approach to measuring public perceptions of CES in urban parks. It extends CES research from the identification of broad service types to the analysis of specific and actionable perceptual elements, thereby establishing a closer connection between public experiences and park planning, construction, and management. The findings provide an empirical basis for advancing Chengdu's park city development, formulating differentiated improvement strategies for different park types, and enhancing the overall service quality of urban parks.