Objective  Urban public green spaces are significantly valuable in terms of ecosystem services, including cultural ecosystem services (CES) featured by high sensitivity and high-level participation, which can well cater to people’s aesthetic preferences and spiritual needs. A meticulous evaluation of the cultural service level of urban public green spaces and the proposal of renovation suggestions for specific landscape areas can effectively address the fundamental cultural requirements of the public and improve the life quality of residents. However, the potential of large-scale participatory social media platforms with extensive user engagement and abundant comment information remains largely untapped for perception evaluation of urban riverside green spaces, which primarily arises from the absence of geographical location information associated with the comments on social media platforms. Consequently, there is currently no established scientific and systematic method for evaluating CES in urban riverside green spaces. Additionally, the evaluation results of CES have not been effectively linked to the actual activities of visitors, presenting challenges in providing clear and specific guidance for site planning and renovation. Therefore, there is an urgent need for a novel approach that can effectively extract valuable information from comment data on large-scale participatory social media platforms, associate evaluation texts with relevant research site information, facilitate the organization of comment data, expand the pool of data sources for CES evaluation, explore the correlation between cultural service level and site usage, and enhance the scientific and rational decision-making process in urban planning.

Methods  Taking the riverside green space in the urban section of the Hutuo River as an example, this research presents a novel approach to evaluate the CES of urban riverside green spaces. The research methodology comprises six distinct steps as outlined below. 1) Collect comment text data from the research site by Python web crawler, and then conduct word frequency analysis to identify frequently used vocabulary associated with the site, based on which a location information lexicon is constructed. 2) Analyze segmented comment texts to extract pertinent information concerning scenic spots and land use, utilizing the established location information lexicon as a reference. 3) Establish a comprehensive CES indicator system, and develop a site CES lexicon through word frequency statistics to extract CES evaluation information from segmented comment texts. 4) Encode scenic spot information and land use data to associate comment texts with specific site spaces for spatial perception mapping of CES. 5) Calculate the kernel density of trajectory data to characterize the level of usage of a certain area of the site. 6) Adopt the method of spatial autocorrelation to investigate the relationship between the overall value of cultural services provided by the site ecosystem and the actual usage of the site. The findings derived from the analysis above can help provide recommendations for site renovation and design.

Results  The findings of this research reveal that the urban riverside green space under research possesses a strong capacity to offer aesthetic experience, sightseeing opportunities and recreational services. However, there is untapped potential for further improvement in the areas of science education and spiritual satisfaction services. Specifically, four distinct categories of areas have been identified on the website: high-service high-usage hotspots(HH), low-service low-usage cold spots(LL), high-service low-usage conversion areas(LH), and low-service high-usage promotion areas(HL). The high-service high-usage hotspots, referred to as HH gathering sites, represent sites of exceptional quality within the research area. It is recommended to maintain their current status and ensure their proper functioning. Conversely, the low-service low-usage cold spots, known as LL gathering sites, exhibit relatively low levels of cultural service and usage. Therefore, renovation and improvement of supporting facilities should be prioritized during the planning and construction of these sites. The high-service low-usage conversion areas, labeled as HL gathering sites, demonstrate a high cultural service level while low usage level. In future planning and construction, it is advisable to leverage the existing social appeal of these sites to enhance the accessibility and optimize the openness of public spaces. The low-service high-usage promotion areas, referred to as LH gathering sites, exhibit relatively low cultural service level despite their high usage level. Therefore, in the process of planning and construction, emphasis should be placed on the development of distinctive landscapes for these areas.

Conclusion  This research represents a valuable contribution to the expansion of data sources for evaluating CES through adopting a novel approach based on social media text. Through the utilization of a multi-level CES perception method based on landscape feature that integrates word frequency analysis and text segmentation technology, the accuracy of CES evaluation is significantly improved. This approach can effectively capture the cultural service perception evoked by the actual site, making it able to comprehensively understand user evaluations. Moreover, the proposed approach can facilitate the visualization of user perceptions concerning open riverside green spaces, enabling a more precise integration of social media text and landscape features. Therefore, this research can provide valuable guidance for the transformation and enhancement of urban riverside green spaces, thereby contributing to the overall improvement of their cultural ecosystem services.