In urban green spaces, ecosystems not only fulfill the ecological function of sustaining the urban environment but also provide an important living area for urban residents. The concept of ecosystem services conceptualizes human environmental interactions through a series of linked components that relate ecological processes to human well-being (MA 2005; Kosanic and Petzold 2020). As evidenced by numerous studies, urban green spaces provide the inhabitants with various ecosystem services (Bolund and Hunhammar 1999; Kremer et al. 2015). For example, regulation services may help urban residents by filtering air pollutants and mitigating flooding (Aram et al. 2019; Grote et al. 2016; Prudencio and Null 2018). Thus, an adequate management of urban ecosystem services is essential to preserve the urban environment (MA 2005; Luederitz et al. 2015; Tzoulas et al. 2007; Zhang and Ramírez 2019).
Cultural ecosystem services (CES) are defined as “all the non-material outputs of ecosystems (biotic and abiotic) that affect physical and mental states of people” (Haines-Young and Potschin 2018). CES are not only the functions provided for humans by the natural ecosystem, but also the interaction between humans and the environment (Oteros- Rozas et al. 2017; Pen ̃a et al. 2015). It plays an important role in the mental and physical well-being of residents based on their activities in the natural environment (Filho et al. 2020; Plieninger et al. 2013). In the urban ecosystem, the demand for ecosystem services has greatly increased, and CES are important due to the frequency and impact of human use (Ko and Son 2018; Wilkerson et al. 2018). At the city level, urban ecosystems such as parks, mountains, and beaches provide scenic beauty and a refuge from everyday busy life (Chen et al. 2019; Martín-López et al. 2018; Schnell et al. 2019; Subiza-Pérez et al. 2020).
Regulating and provisioning services are the most commonly examined aspects of urban ecosystem services (Luederitz et al. 2015; Martínez-Harms and Balvanera 2012). Moreover, urban green spaces are often evaluated for their economic value more than their cultural value, though the latter is more important to residents than the former. (Christie et al. 2012; D’Amato et al. 2016; Spangenberg and Settele 2010). CES are often undervalued compared with other ecosystem services due to the subjective nature of individuals’ perceptions, which presents a challenge for quantitative assessments (Cheng et al. 2019; Lee et al. 2020; Luederitz et al. 2015; Riechers et al. 2016; Stålhammar and Pedersen 2017; Tilliger et al. 2015).
To provide quality CES through limited urban green spaces, it is necessary to understand users’ perception of the CES provided by the city’s green spaces (Andersson et al. 2015; Dickinson and Hobbs 2017). Various ecosystems such as forests, agricultural lands, rivers, lakes, parks, and roadside trees exist in urban areas, and each serves a distinct function. Additionally, CES are not uniformly embedded in all green spaces, but vary based on the type of green space (Plieninger et al. 2013; Ko and Son 2018). Previous studies have shown that differences in CES can exist depending on the characteristics of green areas constituting the urban ecosystem (Beninde et al. 2015; Dade et al. 2020; Threlfall and Kendal 2018).
Socio-cultural approaches are useful for exploring user perception and preferences for CES. Numerous methods have been used to evaluate CES, including questionnaires, photographic analysis, and visitor-employed photography with short comments, with the most common method being the perception survey method, which has temporal and spatial limitations (Haase et al. 2014; Leetaru et al. 2013). This method is most suitable for small regional-scale studies, however, it has limitations in large-scale regional studies (Bragagnolo et al. 2016; Gosal et al. 2019, Paracchini et al. 2014; Peichao et al. 2019).
As an alternative to the perception survey method, studies have increasingly used big data, which can improve existing empirical research methods and identify a variety of previously unknown values based on user-created content (Gosal et al. 2019; Hu et al. 2015; Mckenzie et al. 2013; Scholte et al. 2015). Spontaneously generated social media data differ from data collected for research in a structured manner (Heikinheimo et al. 2017; See et al. 2016; Sonter et al. 2016). The subjective intervention of the researcher can be reduced through social media data, as its collection does not involve direct contact between the researcher and participants (Chen et al., 2018). Such data may provide information on various users’ experiences and preferences over a brief period, which may contribute to improving other empirical methods (Hausmann et al. 2017; Richards and Friess 2015; Yoshimura and Hiura 2017).
Additionally, data in social media include attached metadata, such as a title, descriptions, time stamp, and photo tags; such data contain a considerable amount of information from which interactions with nature or space use can be inferred (Dorwart et al. 2009; Hollenstein and Purves 2010; Johnson et al. 2019; Runge et al. 2020). Some research has confirmed public awareness of ecosystem services and a correlation between landscape features and CES using photographic geographical information (Dunkel 2015; Gosal et al. 2019; Oteros-Rozas et al. 2018; Vigl et al. 2021). Other studies have proposed methods of socio-cultural usage indicators, including tags and other text associated with photos on social media (Barry 2014; Ghermandi et al. 2020; Hamstead et al. 2018; Hollenstein and Purves 2010; J. Retka et al. 2019; Lee et al. 2019). Social media posts provide an information-rich source through which one can not only understand the diverse ways in which people interact with landscapes, but also identify the expressions used to describe CES (Hale et al. 2019, Lee et al. 2020). Some studies have used these newly available data to maximize reliability and validity (Brown et al. 2014; Johnson et al. 2019; Karasov et al. 2020). Previous studies have shown that social network data are similar to data obtained from traditional survey methods (Johnson et al. 2019; Keeler et al. 2015; Richards and Tunçer 2018; Sonter et al. 2016; Wood et al. 2013). Many researches have focused on social media platforms offering easy data access such as Flickr. However, the efficacy of this data differs based on users and the type of social media platform itself (Heikinheimo et al. 2020; Manikonda et al. 2016; Ruiz-Frau et al. 2020; Tenkanen et al. 2017).
Some studies have discussed the representativeness of social media (Li et al. 2013; Liu et al. 2016) and methodological issues related to the representation of CES by crowd sourced data (Zhang et al. 2020). However, some platforms and social media posts make it difficult to understand users’ perception of CES and what they did in the urban green space because photographs and tags do not provide those data (Hale et al. 2019; Richards and Tunçer 2018).
Text, which is unstructured data, accounts for 80% of the data in the world. (Chakraborty and Pagolu 2014; Gantz and Reinsel 2012). However, existing studies have used short sentence limits from social network services (SNSs) (Di Minin et al. 2015; Figueroa-Alfaro and Tang 2017; Johnson et al. 2019; Tenkanen et al. 2017). No existing studies, to our knowledge, has used a character limit from a data-centric approach to quantify CES using long sentences. Long user-written texts can contain important evaluations of products or services, and in some cases provide insightful and important information (Jo et al. 2018; Johnson et al. 2019).
In this study, we examined the relationship between the meanings of words contained in social media posts through natural language processing, to understand user perceptions of CES provided by urban green spaces. To do this, we first extracted keywords representing CES through text network analysis and identified the characteristics of CES for each type of urban green space. This was followed by analyzing the relationship between the keywords and greenery types derived from text mining. We then aimed to determine the possibility of distinguishing between differences in cultural services based on the type of urban green area, through atypical expressions.