Exploring the Importance of Destination Attributes of Sustainable Urban Waterfronts: Text and Data Mining of Tourists’ Online Reviews
1. Introduction
Rapid economic development creates many challenges for cities; among these, balancing ecological, economic, and social sustainability is one of the most urgent [1,2,3]. Polluted rivers and damaged natural systems in urban areas have led to the realization that waterfront environments not only create urban ecological diversity [4] but also improve the climate of inland cities that are increasingly affected by climate change [2]. Urban waterfronts are parts of towns and cities that are adjacent to bodies of water, such as rivers or seas [5,6]. They are functional and interactive spaces that link two different systems: land and water [7,8]. Scholars have noted the benefits of successful urban waterfront redevelopment (e.g., forging connections between local residents and visitors and providing spaces for education, healthy living, and wellbeing) [9]. Given the role of sustainable development goals (SDGs) in the development of sustainable cities [10,11], attention is being paid to improving and enhancing the environmental and landscape quality of existing waterfront areas [2,3] and fully using them to provide recreational and tourism functions [12,13] to satisfy the diverse needs of their users. As part of efforts to enhance the quality of urban ecological and social environments, public agencies are looking to make waterfronts attractive tourist destinations in order to stimulate local economic development [13]. Sustainable urban waterfront districts are places where people of all ages and backgrounds can live, work, play, visit, and learn in ways that enhance and celebrate the natural environment and the beauty, diversity, economic vitality, and creativity of a city [8].
Previous studies on urban waterfront destinations have focused on tourism planning [14], tourist experiences [13], and landscaping [7,15]. Few studies have attempted to understand the relationship between tourism and SDGs from the perspective of how sustainable urban waterfronts attract tourists. This question is becoming important given the positive impacts that influencing positive tourist behavior has on economic sustainability (e.g., positive evaluations, word-of-mouth (WOM) intentions, and behavioral intentions) [16,17]. Furthermore, understanding the economic, environmental, and social impacts of tourism requires a detailed examination of a wide range of data [18].
Destination attributes refer to the characteristics and features of a destination that satisfy tourists’ needs [19]. Information about these characteristics and features influence tourists’ behavioral intentions through psychological processes (i.e., perceptions, feelings, and attitudes) [20]. Such information could be obtained through WOM (i.e., face-to-face exchanges of information between tourists’ friends and relatives) [21]. However, with the emergence of the internet, online user-generated content (UGC) and consumer-generated media [21,22], especially comments and ratings, have stimulated information exchange among a much wider range of potential tourists. Well-known travel and search websites such as TripAdvisor, Google Maps, and Lonely Planet provide a large number of online reviews and ratings that could reshape potential tourists’ ideas about a destination, influencing or modifying their destination choices [23]. Scholars have claimed that reviews on these websites represent a more general and widespread form of WOM, namely electronic WOM (eWOM) [24,25], and that the ratings represent customers’ or tourists’ overall satisfaction with the quality, service, or experience of a product or destination [26,27]. For sustainable destination development, eWOM might be more cost-effective [24,28] and influential [29,30] than other sources of information because of the intangible nature of tourism services [31].
Scholars have argued that online reviews could be used as an alternative data source for evaluating sustainable tourism [18]. In the era of big data, tourists leave downloadable qualitative data on the internet through online reviews, photos, and other forms of interaction. However, the influence of online reviews and feedback on destinations offered by public agencies or destination management/marketing organizations (DMOs) has not been explored [32]. Current UGC studies [33,34] have attempted to understand this impact by applying two different approaches (i.e., structured or unstructured), to capture tourist perceptions or awareness [30]. The structured approach relies on quantitative ratings of individual attributes to understand tangible and emotional attributes (e.g., [35]), whereas the unstructured approach relies on open-ended questions that allow respondents to use their own words to freely describe their feelings and perceptions of a destination [36]. The latter approach has been increasingly emphasized by researchers in recent years. However, one of the key challenges is extracting meaningful insights from the large amount of shared text and finding possible patterns and models in this unstructured information [37]. Sparks and Browning [38] have suggested that combining structured quantitative ratings with unstructured textual data could yield a more complete evaluation of users’ online responses and, hence, more useful research results. However, to the best of our knowledge, except for Moro, Rita, and Coelho [39], few researchers have attempted to combine these two types of data to improve our understanding of destination attributes and online ratings in tourist reviews.
This study aims to fill this research gap by exploring and categorizing the most important attributes of sustainable urban waterfront destinations using text mining. Furthermore, data mining techniques are employed to identify the attributes that best predict tourist ratings based on stimuli–organism–response (S–O–R) theory, and the relationships among the attributes. In this theory, the important attributes act as the stimulus (S), tourists as the organism (O), and the ratings for the response (R). These results have useful implications for the planning and marketing of sustainable urban waterfront destinations, sustainable goal management, and marketing strategies of public agencies and DMOs.
3. Methods
The two study sites were the Liuchuan Waterfront Trail and the Luchuan Waterfront Corridor in Taichung city, Taiwan (Figure 1). We chose Taichung as the setting of the study because it is the second-largest city in Taiwan, with millions of tourists every year. In recent years, Taichung’s water resource rehabilitation has been internationally recognized (e.g., Japan’s Good Design Award (Best 100) and the FIABCI World Prix D’excellence Awards) [87]. In addition, we chose to study these waterfront destinations because one of them, Liuchuan, is one of the four major rivers in the center of Taichung city [88,89]. As the city developed, the water quality of these two rivers became polluted due to wastewater discharge, and their ecosystems deteriorated, eroding the urban waterfront environment and negatively affecting the city’s image [90]. The Liuchuan river waterfront was improved in 2015 by installing a fully underground water purification plant to treat domestic sewage; the quality of the river’s water environment was improved through sustainable ecological remediation [88,89]. In addition, extending the surrounding green belt connected the riverbank space and facilitated flood control, water quality improvement, and landscape creation to provide tourists with a more natural and ecological experience [90]. The scenic riverbank has started to become an emerging popular waterfront destination in urban areas, with pleasant scenery during the day and bright lights at night, combining aesthetics and recreational use [87,88]. The Luchuan waterfront is the highlight of Taichung’s “Little Kyoto” project dating to the Japanese occupation period in 1900 and was one of eight major nocturnal scenic areas in Taichung at that time [88,89]. The improvement of the water environment prioritized water quality, flood control, and the creation of ecological habitats [90]. To provide tourists access to the river, the surrounding recreational green space was expanded and developed into an inclusive playground, the “Luchuan water purification park”, which is to become a shaded river walkway with a people-friendly, scenic riverbank to attract tourists [89]. Thus, the remediated waterfront landscape was ascribed a new meaning and symbol, relying on the provision of recreational and tourism functions as key elements of its reshaping into a visually appealing and sustainable urban waterfront destination [90]. In addition, these elements have the potential to generate large amounts of data.
The data for this study were obtained from searchable and freely downloadable online sources, namely 2765 Chinese comments posted by domestic and international tourists on TripAdvisor (https://www.tripadvisor.com.tw/ (accessed on 4 November 2023), search keyword = 柳川藍帶水岸, 新盛綠川水岸廊道) and Google Maps (https://www.google.com.tw/maps/ (accessed on 4 November 2023), search keyword = 柳川水岸景觀步道, 綠川水岸景觀步道) relating to the Liuchuan Waterfront Trail and Luchuan Waterfront Corridor in 2018–2023, which were randomly collected over a period of three months; the collected data were used to construct the variables needed to establish the model [91]. Tourist reviews were selected based on their completeness, usefulness in terms of information, and variability across travel seasons. There were many features to choose from in online reviews; TripAdvisor and Google Maps’ online tourist reviews each have data relating to multiple variables. For this study, we selected destination attributes in online reviews (nominal variable: 1 = not mentioned, 2 = mentioned) as the independent variables in the model, and tourists’ ratings (1 = not very satisfied, 5 = very satisfied) were treated as the dependent variables. However, collecting all of the attributes made it difficult to analyze how each feature affected tourist ratings in the subsequent data-mining model using a support vector machine (SVM). Therefore, the number of destination attributes and their categorization were determined in this study based on word frequency and discussions between two experts on destination attributes and sustainable development [91]. SVM is a data-driven approach that attempts to minimize the upper bound of the prediction error and has better prediction results [92], which helps clarify the relationship between destination attributes and tourists’ ratings in this study (Figure 2).
5. Discussion and Implications
This study used data mining techniques on tourists’ online reviews to understand the relationship between destination attributes of sustainable urban waterfronts and tourists’ online ratings. The results show that tourists’ reviews mentioned both tangible and intangible destination attributes, relating to both sustainable landscapes (i.e., aesthetics, water resource rehabilitation, sustainable lighting, emotional experience, and LID waterfronts) and sustainable recreational spaces (leisure activities, festivals, inclusive destinations, photography, and tourist experience). Two destination attributes were common to both categories: nightscape and waterfront landscape. These 12 attributes were the best predictors of how tourists rated sustainable urban waterfront destinations. Furthermore, sensitivity analyses showed that sustainable landscape-type attributes had a greater impact on tourists’ ratings than sustainable recreational space-type attributes. In addition, three important rules of association between the two types of destination attributes were identified. These findings suggest that public agencies should consider tourists’ needs for recreation and landscapes to provide a unique and meaningful tourism experience when planning and designing sustainable urban waterfront destinations. When satisfied, tourists’ needs could become a competitive advantage for sustainable waterfront destinations and help realize sustainable economic development. The following section identifies the key contributions of this study.
Tripadvisor and Google Maps are two of the largest online review platforms used by tourists; their online reviews influence the perceptions and choices of potential tourists. The destination attributes identified in online reviews are often related to tourists’ perceptions and satisfaction [64]. Although tangible attributes could be easily defined by researchers and assessed by tourists, the collection of intangible attributes, such as emotional experiences, is more difficult to identify because they are generally subjective and expressed in various terms during the travel period [16,19]. This study relies on tourists’ ability to express their opinions, views, and feelings about a destination in their own words in free-text online reviews. The results yielded a list of tangible and intangible attributes. The sustainable landscape-type attributes were similar to those identified by [2,4]. They suggested that the protection of the water environment and the transformation of emerging consumer landscapes into landscapes meant for new uses contribute to the establishment of urban waterfront sustainability. The sustainable recreational space-type attributes are similar to those found by Xie and Gu [3], Aiesha and Evans [12], and Griffin and Hayllar [13]. These attributes suggest that waterfront areas are finite, non-renewable, and valuable natural assets that attract and provide recreation for both urban residents and tourists [8,40]. Thus, this study makes a valuable contribution to the SDGs, specifically targeting SDG 15 (Life on Land) and SDG 11 (Sustainable Cities and Communities), with a focus on urban waterfront destinations. Organizations responsible for managing urban waterfront destinations can enhance tourist satisfaction and further encourage repeat visits by implementing of sustainable management strategies. This becomes particularly crucial in addressing the lasting deprivation of urban waterfronts and informing tourists’ decision-making processes. The destination attributes common to both types were nightscapes and waterfront landscapes. This is similar to the study by [40], who recognized that urban waterfront destinations perform well in landscapes, nightscapes, and walkway spaces.
This finding follows Kostopoulou’s [99] observation that sustainable urban waterfronts have the potential to serve as creative environments that attract tourists. As Giovinazzi and Moretti [7], Rahana and Nizar [100] have shown, although the consequences of sustainable urban waterfront development are related to the original characteristics of the host city, other factors (water and environmental quality, mixed use, and recreational development) are necessary considerations when achieving sustainability in urban waterfront development. Against this background, by clearly identifying important destination attributes in tourists’ reviews, the second main contribution of this study is that its findings show how public tourism agencies need to shift their reliance on traditional data collection methods (e.g., questionnaires and surveys). The opinions and perceptions of tourists about sustainable waterfront destinations in online reviews are seldom collected or analyzed by public agencies [13]. Different data collection methods also contribute to S–O–R theory. In this study, destination attributes were collected through online reviews, which are unstructured data used to extract environmental stimuli and evaluate responses to these stimuli. This is in contrast to past studies [53], wherein S–O–R was understood in terms of structural information (e.g., questionnaire surveys). Social media information has become an important reference channel for travel decision-making. In the future, if S–O–R research is combined with text mining to explore environmental and behavioral relationships, more marketing information and market potential could be obtained.
TripAdvisor and Google Maps users provide online reviews in two main ways: by entering a text review in the free-text area and by offering a quantitative rating between one and five. Textual reviews often contain interesting but hidden user sentiments; however, relatively little research has linked ratings to textual reviews. Therefore, this study conducted knowledge extraction by modeling tourists’ ratings on TripAdvisor and Google Maps to examine how each destination attribute in reviews affected rating scores. The results show that the 12 attributes of sustainable urban waterfront destinations significantly predicted tourist ratings. Thus, this study concludes that tourists’ perceptions of destination attributes, which contribute to memorable travel experiences [55], are the best predictors of their ratings of sustainable waterfront destinations in online reviews. Another major contribution of this study is its determination of how perceptions of destination attributes affect tourists’ online ratings of sustainable urban waterfront destinations and to understand what motivates tourists to post ratings for them. These findings are interesting because, unlike previous research that analyzed a small number of destination attributes [55,61], this study proposed a single valid model based on many destination attributes.
In addition, the sensitivity analysis we conducted identified correlations between attributes, which helps explain how each attribute affects ratings on TripAdvisor and Google Maps. The results show that sustainable landscape-type attributes had a greater impact on tourist ratings than sustainable recreational space attributes. Thus, to promote and to realize more effective sustainable economic development and competitive advantage, public agencies could use these findings (i.e., tourist demand for sustainable landscapes and recreational spaces) as a basis upon which to plan and design sustainable urban waterfront destinations.
Finally, to further understand the destination attributes frequently mentioned in tourists’ online comments, the results identified three important rules of association that link sustainable landscape-type attributes to sustainable recreational space attributes. These findings imply that tourists increasingly demand unique and meaningful travel experiences that are generally associated with intangible destination attributes [101]. In addition to the tangible planning of infrastructure, urban waterfront planning also includes an intangible dimension, which could be a place of/about and for cultures, so that dynamic spatial changes can provide people with knowledge regarding changes in traditional culture and diversity [43]. The planning and design of sustainable urban waterfront destinations should focus on a small number of combinations of tangible and intangible destination attributes instead of adopting an all-encompassing strategy [102]. A final major contribution of this study is the identification of a key set of destination attributes that should be the focus of sustainable urban waterfront development.
6. Limitations and Further Research
Despite its contributions, this study has the following limitations. First, tourists’ general perceptions of Taichung’s sustainable urban waterfront destinations might affect their beliefs regarding the relative importance of destination attributes. For example, past research has found that the location of waterfront destinations affects tourists’ ratings [63]. Thus, the wider context of tourists’ comments and ratings might have biased the text and data mining approach and affected the rankings of destination attributes and their correlation with ratings. Given that this study only collected comments related to Liuchuan and Luchuan in Taichung to confirm the model’s generalizability, further research in other locations is required. In addition, the attributes of the 12 destinations were based solely on the opinions of two experts. For example, future research could use (hierarchical) k-means to determine the optimal number of clusters in different settings, resulting in different (non-overlapping) attributes. Second, tourists’ online reviews in this study are limited to TripAdvisor and Google Maps in Taiwan; future research could incorporate reviews from other websites in other countries. The data collected online could be subdivided according to the country of origin. Therefore, future research should compare the attributes identified in Taiwan’s tourist reviews with those of other countries to understand how sustainable urban waterfront destination attributes attract tourists from other countries. Third, as not all reviews have the same impact on potential visitors, concepts such as reviewer trust or helpfulness could be used to categorize reviews [30]. Because both of these concepts could be collected from websites, future research could focus on the highest-scoring reviews (as rated by other members of the community) or reviews posted by trusted reviewers. Finally, considering the ability of the SVM to decompose relationships between many different attributes, there is a possibility that the pattern might contain more attributes from other sources [39]. Future research could be conducted at different locations or with different attributes to understand how to minimize errors in the identified pattern or how to adjust the pattern to improve prediction. Finally, as many scholars have pointed out, there are still some limitations of the traditional text mining techniques used in this study (e.g., [103]), considering the fact that there are more advanced artificial intelligence (AI) models for textual understanding and topic extraction. It is suggested that future work could explore other text embedding approaches to improve the quality of data processing.
