Insights into People’s Perceptions Towards Urban Public Spaces Through Analysis of Social Media Reviews: A Case Study of Shanghai ^†

Abstract

Urban public space is a crucial constituent of livable city construction. A pleasant and comfortable public space is not simply spacious, bright, and accessible but also subjectively preferred by citizens who use it. Efforts to understand how citizens experience and perceive therein thus matters and would significantly aid urban design and well-being improvement. This research constructs a perception lexicon for 129 sites of public street space, a significant type of public space, in Shanghai and identifies how citizens comment on these sites through sentiment analysis based on social platform texts. A Chinese natural language processing (NLP) tool is applied to sort out the extent of citizens’ feelings on the urban street environment through a 0–1 scoring system. Six types of built environment elements and five categories of urban public spaces are identified. Pleasantly perceived sites primarily locate in the urban center and sporadically distribute in the outskirts and are normally “high-density” and “multi-function” in nature. Among the five categories of urban public spaces, sites that are commercially dynamic with culture, arts, and historical elements or that have gourmet food and good walkability generally receive the higher sentiment scores, but scores of ancient town commercial streets (many are antique streets), once popular and contributing much to tourism economy, are not satisfactory. The NLP-based text analysis also quantifies the intensity of emotional perceptions toward the six types of built environment elements and their associations with the general perception. This study not only offers insights for designers and policy makers in public space optimization but also showcases a scalable, data-driven approach for integrating public emotional and experiential dimensions into urban livability assessments.

1. Introduction

Urban public spaces are built in diverse forms such as parks, green spaces, and streets. They are vital to promoting people’s health and local vigor [1,2], and thus are at the heart of city planners’, park managers’, and landscape architects’ actions in building for city livability. Livability can be measured by various factors that matter for quality of life, but people’s perceptions and feelings toward the city, especially the built environment, are considered fundamental [3,4]. How can we promptly evaluate the perceptions of the general public on urban public space, and how do their perceptions associate with particular built environment elements? This research attempts to explore these questions. Though measuring subjective perception through large number of samples is not easy, in recent years, with the development of new data and technology, research using social media data to evaluate people’s perceptions towards these public spaces has emerged [5,6]. Most of the studies, however, focus on parks and green spaces [7,8], with few looking into how people feel about different types of street environment. Indeed, many popular check-in spots are urban streets mixed with diverse built environment elements [9] such that the classification of park or green space alone may be insufficient to capture their livability. Urban streets are significant in shaping public life and are the traditional subject in urban planning studies. Classic research tends to provide nuanced, often qualitative analysis on how people interact with physical entities on the streets and how these interactions contribute to the richness of public life, which generates fruitful insights into human-centered design [10,11]. But this approach, normally based on small samples, would be a challenging way to learn the views of the general public. Social media platforms may supplement in this regard. Thus, this study focuses on urban public streets and applies check-in and review data from a social media app to measure the perceived quality of the street built environment from the general public’s perspective. Shanghai, a metropolis with an astonishingly high number of internet-famous streets in China, is selected as the study area for this research.

Urban streets are public in nature and a significant constituent in urban design [12,13]. A well-designed street has the potential to foster urban diversity and vitality, and thus improve livability, whereas unpleasant streets impair urban experiences and happiness [14]. Against the backdrop of the internet celebrity economy, urban streets are also the main check-in spots and popular culture gathering places. While different groups of people perceive urban streets differently [12], street characteristics also shape people’s perceptions in diverse ways. It is widely acknowledged among urban designers that appropriate commercial activities can improve vitality of the public streets. Cafés, retail shops, and convenience stores are classic design elements for creating a pro-business environment for social interaction [15]. Commercial streets thus are the focus of this research. Noteworthily, with the diversification of street design, streets also integrate a variety of features, forming differentiated street labels and categories. Arts and cultural icons are common internet celebrity factors that empower street vitality [16,17]. They often mix with historical elements to improve street attractiveness but are also likely to push up property price and elicit gentrification [18,19]. Historical elements could be protected remains, memorial halls, or any intangible cultural heritage, which assist design to enrich street experience. In China, some well-preserved ancient towns such as Lijiang or Zhouzhuang are also commercialized [20]; their economic success has led to widespread imitation and the emergence of antique streets. Walkability was introduced and has been stressed in the planning literature since the mid-1990s, as professionals and scholars believe that good walkability facilitates more walking and thus improves people’s physical and mental health [21]. Walkability plus outdoor dining represents a type of popular design to enhance street vigor and the interactive experience. There are also other stylish features, e.g., European-style facades, night markets, and creative elements, that are usually incorporated into the street design and planning to shape uniqueness and attractiveness. These features are often labeled in the street names or brief introduction, which help with identification. A street may have more than one feature, and the idea is to identify the major one for reclassification. In general, this research sorts over 200 street sites in Shanghai and categorizes them into five groups for follow-up analysis: they are commercial streets mixed with (1) culture, arts, and historical elements; (2) gourmet food and good walkability; (3) ancient towns; (4) stylish features and a night market; and (5) creative elements.

Researchers generally agree that livability is not simply about offering a spacious, bright, and accessible environment, but associating deeply with individuals’ perception and emotion [4,6,22]. The former can be objectively measured through, e.g., geographical information systems (GIS) or devices such a lux meter, while the latter often rely on respondents’ subjective judgments, which are not easy to collect [12]. To researchers and professionals, real-time consciousness and the felt quality of the built environment might be more convincing than objective physical environment parameters alone to capture livability [23]. This has been stressed in theories such as sense of place, which takes sensate place as a significant constituent in assisting urban designers to create pleasant public urban spaces [24]. The emergence and rapid development of mobile phones and the associated social media apps lend a hand. Social media data have increasingly become popular among researchers given their richness, high volume, and dynamically updated nature [25,26]. Popular social media platforms such as Facebook, Tripadvisor, Twitter, and Instagram collect photos, reviews, and videos from their users every day; these data are frequently applied to assist policy makers and analysts in understanding public opinions [27,28]. Different from traditional surveys that are usually designed with targeted questions and focused groups, user-generated, self-reported social media data are often characterized by a free-and-easy style without specific intentional purpose [29]. This, though it may impact data quality to some extent, enables a more realistic reflection on how the masses perceive the environment in a natural state. It is thus necessary to care about review quality when adopting social media data in a study.

The application of social media data in planning and design-related research has emerged in recent years. For instance, Yang Song and Bo Zhang adopt social media data, Instagram posts in particular, to understand how Seattle Freeway Park is used as a public space and in what way people recognize the built environment of the park [30]; similar research is also conducted in New York Bryant Park using Tripadvisor reviews, Shinsui parks in Tokyo using Twitter comments, and Chicago’s park management using Google reviews [31,32,33], etc. There are also numerous studies focusing on green space or neighborhoods. For instance, Zhifang Wang et al. examine people’s satisfaction toward green space at the fine-grained scale through machine learning of content analysis of Dianping, a popular app in China [34]; similar research is also conducted in assessing the seasonal variation of Birmingham’s 46 green space sites using Twitter reviews [35]. Gibbons et al. analyze twitter reviews to help forecast health status of resident population at a census tract level [36]. These studies, however, barely pay adequate attention to the street environments, which are the main constituents of urban public spaces shaping livability. In addition, while it is possible to record general perception with the analysis of social media review data, how specific built environment elements quantitatively associate with such perception is equally important but less understood so far. Early in the 1960s, Lynch proposed paths, edges, districts, nodes, and landmarks as five key elements shaping people’s perception on physical environment [10]. The street design literature also emphasizes features such as connectivity, arts, waters or fountains, greenery, squares and parks, sidewalks and pavements, stores and public facilities, etc., in the development of urban and community engagement [12,37,38,39]. Though follow-up studies have developed numerous ideas or concepts such as resilience and sustainability to establish a design paradigm thereafter, they are either too general or too formal to link people’s perceptions. Using small samples, an independent qualitative study is valuable, but it is difficult to present general public perception without bias. Given these reviews and the data availability of social media platforms, this research identifies six types of built environment elements encompassing major features in street designs for lexicon framework construction (see research method).

2. Research Methods

2.1. Study Area

Shanghai, one of the fast-growing dense cities in China, is well known for its dynamic urban environment and as the home for many popular social media platforms such as RedNote, Bilibili, Weibo, etc. By the end of 2024, Shanghai had a registered population of 14.873 million and a permanent population of 24.871 million, forming a large social media user base. According to the Statistical Report on the National Economic and Social Development of Shanghai in 2024, the city has 33.5 million mobile Internet users, the density of 5G base stations ranks among the top in China, and the penetration rate of social media exceeds 85%. Platforms such as Dianping and RedNote generate over 200,000 comments on urban environment per day. As an international, high-density metropolis, Shanghai’s urban public spaces are well built and popularly geo-tagged on social media. Many internet celebrity check-in spots are in the forms of urban streets, for instance, Wukang Road Historical and Cultural Street, the University Road Pedestrian Street, the Xuhui Riverside Pedestrian Street, etc. These streets are filled with commercial functions and are diverse in categories of urban public spaces, encompassing historical, creative, recreational, gourmet, leisure and stylish, and cultural and art types. Citizens and visitors coming to these streets generate abundant reviews each day, offering rich samples for analyzing people’s perception on urban public spaces. All these make Shanghai an ideal case for this study. As most popular public space check-in spots in Shanghai are located in the downtown area, which is normally delineated within the outer ring road, special attention is given to the area during analysis. This study also carefully selects the most popular 129 urban public streets with high numbers of reviews among all the urban streets in Shanghai to guarantee data and analysis quality (see Figure 1).

2.2. Data Collection and Analysis

2.2.1. The Data

This study draws upon data from one of China’s leading social media platforms, Dianping, which disseminates urban lifestyle and users’ experiences in cities. The platform not only provides urban sites for leisure consumption but also aggregates people’s reviews across these sites. Services shared by this platform are classified into several items, including gourmet, entertainment and shopping, outgoing trip, beauty and fitness, life convenience, and welfare. The diverse street areas are compiled in the outgoing trip item and are further classified into many types, in which urban streets with distinctive public life are picked up for this research while sites of natural landscape, zoos, and botanical gardens are excluded. There were rounds of manual selection to ensure that the chosen sites suited this study and contained enough user reviews. As Dianping is an open platform, its user groups are diverse, including local residents and short- and long-distance visitors. For urban street sites with public activities, people can share their experiences, views, and comments of the public space and built environment. These user-generated review texts associated with various “urban public streets” are valid raw data to analyze how people perceive these public spaces. Key advantages of this platform include a substantial user base, high update frequency, and the integration of geolocation data and user-generated ratings. To facilitate data collection, an Application Programming Interface (API) was employed to retrieve review texts published under the “Streets” category between 1 September 2023 and 1 September 2024 in Shanghai. The dataset comprises usernames, review dates, and the textual content of the reviews. Then, sentiment analysis was conducted using interfaces and pre-trained machine learning models. In particular a Chinese natural language processing (NLP) tool, provided by a third-party artificial intelligence service platform named Baidu AI Cloud (Baidu, Inc., Beijing, China), is applied. This generated a sentiment score for each review, automatically classified its sentiment polarity (i.e., positive, negative, or neutral), and assigned a corresponding confidence level. The sentiment scores ranged from 0 (indicating highly negative sentiment) to 1 (indicating highly positive sentiment). The use of deep learning algorithms to calculate sentiment values substantially reduced the need for manual interpretation and provided an efficient means of estimating visitors’ perceived emotional inclination, which is conceptualized in this context as their subjective level of preference [40].

Following the computation of sentiment scores for all textual data, reviews with confidence levels exceeding 90% were retained for analysis. This filters out over 63,000 valid user reviews across 223 “streets” in Shanghai, including 57,020 positive reviews, 5987 negative reviews, and 575 neutral reviews. These data were used to construct the Shanghai Urban Public Streets Sentiment Analysis Text Database. Then, to improve the usability and quality of the data, the research team excluded sites with less than 20 reviews and maintained 129 streets for further analysis. With the support of the database, the research calculates the average sentiment score of all user reviews for each selected street as the final sentiment rating representing the overall emotional valence of user experiences within that street.

2.2.2. Data Analysis

This research intentionally identifies six built environment elements upon review of related research and industrial reports for lexicon framework construction. As mentioned before, the selection avoids elements that are too general or broad to deliver specific meaning, e.g., resilience, sustainability, or smart cities, but focuses on concrete physical space entities that are shapeable and modifiable in progressive urban design or micro-regeneration. The identified elements include water bodies, roads, buildings, green spaces, squares, and recreational facilities; each is lexically constructed by package of associated words (see

Table 1. Classification of built environment elements for urban public streets.

Built Environment Elements	Lexicon	Associated Words
Water bodies	Natural or artificial water-filled areas, including rivers, lakes, waterfronts, canals, etc.	33
Roads	Paved or unpaved paths used for vehicular and pedestrian travel, including footpaths, alleys, lanes, bikeways, etc.	27
Buildings	Buildings or structures used for residential, commercial, industrial, artistic, or recreational purposes.	33
Green spaces	Areas that are covered with grass, trees, and other vegetation, including parks, gardens, lawns, etc.	26
Squares	Public open spaces, typically paved and surrounded by buildings, used for social or recreational activities.	16
Recreational facilities	Locations and structures designed for leisure activities, including pavilions, benches, seats, gazebos, kiosks, etc.	13

) and has a total of 148 words.

As Chinese language has no space in text writing, which is different from English-language text, word segmentation is needed before the texts can be analyzed. And similar to English, other tasks such as part-of-speech (POS) tagging, named entity recognition (NER), syntactic parsing, and anaphora resolution (AR) are also incorporated into the Chinese natural language processing (NLP) model. With the help of the Chinese NLP and well-trained machine learning tools, this research conducted sentiment analysis of the collated texts, as a whole and by type, with scores ranging from 0 (negative sentiment) to 1 (positive sentiment) under accurate calculation.

The review comments for each word associated with the built environment elements are sorted and analyzed, assigned with a sentiment score ranging from 0 to 1. As this research assumes that people’s sentiments on built environment elements matter for the overall perception, regression analysis is conducted to quantify the relationship between people’s overall sentiment scores and sentiment scores of water bodies, roads, buildings, green spaces, squares, and recreational facilities. The research applies SPSS (Statistical Product and Service Solutions, produced by IBM, Armonk, New York, NY, USA) to conduct the regression analysis. In particular, this research constructed a multiple linear regression model, where the dependent variable was the total sentiment score of each public street. The independent variables consisted of sentiment scores assigned to six types of built environment elements within each street. All sentiment scores were derived from textual data associated with each street. The model was specified as

$$Y={\beta }_0+{\beta }_1{X}_1+{\beta }_2{X}_2+{\beta }_3{X}_3+{\beta }_4{X}_4+{\beta }_5{X}_5+{\beta }_6{X}_6+\epsilon$$

(1)

where $Y$ is the total sentiment score, $X_1$ through $X_6$ represent sentiment scores for the six built environment element types, ${\beta }_0$ is the constant (intercept), ${\beta }_1$ to ${\beta }_6$ are regression coefficients for each predictor, and $\epsilon$ is the error term. Model results are interpreted upon the regression effects and coefficients. In general, the research flow include three steps that starts from identifying study area and data collection, then conducting sentiment analysis and followed regression statistics analysis (Figure 2).

3. Results

3.1. Perceptions Toward Different Categories of Urban Streets

Filtering districts with user reviews more than 20, this study extracted 129 valid samples from a total of 223 “urban public streets” in Shanghai. It first quantified sentiment inclination toward the street as a whole, followed with a mapping of the spatial heterogeneity of the sentiment scores and by categories. The result unveils that sentiment scores of all the urban public streets range from 0.53 to 1.0, and four levels of sentiment values were categorized to reflect public sentiment differences across five types of street (Figure 3 and Figure 4): the four levels are extremely positive (0.9 ≤ sentiment score < 1.0), highly positive (0.8 ≤ score < 0.9), moderately positive (0.7 ≤ score < 0.8), and slightly positive (0.5 ≤ score < 0.7).

Among all the 129 urban street sites, 49 fall into the “extremely positive” sentiment category (with sentiment scores ranging from 0.9 to 1), in which 17 distinguish themselves by culture, art, and historical elements (34.69%); 12 are labeled with gourmet food and good walkability (24.90%); 2 are ancient towns (4.08%); 9 are characterized by stylish features and a night market (18.37%); and 9 are featured with creative elements (18.37%). In user-generated review texts of this category, high-frequency descriptions include “sense of history”, “creativity”, “convenient transportation”, “ideal leisure destination”, etc., reflecting widespread, highly recognized experiences of the public in these streets. For instance, Wukang Road, a historical and cultural street (sentiment score: 0.937), is marked by the iconic Wukang Mansion and retains a 20th-century building style. The integration of art galleries and independent cafés fosters a unique atmosphere where “Old Shanghai charm” meets contemporary art. Comments such as “architectural aesthetics,” “café culture,” and “ideal for walk” frequently appear in the review texts, showcasing its attractiveness as a historical and cultural landmark. Similarly, Xintiandi (score: 0.937), one of the earliest property-led commercial real estate developments, is a street blending traditional Shikumen architecture with modern commercial elements. Its historical-meets-contemporary ambiance offers an immersive, mixed experience of a high-end, fashionable consumption place and grassroots cultural markets (Figure 5). Nightlife in Xintiandi is also active; bars and light arts facilitate the “night-time economy”. The relatively high consumption level does not prevent Xintiandi from being a popular urban street with a high emotional rating; instead, it turns out to be a landmark combining historical memory with metropolitan vitality. Daxue Road Xianshi Pedestrian Street (sentiment score: 0.936), adjacent to Fudan University, is another example illustrating how creativity and walkability play a role in a vibrant urban street. Capitalizing on college students, the street has evolved into a cluster of creative bookstores, designer boutiques, and diverse dining spaces, forming a youthful and international leisure ecosystem. Public reviews such as the “creative vibe” and “transportation accessibility” are emphasized, with over 70% of the positive comments referring its vibrancy and suitability for social activities among the youth.

There are 49 urban street sites falling within the “highly positive” sentiment category (0.8 ≤ sentiment score < 0.9, account 37.98% of all the samples), in which 11 distinguish themselves by culture, art, and historical elements (22.45%); 7 are labeled with gourmet food and good walkability (14.29%); 14 are ancient towns (28.57%); 8 are characterized by stylish features and a night market (16.33%); and 9 are featured with creative elements (18.37%). Tian’ai Road, a representative example in this category (see Figure 6), is labeled with a romantic theme and featured with symbolic landmarks such as a “love mailbox” and a graffiti wall, making it a popular photo spot for couples. While many praise the street’s romantic atmosphere and artistic elements, a small number of people also express their dissatisfaction with the dilapidated streets, narrow sidewalks, and lack of distinctive qualities. Such contrasting perceptions suggest that shortcomings in infrastructure and diverse experience might prevent these streets from reaching broader groups of people, limiting their attractiveness to the locals. In other words, these streets are yet to form an environment integrating history, commerce, leisure consumption, and modern fashion like those “extremely positive” counterparts. Another illustrative case is the Korean Street on Hongquan Road (sentiment score: 0.852), which is widely recognized as a focal point for Korean pop culture in Shanghai. Korean BBQ restaurants, beauty stores, and K-pop-themed cafés are densely concentrated in this area, attracting a substantial number of young consumers. However, “traffic congestion during peak hours” and “inadequate sanitation standards in restaurants” occasionally emerge in review comments, indicating that the development of supporting infrastructure might not cope with the increasing popularity of the street. Consequently, despite its thematic coherence and popularity among youth, sentiment score of the area remains “highly positive” and fails to go higher.

There are 20 urban street sites falling within the “moderately positive” sentiment category (0.7 ≤ sentiment score < 0.8, account 15.50% of all the samples), in which 2 distinguish themselves by culture, art, and historical elements (10.00%); 1 is labeled with gourmet food and good walkability (5.00%); 7 are ancient towns (35.00%); 6 are characterized by stylish features and a night market (30.00%); and 4 are featured with creative elements (20.00%). Tianzifang, a cultural and creative street regenerated from traditional lanes and alleys, gathers craft shops, cafes, and specialty food stalls and represents a cultural icon for traditional Shanghai. The street, however, is also criticized as being too “crowded” and with too much “homogenization” under excessive commercialization, resulting in a low sentiment perception score. This brings questions regarding the balance between cultural development and public experience. Cultural IP may improve popularity of the streets but can also lead to over-crowdedness and commercialization. Similarly, Qibao Night Market in Minhang District also received polarized review comments. As a popular ancient town in Shanghai, Qibao attracts people for its special local snacks and traditional town scenery, but as it is located in a suburb and the business market is small, it primarily serves local residents and neighboring visitors, limiting its serving groups. Moreover, “hygiene conditions” is also an important factor constraining Qibao from receiving a higher score. Another typical example is Huanghe Road (see Figure 7), which gained public attention due to the popularity of the TV series Blossoms (Fanhua) and attracted quite a few people to take photos. Taisheng Garden is one of the photo spots. However, the actual experience varies, as reflected in the review comments: some mention that the area lacks the bustling atmosphere shown in the series Blossoms, with few visitors and unremarkable food; other feel that “the old has gone,” noting that long-established restaurants have been converted into guesthouses, leaving behind a sense of faded prosperity. All these uncover the limitations and shortcomings of streets receiving “moderately positive” comments: historical and cultural elements or film and TV exposure may increase popularity of the streets but it is the intrinsic quality that supports long-lasting attractiveness.

There are only 11 urban street sites being categorized as “slightly positive” (0.5 ≤ sentiment score < 0.7, account 37.98% of all the samples), in which 1 is labeled with gourmet food and good walkability (9.10%); 3 are ancient towns (27.27%); 6 are characterized by stylish features and a night market (54.54%); and 1 is featured with creative elements (9.10%). In user-generated review texts of this category, high-frequency descriptions include “outdated facilities”, “mediocre experience”, “barely featured”, etc., indicating a low level of public satisfaction. Hong Kong Famous Stores Street (People’s Avenue branch), located in an underground commercial space, is criticized for its poor ventilation, depressing environment, and low-end stores, leading to the perception of a “poor experience”. Zhoupu Old Street, characterized by its dilapidated buildings and disorganized layout, lacks distinctive shops or cultural activities, with people labeling it an “uninspiring” street. The Jinjiang Amusement Park Night Market confronts challenges such as aging facilities and poor sanitation; review evaluations of “dirty, disorganized environment” and “lack of effective management” are universal. Zhuangxing Ancient Town, located in the outer suburb, suffers from inconvenient transportation, a monotonous business structure, and a barely developed commercial atmosphere, which is criticized for its “lack of vitality”. Likewise, Shanghai Sino-European Street (see Figure 8), with infrastructure shortcomings, chaotic sanitation management, and poor operational conditions, receives low satisfaction from the public. These urban streets are often located in a marginal area of the city with outdated infrastructure, poor management, and barely distinctive characteristics, and can hardly satisfy the public needs for high-quality leisure experiences.

3.2. Associations Between Perceptions on Urban Streets and Built Environment Elements

In general, people’s perceptions toward the six built environment elements are close but still exhibit difference, ranging from 0.846 to 0.892 (

Table 2. Classification and statistics of built environment elements.

Elements	Associated Words	Sentences (Containing the Words) Frequency	Words Frequency	Sentiment Score
Water bodies	33	54.87‰	14.83%	0.871
Roads	27	182.62‰	49.60%	0.846
Buildings	33	101.37‰	25.46%	0.877
Green spaces	26	31.72‰	7.72%	0.892
Squares	16	9.44‰	1.99%	0.864
Recreational facilities	13	1.94‰	0.40%	0.862

). Words associated with roads appear most frequently, reaching 49.6% of the total, and the sentences containing this element are also the highest, accounting for 182.62‰ in all the review sentences. But the sentiment score for roads was the lowest among all (only 0.846), implying that people concerned this element most but are generally least satisfied with it compared with other elements. The second frequently appearing element is buildings, with word frequency of 25.46%, and 101.37‰ of the sentences containing words associated with this element; sentiment score for buildings is 0.877, indicating that people care about buildings a lot in the urban street experience and feel quite positive about them. Words associated with water bodies rank third in word frequency (14.83%) and appear in 54.87‰ of all the sentences; sentiment score for water is also high (0.871), suggesting that people pay relatively high attention to water and are basically satisfied with this element. Green space is the fourth concerned element among the six, with 7.72% associated words frequency and appear in 31.72‰ of all the review sentences. Though words related to green spaces are not most frequently referred, their overall sentiment score is the highest, reaching 0.892, revealing that people value this element and are happily impressed in this regard during their street experience. The least two frequently appearing elements are squares and recreational facilities, each with word frequencies of merely 1.99% and 0.40%. Sentences containing associated words of squares and recreational facilities only account 9.44‰ and 1.94‰ of the total, respectively, and sentiment scores of the two are close to the average value of the six elements (0.864 and 0.862), showcasing that people’s attention to these two elements are limited and their perceptions towards the two are relatively neutral among all.

Multiple linear regression analysis is conducted to identify how people’s perceptions on built environment elements relate to their general perception toward the public street, which are summarized in

Table 3. Multiple regression model results.

Elements	B	SE	Beta	t-Value	p-Value	VIF
Water bodies	0.038	0.029	0.075	1.296	0.198	1.134
Roads	0.495	0.065	0.478	7.579	0.000	1.341
Buildings	0.223	0.036	0.368	6.110	0.000	1.222
Green spaces	0.048	0.020	0.148	2.373	0.019	1.317
Squares	0.000	0.013	−0.001	−0.020	0.984	1.129
Recreational facilities	−0.001	0.012	−0.007	−0.114	0.910	1.228
Constant	0.204	0.049		4.170	0.000
R2	0.638
Adjusted R2	0.621

. The highest variance inflation factor (VIF) value is 1.341, indicating that there is no significant multicollinearity in the model. The Durbin–Watson statistic is 2.247, indicating no significant autocorrelation in the residuals. The F statistic is 35.892 (p < 0.001), indicating that the overall model is statistically significant and explains a non-random portion of the variance in the dependent variable. In the regression model, the unstandardized coefficient (B) reflects the raw effect size, while the standardized coefficient (Beta) allows for comparison of the relative influence of each independent variable.

The sentiment for the built environment elements explains 62% (R² = 0.638, adjusted R² = 0.621) of the variation in the total sentiment scores of each street. All statistically significant elements, including roads, buildings, and green spaces, positively contributed to the total sentiment scores. Among them, roads exhibited the largest standardized coefficient (Beta = 0.478, p < 0.001), suggesting a strong positive influence on overall sentiment. This result indicates that the quality and design of road infrastructure have a significant impact on the emotional perception of urban streets. Buildings also had a significant positive relationship with overall sentiment (Beta = 0.368, p < 0.001). This suggests that perceptions of buildings, however aesthetically pleasing, well-maintained, or architecturally impressive, are important contributors to people’s emotional experience in urban public space environments. This finding highlights the role of architectural design in promoting urban emotional appeal. Green spaces demonstrated a smaller but still statistically significant positive effect on sentiment (Beta = 0.148, p = 0.019). While the influence of green spaces is less than that of roads or buildings, it still plays a meaningful role in shaping urban emotional experiences. The positive effect may reflect the well-documented benefits of nature and green in urban public streets, e.g., the improved mood or reduced stress levels. As water bodies, squares, and recreational facilities had small Beta values and p-values greater than 0.05, they had no meaningful impact on the total sentiment scores in this case. These findings unveil the synergistic role of infrastructure, architecture, and natural elements in fostering positive urban street experiences.

4. Discussion and Conclusions

Taking advantage of social media reviews from a leading platform, this research analyzes people’s perceptions toward urban public spaces and their built environment factors. Data sources of 129 urban street sites are collated to construct a perception lexicon. Five categories of street based on their features and six types of built environment element (water bodies, roads, buildings, green spaces, squares, and recreational facilities) are identified and their associations are analyzed. Accounting for approximately 36.74% of all the samples, urban streets with “extremely positive” sentiment mainly are located within Shanghai’s inner ring road such as the Huangpu and Xuhui districts and are “high-density” and “multi-function” in nature. Their key competitive advantages lie in the synergy of culture, creative art, historical heritage and walkability, as informed by the distribution of the sites in five street types. Compared with the “extremely positive” streets that primarily cluster in the central city (about 90.00%), streets with “highly positive” sentiment score are relatively dispersed, appearing in both the central city (61.21%) and peripheral and outskirts (38.79%) without clear cluster features. These streets are normally characterized by a distinct feature, e.g., a cultural theme or a specialized cluster of dining establishments, to attract niche consumer groups, but occasionally constrained by functional integration and environmental refinement. Urban streets with “moderately positive” sentiment mainly include local night markets (e.g., Qibao Night Market), traditional commercial streets (e.g., The City God Temple), and some regenerated old spaces (e.g., Hong Kong Famous Stores Street) and are primarily distributed within the suburban ring of Shanghai. Homogeneous competition and aging infrastructure are the main challenges faced by these streets, which can be reflected in negative comments such as “crowdedness”, “excessive commercialization”, and “hygiene problems”. Such unpleasant reviews partly reflect how the history, location, and transition matter in the production of a street’s atmosphere, and that street features alone may be insufficient to create an excellent perception from the general public or receive high scores. Streets with “slightly positive” sentiment score are spatially sporadic across the city. They mostly suffer from deficiencies in areas such as infrastructure, management, and uniqueness, resulting in relatively low emotional ratings.

In view of the association between built environment elements and perceptions of streets, this research identifies that sentiments on roads, buildings, and green spaces significantly influence overall sentiment. Roads including sidewalks, pavements, bikeways, pedestrian paths, etc., provide different groups of people with space to move (or stay). They are the foundational framework forming space of flows. Many studies have noted the close relationship between road space and pedestrians’ perception [41,42,43,44]. A study from Eskişehir road in Ankara, Turkey reveals that sidewalks along main traffic routes are often perceived negatively by pedestrians due to traffic noise, unsafety, and dust [41]. At macroscale, disconnectedness of roads is likely to decrease population density and vitality [42], whilst at meso- and microscales, a proper increase in density of intersections, bus stops, crossings, and in width of sidewalks can possibly improve pedestrians’ satisfaction [43]. In addition, designing shared spaces between traffic routes and streets for different groups of people is possible so as to improve safety and user experiences [44]. Buildings with the architecture and structure for residential, commercial, industrial, artistic, or entertainment purposes are the main functional spaces shaping the boundaries of the street areas. Studies on buildings and people’s perceptions mainly focus on building metrics such as density and design [45]. For instance, a study on residents’ attitudes toward suburban development in London unveils that vernacular design makes some increase in density tolerable, but significantly higher density might not be acceptable [45]. Green spaces covered with grass, vegetation, and trees or characterized by gardens and parks are soft landscape elements for street areas. A number of studies have recorded the positive impacts of greenery, e.g., lawns, trees, vegetation, or parks, on people’s perception toward the public space [46,47]. Noteworthily, the value of experiencing nature in human health is receiving increasing attention, whereby scholars and professionals argue for fine-grained ecosystem services design to improve urban public space quality [46].

In all, this research draws review comments from China’s leading social media platform to establish a perception lexicon database to provide an approach to analyze public perception of built environment [48]. The findings generate significant implications for urban public space design and management. First, locating in the central city represents a main advantage but exceptions occasionally happen. Urban streets located within the inner ring of the city benefit from a combination of historical resources, commercial vitality, and policy support, forming high-density clusters with high emotional ratings. This phenomenon reflects a positive correlation between “resource endowment”, “development intensity”, and “experience quality”. Some exceptional streets in suburbs receiving highly positive comments are almost developed upon original ancient towns, e.g., Jinze, Zhujiajiao, and Liantang in Qinpu or Shiyan in Jinshan. Second, there is an obvious polarization regarding themed urban streets. Streets located outside the central city solely rely on a particular IP or media exposure (e.g., thematic concepts or short-term online flows) but lack functional diversity or quality service and tend to find difficulty in maintaining long-term reputation. These streets might require an upgrade of business diversification and refined management to enhance their sustainability. Third, hardware remains significant for street experience, especially for those with low ratings. Built environmental factors such as “crowdedness”, “aging facility”, and “environmental hygiene problems” significantly impact people’s perceptions, and such impacts are much more prominent than those that are highly rated. For urban designers and managers, particular attention should also be given to roads, buildings, and green space before the development of streets’ distinctive characteristics. Then, repositioning the streets’ value in the urban leisure and public space system through, for example, business mode renewal would be an alternative to overcome the existing marginal status. However, this study acknowledges that the analysis has its own limitations. For instance, constrained by data availability and privacy, this study cannot distinguish the platform user groups, e.g., local residents or tourist visitors. This prevents us from a further understanding toward how different groups of people perceive the built environment and how their perceptions vary from one and the other. There is also a possible selection bias in reflecting opinions of active users, as people who visited the sites but did not leave a comment on the platform cannot be recorded and analyzed. Moreover, though Shanghai with its rich urbanity and developed online social network is an ample data source, it can hardly capture the diversity in cities: it may be difficult for small and medium-sized cities with monocentric urban cores to develop diverse types of streets; cities characterized by mountainous topography such as Chongqing may not follow the downtown–suburb distribution pattern and develop its own locally featured street areas. In other words, Shanghai as a renowned Chinese metropolis can offer learnable experiences for other cities, but such experience has its limitations in applying to a different urban setting and simple imitations cannot guarantee success. In addition, marketing strategies either from the platform or the nearby merchants may, to some extent, have an impact on people’s inclination when commenting. Though such impact might be less an issue for urban public space than for stores or shops, it remains nonnegligible when reviewing the results of social media sentiment scores.