Rural Image Perception and Spatial Optimization Pathways Based on Social Media Data: A Case Study of Baishe Village—A Traditional Village

Abstract

The sustainable development of traditional villages faces a core challenge stemming from the disconnect between public perception and spatial planning. To address this issue, this study, taking Baishe Village—a national-level traditional village—as a case study, constructs and applies a “Digital Humanities + Spatial Analysis” research paradigm that integrates text mining, sentiment analysis, visual coding, and spatial analysis based on multimodal social media data (Sina Weibo and Xiaohongshu) from 2013 to 2023. It aims to conduct an in-depth analysis of tourists’ rural image perception structure, emotional tendencies, and their spatial differentiation characteristics, and subsequently propose spatial optimization pathways that promote the revitalization of its cultural landscape and sustainable land use. The main findings reveal the following: (1) In terms of cognitive structure, the rural image presents a ‘settlement-dominated’ four-dimensional structure, with settlement elements such as pit kilns (accounting for more than 70%) as the absolute core. (2) In terms of emotional tendencies, a cognitive tension is formed between the high recognition of architectural heritage value (positive sentiment: 57.44%) and significant dissatisfaction with service facilities. (3) In terms of spatial patterns, a “dual-core-driven” pattern of perceived hotspots emerges, with 83% of tourist activities concentrated in the central–southern main road area, revealing a “revitalization gap” in village spatial utilization. The contribution of this study lies in translating abstract public perceptions into quantifiable spatial insights, thereby constructing and validating a “Digital Humanities + Spatial Analysis” paradigm that fuses multimodal data and links abstract perception with concrete space. This provides a crucial theoretical basis and practical guidance for the living conservation of cultural landscapes, the enhancement of land use efficiency, and refined spatial governance.

1. Introduction

Traditional villages, as spatial units embodying unique architectural heritage and historical–cultural value, are key carriers for preserving nostalgic memory and cultural DNA. In the wave of global rural transformation, marked by phenomena such as rural depopulation and the emergence of ‘shrinking villages,’ they play an increasingly vital role [1,2,3]. With the in-depth promotion of rural revitalization strategies, the conservation and revitalization of traditional villages have become a focal point for both academia and practice [4,5,6,7]. However, current conservation practices commonly face a core challenge—a significant disconnect between public perception and spatial planning [8,9,10]. On one hand, substantial investments are directed towards the restoration and reconstruction of physical spaces; on the other hand, the resulting spatial experiences often misalign with the expectations of tourists and the public, leading to a dilemma of “high investment, low recognition” [8,11,12]. This disconnection not only leads to the waste of resources but also reflects the neglect of human experience and real local needs. Therefore, how to scientifically interpret the public’s perception and preference of the rural environment and effectively integrate it into the spatial planning decision-making process has become a key bottleneck problem for traditional villages to achieve living conservation and sustainable development.

To address the aforementioned challenges, academia has explored various theoretical perspectives. Kevin Lynch’s image theory, originating from urban studies, provides a foundational framework for the structural analysis of spatial perception. Through his research on urban residents’ mental maps, Lynch pioneered the identification of five physical elements constituting the urban image—paths, edges, districts, nodes, and landmarks. The core contribution of this theory lies in providing a systematic vocabulary and framework for understanding and analyzing the “legibility” and perceptual structure of urban spaces, and it remains widely applied today [13,14,15,16,17]. Meanwhile, Tourism Destination Image (TDI) theory and Landscape Perception theory have enriched the understanding of tourist experiences from the perspectives of affective evaluation and human–environment interaction, respectively. TDI theory emphasizes that a destination’s image is jointly composed of a cognitive image (knowledge and beliefs about “what it is”) and an affective image (emotions and attitudes about “how it feels”). This theory positions “emotion” as a key variable, offering an important tool for quantifying and analyzing tourists’ subjective experiences [18,19,20]. Concurrently, Landscape Perception theory, from the perspective of human–environment interaction, highlights the decisive role of individual cultural background, experience, and values in shaping landscape preferences and imbuing a “Sense of Place” [21,22,23]. However, when directly applied to traditional village research, the limitations of these theories are increasingly evident. Traditional villages are not merely urban spaces or tourist attractions, they are “living” human settlement systems highly integrated and intertwined with production, life, and ecological functions, and their human experience is more diverse and complex. Their spatial boundaries are often blurred, path networks organic, and cultural connotations deeply embedded in daily life. Therefore, directly applying urban-centric image models struggles to capture their unique cultural landscape characteristics and intrinsic human–land relationships, and struggles even more so to fully reflect tourists’ deep experiences. This indicates an urgent need in traditional village image research to develop analytical frameworks that better reflect their composite and “living” characteristics, and to focus on the dynamic and multidimensional nature of experience.

With the rapid development of information technology, particularly the widespread adoption of social media, a paradigm shift has occurred in perception research [24]. Big data represented by User-Generated Content (UGC) on platforms such as Sina Weibo, Xiaohongshu, and Flickr, due to its massive sample size, strong timeliness, low acquisition cost, and ability to reflect “bottom-up” authentic voices, have become a valuable data source for gaining insights into public perception and behavior [24,25]. Currently, social media-based perception research has been widely applied in various fields. In urban studies, scholars use geotagged photos or textual data to identify urban functional zones, evaluate the quality of built environments, and measure urban vitality and sentiment [24,26]. For example, Viriya Taecharungroj et al. used Flickr photos to perceive the image of 222 cities globally, proposing a new way to interpret urban images [26]. Chen et al. captured urban park visitors’ perception of cultural ecosystem services and landscape factors by mining social media data [21]. Huang et al. compared “Big Data” and “Small Data” methods in studying the image of the city on social media [16]. Hao Chen et al. studied users’ landscape preferences using TikTok short video content [27]. These studies commonly employ techniques such as text mining (e.g., word frequency analysis and Latent Dirichlet Allocation (LDA) topic modeling), sentiment analysis, and spatial kernel density analysis, which have greatly enhanced the objectivity and granularity of perception research.

Despite these theories laying a crucial foundation, existing research still presents three major limitations that constrain their effective application in the planning practice of traditional villages, especially in fully capturing human experience. First, theoretical adaptability is insufficient; directly applying urban image models to rural contexts struggles to fully capture the unique, composite landscape characteristics woven from production, life, and ecology in villages [28]. Second, there is a methodological reliance on traditional approaches; existing research predominantly relies on small-sample methods such as questionnaires [29,30] and interviews [31,32], which have limited timeliness, sample representativeness, and capacity to reveal spatial details. Third, the dimensionality of data is often singular; even studies that have begun to use social media data are mostly confined to single-dimensional analysis of either text or images [33,34,35], failing to fully leverage the potential of multimodal data in revealing the linkage effects among cognition, emotion, and space. To more clearly illustrate the progress and shortcomings of existing research,

Table 1. Comparison of key studies on spatial perception and optimization based on social media data.

Research (Author, Year)	Data Sources	Main Method	Concerns/Contributions	Limitations (Relevant to This Study)
Carlos Garcia-Palomares et al. (2015) [15]	Flickr photos	GIS spatial analysis	Identify tourist hotspots in European metropolises	Image data only, focusing on urban areas, without delving into emotional/cognitive structures.
Huang et al. (2021) [16]	Flickr photos; questionnaire	Comparison between big data and small data	Urban image research; methodological comparison	Focusing on urban areas, the data integration is not high and has not been applied to rural spatial optimization.
Chen et al. (2024) [21]	Social media data	Text mining; sentiment analysis	Perception of cultural ecosystem services in urban parks	Focusing on urban parks, without delving into the spatial correlation of multimodal data.
Liu and Guo (2023) [22]	Social media data	Text mining; sentiment analysis	Perception of ice and snow tourism imagery	Focusing on macro imagery, lacking refined spatial analysis.
Munawir et al. (2019) [23]	Google Maps reviews	Text mining	Theme park visitor perception and brand strategy	Focus on text, underutilizing image data for spatial perception.
This study	Weibo, Xiaohongshu (multimodal)	Text mining; sentiment analysis; visual encoding; spatial analysis	Construct a “digital humanities + spatial analysis” paradigm to connect perception and space, proposing optimization pathways	Data source representativeness (tourist perspective); spatial analysis accuracy; cross-sectional study.

summarizes key studies on spatial perception and optimization based on social media data, along with their main characteristics and limitations. In summary, there is now an urgent need for an innovative research paradigm capable of integrating multimodal data, systematically analyzing public perception, and precisely linking these perceptions to geographical space in order to more comprehensively and deeply understand and enhance the human experience in traditional villages.

To fill the aforementioned research gap, this study, taking the national-level traditional village of Baishe as a case study, aims to construct and apply an integrated research framework that fuses multimodal social media data with multi-technique analysis. Specifically, this study will address the following core research questions:

• Based on social media data, what are the core constituent elements of tourists’ perceived rural image of Baishe Village, and what is its internal structure?
• What is the overall emotional tendency of tourists towards the rural image of Baishe Village, and what specific image elements or experiences are the main sources of positive/negative emotions?
• What geographical distribution patterns do tourists’ spatial activities and visual focal points in Baishe Village exhibit, and which are the perceived hotspot spaces?
• What is the relationship between the formation of these spatial hotspots and their intrinsic constituent image elements?

To answer these questions, the main objectives of this study are (1) to systematically analyze and structure the elements of tourists’ perceived rural image; (2) to quantitatively evaluate tourists’ emotional tendencies and diagnose experiential shortcomings; (3) to accurately locate and visualize the spatial patterns of perceived hotspots; and (4) based on the above analysis, to propose data-driven spatial optimization pathways. The primary contribution of this study is not merely to depict the image of a specific village, but to provide a replicable “diagnostic tool” for the conservation and development of traditional villages. It achieves this by translating abstract public perceptions into quantifiable spatial indicators, constructing and validating a “Digital Humanities + Spatial Analysis” paradigm that fuses multimodal data and links abstract perception with concrete space. This provides a crucial theoretical basis and practical guidance for the living conservation of cultural landscapes, the enhancement of land use efficiency, and the promotion of more responsive, refined spatial governance, with a particular emphasis on people-centered experience enhancement.

2. Materials and Methods

2.1. Study Area

This study selects Baishe Village, a national-level traditional village located in the Guanzhong region of Shaanxi Province, as the case study area (Figure 1). The village was included in the second batch of the “Directory of Chinese Traditional Villages” in 2013 and was designated as a “National Forest Village” in 2019, forming a unique “cultural–natural” dual-heritage landscape. As the core protection zone for the “sunken Yaodong” (underground cave dwelling) architectural complex in Northern China, Baishe Village is renowned as a “living museum of raw earth architecture.” Its unique “surface-hidden” spatial form, characterized by the saying “trees are seen but no village, the village is entered but no houses are seen,” along with its rich cultural relics, provides an ideal case for exploring the rural image perception and spatial characteristics of this special type of traditional village [36,37,38].

2.2. Data Sources and Pre-Processing

To comprehensively capture public perception of Baishe Village, this study selected two mainstream Chinese social media platforms, Sina Weibo and Xiaohongshu, as sources for multimodal data. The former is noted for the immediacy and breadth of its textual information, while the latter is characterized by high-quality visual content and in-depth experiential sharing. Combining both allows for a more three-dimensional reflection of the rural image. Although this study was unable to obtain precise Global Positioning System (GPS) coordinates for each data point, we achieved spatial positioning by visually identifying photo content and referencing explicit landmarks mentioned in the text (such as ‘Earth Pit Kiln No. 1′ and ‘Niangniang Temple’). This provided a reliable foundation for hotspot pattern analysis.

Data collection was conducted for the period from December 2013 to December 2023. Using keywords such as “Baishe Village,” a total of 2378 texts and 986 photos were initially retrieved through a targeted web crawler based on the Scrapy framework in a Python 3.10 environment. To ensure data quality and analytical validity, a rigorous data cleaning and filtering process was designed (

Table 2. Data cleaning and filtering process.

Data Type	Filtering Stage	Specific Criteria
Text Data	Automated Cleaning	Removal of advertisements, duplicate content, and invalid text (e.g., pure emojis).
	Manual Semantic Check	Exclusion of texts irrelevant to the theme of Baishe Village.
Image Data	Initial Content Screening	Exclusion of images where the geographical scene is unrecognizable or dominated by selfies (human face > 70% of frame).
	Manual Scene Check	Ensuring image content pertains to landscapes, architecture, or activities within Baishe Village.

). After pre-processing, a multimodal database containing 978 valid texts and 263 valid landscape photos was finally constructed, providing a sample basis with both historical depth and representativeness for subsequent analysis.

2.3. Research Framework and Methods

The methodological framework of this study integrates the classic theory of the city image by Kevin Lynch, the affective and cognitive analysis paradigms from Tourism Destination Image (TDI) research [39], and recent cutting-edge practices in landscape perception and spatial analysis using social media big data [40,41,42,43]. Through the organic integration of these methods, this study innovatively constructs a new process for the multi-dimensional perceptual analysis and spatial diagnosis of traditional villages.

The analytical framework consists of four interconnected steps (Figure 2). First, we employed text mining techniques to identify the core image elements of Baishe Village and their structural relationships. Next, sentiment analysis was used to quantify tourist emotional tendencies and explore their driving factors. Then, we coded the visual data to construct a thematic model of the rural image elements and to identify spatial hotspots. Finally, in the key integration step of this study, we conducted a spatialized correlation analysis of the above findings to reveal perceptual differences across various spatial zones and subsequently propose targeted spatial optimization pathways. The specific methods are detailed below.

2.3.1. Image Element and Structure Analysis

To identify the core elements of tourist perception, this study processed the 978 valid texts. First, the Jieba library was used for Chinese word segmentation and stop-word removal. Then, KH Coder 3.0 software was utilized for frequency analysis (TF-IDF algorithm) to extract the most representative high-frequency keywords. To further explore the intrinsic connections between these elements, a keyword co-occurrence network was constructed. In this network, nodes represent keywords, and the weight of the edges represents the frequency of their co-occurrence within the same text [44]. To reveal the network’s topological features and semantic structure from multiple dimensions, we employed several algorithms, including Community Detection, Correlation Analysis, and Centrality Analysis. The visualization and specific analysis of these networks will be presented in Section 3.

2.3.2. Sentiment Analysis

To quantify the emotional experience of tourists, this study used the SnowNLP library in Python to assign a sentiment score to each text i. SnowNLP is a model trained on a sentiment lexicon and a Naive Bayes classifier, capable of calculating a sentiment score S(i) that ranges from 0 (completely negative) to 1 (completely positive) [45,46]. The sentiment category C(i) of text i was determined by setting thresholds for its sentiment score S(i), defined as follows:

$$C(i)=\left\{\begin{array}{c}Positive,if S\left(i\right) > 0.6\\ Neutral,if 0.4\le S\left(i\right) \le 0.6\\ Negative,if S\left(i\right) < 0.4\end{array}\right.$$

where S(i) represents the sentiment score of text i. The proportional distribution of sentiments across the entire sample set can be calculated from all N texts. For example, the percentage of positive texts, Ppositive, is calculated as follows:

$$P_{positive}={\displaystyle \frac{\sum _{i=1}^N[C\left(i\right)=\mathrm{Positive}]}{N}}\times 100\%$$

where [C(i) = Positive] is an indicator function that equals 1 if text i is classified as positive, and 0 otherwise. To further investigate the specific reasons driving different emotions, we conducted separate keyword extraction and frequency analysis on the “positive” and “negative” text subsets to identify the core image elements closely associated with each sentiment tendency.

2.3.3. Visual Content and Spatial Hotspot Analysis

To gain an in-depth understanding of tourists’ visual focal points, this study employed the qualitative analysis software NVivo 20.0 to analyze the 263 photos. The process followed the three-stage coding procedure of Grounded Theory [47,48,49,50]: (1) Open Coding: Each specific landscape element in the photos (e.g., architecture, plants, decorations) was tagged to form initial concepts. (2) Axial Coding: Related initial concepts were categorized and merged to form higher-level categories (e.g., “architectural forms,” “courtyard plants”). (3) Selective Coding: All categories were interconnected and refined to finally construct a visual thematic model of the rural image that reflects the overall visual perception structure. Simultaneously, during the coding process, the identifiable shooting location of each photo was recorded. By calculating the frequency of each location, the most concentrated perceived spatial hotspots for tourist photography were identified.

2.3.4. Image–Space Correlation Analysis

This stage represents the integration and deepening of the research. We created a “Hotspot-Landscape Element” correlation matrix. The rows of this matrix represent the main spatial hotspots identified in Section 2.3.3, and the columns represent the high-frequency core visual elements identified through coding. By analyzing which high-frequency visual elements constitute each hotspot, we were able to reveal the underlying reasons for the attractiveness of different spaces, thereby achieving an effective linkage between abstract images and concrete geographical spaces.

2.4. Methodological Validation

To ensure the rigor and validity of our integrated research framework, we employed a multi-pronged validation approach. First, we performed a cross-validation between our different data modalities. The core image elements identified through text mining (e.g., “Yaodong,” “Dikengyuan”) showed a high degree of consistency with the most frequently photographed landscape elements identified through visual content analysis. This consistency between what visitors write about and what they photograph validates the robustness of our findings on the core perceptual elements. Second, the methods adopted in this study have been widely proven to be effective in previous research. The use of social media data for sentiment analysis and hotspot identification is a well-established methodology in urban studies and tourism research, confirming its applicability to our case [51,52]. Finally, our findings were validated against real-world evidence. The identified spatial hotspots (e.g., Earth Pit Kiln No. 1 and Touristic Earth Pit Kiln A) align closely with the recommended tour routes provided by the Baishe Village tourist center and official promotional materials, confirming that our data-driven results accurately reflect on-the-ground visitor behavior patterns.

3. Results

3.1. Composition and Structure of the Rural Image

This section aims to answer the first research question: What are the core constituent elements of tourists’ perceived rural image of Baishe Village, and what is its internal structure?

3.1.1. Identification of Core Image Elements

The frequency analysis of 978 valid texts (Figure 3) clearly reveals the core of tourist perception. The terms “Yaodong” (cave dwelling, 296 times), “Dikengyuan” (pit-courtyard, 253 times), and “Diyao” (earth-kiln, 194 times) form the highest-frequency triad of image elements, with their total frequency far exceeding all other keywords. This finding directly confirms that settlement architecture, represented by the Yaodong, is the absolute dominant element constituting the rural image of Baishe Village. It is noteworthy that the mention frequency of “Dikengyuan” even surpasses that of the village name “Baishe Village” (152 times), further highlighting the iconic status of this unique architectural form in tourist cognition.

Beyond architectural elements, tourist perception also exhibits multidimensional characteristics (Figure 4). Geographical and Spatial Traits: Keywords such as “Sanyuan” (128 times) and “underground” (64 times), along with unique descriptions like “seeing trees but not the village” and “hearing sounds but not seeing people,” collectively construct a clear geographical positioning of the village and a profound impression of its “surface-hidden” human settlement environment. Cultural and Historical Value: The appearance of words like “revolution,” “raw earth architecture museum,” and “intact” reflects tourists’ awareness of the village’s red culture attributes and heritage conservation value. Natural and Ecological Elements: Terms like “tree” and “cypress tree” constitute the basic units of perception regarding the village’s natural environment.

3.1.2. Semantic Structure of Image Elements

To explore the intrinsic connections among the core image elements, a keyword co-occurrence network was constructed (Figure 5). The network consists of 83 nodes and 81 edges, exhibiting a clear community structure. A multi-dimensional network analysis approach was adopted to systematically reveal its topological features and semantic structure.

First, the community structure of the network reveals the multidimensional themes of the image. As shown in Figure 5, the network was clearly divided into four main semantic clusters using the Modularity algorithm. Cluster 1: Dikengyuan Architectural Core (red nodes). This community, centered on words like “Dikengyuan,” “Yaodong,” “Siheyuan” (courtyard house), and “sunken,” is the largest and most densely connected area in the entire network, reflecting the absolute center of gravity of tourist perception. Cluster 2: Geographical and Spatial Traits (blue nodes). This cluster links geographical identifiers like “Baishe Village” and “Sanyuan” with spatial experience descriptions such as “hearing sounds but not seeing people.” Cluster 3: Historical and Cultural Value (green nodes). This cluster includes words like “museum,” “revolution,” “history,” and “conservation,” embodying tourists’ cognition of the village’s cultural value and heritage attributes. Cluster 4: Natural and Ecological Environment (yellow nodes). This cluster is composed of natural elements like “tree,” “loess,” and “ground.”

Second, centrality analysis identifies the key elements for image integration (Figure A1). The calculation of Betweenness Centrality shows that nodes such as “Diyao” (centrality value: 2438), “courtyard” (1728), and “Siheyuan” (1746) have the highest scores. These nodes act as “structural holes” connecting different semantic clusters, playing a crucial bridging role in integrating the cognition of architectural forms, spatial perception, and cultural value.

Finally, correlation analysis reveals a potential cognitive conflict (Figure A2). The analysis indicates that most words describing architecture and landscape are positively correlated. However, it is noteworthy that the keywords “revolution” and “vernacular residence” show a statistically significant negative correlation (r = −0.17, p < 0.05). This finding suggests that in tourists’ perception, there may be a certain value tension or cognitive separation between the village’s “red revolutionary heritage” attribute and its “traditional residential function” attribute. This phenomenon particularly merits attention in future revitalization efforts.

In summary, the rural image of Baishe Village presents a complex semantic network with the Yaodong architecture as its absolute core, radiating outwards to multiple dimensions including geographical space, history and culture, and the natural environment, and is characterized by a complex structure with internal cognitive tensions.

3.2. Sentiment Analysis of the Rural Image

3.2.1. Overall Distribution of Tourist Sentiment

To quantify the emotional experiences expressed by tourists on social media, this study performed sentiment analysis on the 978 valid texts. As shown in Figure 6, the results indicate that the overall impression of Baishe Village is predominantly positive. Among all texts, positive sentiment accounted for the highest proportion at 57.44%; neutral sentiment followed at 28.93%, while negative sentiment had a relatively low share at 13.63%. This asymmetrical distribution of sentiment preliminarily reveals that the core attractions of Baishe Village as a tourist destination have been generally recognized by tourists.

3.2.2. Analysis of Driving Factors for Positive and Negative Sentiments

To further investigate the specific reasons driving different emotions, this study conducted separate keyword extraction and analysis on the “positive” and “negative” text subsets (

Table 3. Comparison of core keywords for positive and negative emotions (selected vocabulary: frequency > 10).

Emotion Type	Percentage (%)	Evaluation Adjective (Quantity)
Positive Emotion	57.44%	Ancient (36), warm in winter and cool in summer (29), intact (25), simple and unadorned (25), complete (22), mysterious (19), natural (15), unique (15), tranquil (13), lush (10).
Negative Emotion	13.63%	General (15), inconvenient (13), far (13), not tasty (12), abandoned (12), insufficient (12), deep (11), poor (10), backward (10), tired (10).

). The results accurately pinpoint the strengths and weaknesses in tourist perception.

The core driving force for positive sentiment stems from a high degree of recognition for the value of architectural heritage and the unique environmental experience. In positive texts, keywords such as “ancient” (36 times), “warm in winter and cool in summer” (29), “intact” (25), “simple and unadorned” (25), and “mysterious” (19) appeared most frequently. These words are highly focused on the historical depth, ecological wisdom, conservation status, and unique spatial feeling of the Yaodong architecture, indicating that the core positive experience for tourists comes from a deep appreciation of this cultural heritage.

The main source of negative sentiment is concentrated on the inadequacy of tourism service facilities and a sense of decay in some areas. In negative texts, “inconvenient” (13 times), “not tasty” (12), “insufficient” (12), and “abandoned” (12) became high-frequency words. Through further interpretation of the relevant texts, these negative comments mainly point to issues with transportation accessibility, dining quality and options, and the dilapidated state of the undeveloped northern part of the village.

In conclusion, the emotional experience of tourists presents a significant “cognitive tension”, whereby they come for “what to see” (the core cultural heritage) and give it high praise, but at the same time, they express considerable dissatisfaction with the process of “how to get there, what to eat, and how to get around” (the tourism supporting services). This finding points to a clear direction for the future conservation-oriented development and service quality improvement of Baishe Village.

3.3. Hotspot Patterns and Element Composition of the Image Space

3.3.1. Identification and Distribution of Perceived Spatial Hotspots

To investigate the geographical distribution characteristics of tourist spatial behavior, this study conducted a frequency analysis of identifiable shooting locations from the 263 photos to identify the most focused-on perceived spatial hotspots. As shown in Figure 7, tourists’ visual attention is highly concentrated. “Touristic Earth Pit Kiln A” (photographed 53 times) became the core photo-taking spot by a clear margin, followed by generic references to “Yaodong” (cave dwelling, 45 times), “Dikengyuan entrance” (38 times), “Earth Pit Kiln No. 1 (Red Revolution Site)” (16 times), and the “Niangniang Temple area.”

By mapping these hotspots onto the village base map (Figure 8), a clear spatial pattern emerges. The tourist activities exhibit a significant “dual-core-driven” pattern, with over 83% of photographic activities highly concentrated in the southern experiential zone centered on “Earth Pit Kiln No. 1” and the central cultural exhibition zone centered on “Touristic Earth Pit Kiln A.” There is a clear “main road dependency” in spatial use, as all hotspots are located adjacent to the central–southern main road, forming a linear cluster. The northern area has become a “perception blind spot” and a “revitalization gap”; despite having numerous Yaodong, this area is almost absent from tourists’ perceptions, visually revealing a severe imbalance in the village’s spatial revitalization and use. The typical landscapes of hotspot and coldspot areas are shown in Figure 9.

3.3.2. Landscape Attraction Composition of Hotspot Areas

To answer the key question of “why these places become hotspots,” we conducted an in-depth analysis of the visual element coding from the photos of major hotspots to deconstruct their “landscape attraction formula” (

Table 4. Analysis of landscape element composition in key hotspot areas.

Hotspot Space	Core Landscape Elements (from NVivo Coding)	Attraction Type Classification
Earth Pit Kiln No. 1	Entrance stone steps (high frequency), courtyard trees (high frequency), cave dwelling interior (medium frequency), decorations (lanterns/corn) (medium frequency)	Comprehensive experiential type
Touristic Earth Pit Kiln A	Red cultural symbols (high frequency), architectural forms (medium frequency)	Culture-empowered
Around the Niangniang Temple	Iconic temple architecture (high frequency), public square space (medium frequency)	Public node type

). The analysis found that different hotspots have distinct sources of attraction.

Earth Pit Kiln No. 1 (Culture-Empowered Hotspot, Figure 9a): The formation of this hotspot, in addition to its basic architectural form, benefits greatly from the powerful empowerment of its “Red Revolution Site” cultural label. In photos and texts about this location, the mention rate of elements like “history” and “revolutionary stories” is much higher than in other areas, indicating that cultural narratives significantly enhance the space’s attractiveness.

Touristic Earth Pit Kiln A (Comprehensive Experiential Hotspot, Figure 9b): Its high popularity stems from the richness of landscape elements and the completeness of the scene. It successfully integrates multiple high-frequency visual elements that tourists focus on—such as a “well-preserved entrance stone staircase,” a “shady tree in the center of the courtyard,” “lifelike furnishings inside the cave dwelling,” and “atmosphere-creating decorations like hanging red lanterns and corn cobs”—into a single space, providing tourists with a “one-stop,” most-representative Yaodong experience.

Niangniang Temple Area (Public Node Hotspot, Figure 9c): The popularity of this area is mainly driven by the landmark status of the “Niangniang Temple” building itself, as well as the public space attributes of the open area factors (in front of it), which is suitable for gathering and hosting events.

In conclusion, whether a space can become a beloved hotspot for tourists depends not only on possessing a single high-quality landscape element but, more critically, on its ability to provide a rich combination of landscape elements, unique cultural value empowerment, or important public space functions. This finding offers valuable practical insights into how to activate “coldspot” areas within the village through landscape design and functional implantation.

4. Discussion

4.1. Construction and Interpretation of the Rural Image Perception Model for Baishe Village

Based on the comprehensive analysis of multimodal data, this study ultimately constructed a thematic model of the rural image of Baishe Village that systematically reflects tourist perceptions (

Table 5. Thematic model of the overall image of Baishe Village.

Image Theme	Image Cluster	Core Image Elements (Frequency/Value)
Settlement Imagery	Road System	Country path (2)
	Architectural Types	Touristic Dikengyuan (16), Village entrance (4), Village committee (1), Dikengyuan entrance (38), Dikengyuan (53), Dikengyuan homestay (3), Above-ground brick building (8), Farm-stay Dikengyuan (8), Water well (3), Yaodong (45)
	Textures and Decorations	Lantern (16), Woven handicraft (2), Haystack (6), Door couplet (3), Wooden pavilion (10), Wall carving (2), Entrance stone steps (24), Stone table/stool/pier (12), Outdoor screen (2), Eaves decoration (2), Chimney (3), Woven corn cob string (15)
	Village Layout	Above-ground Dikengyuan (16), Overall layout (14), Village landscape appearance (23)
Life Imagery	Religious Beliefs	Niangniang Temple (1), Chastity Arch (2)
	Culinary Culture	Delicacies/Food (2)
	Folk Art	Paper-cutting (3)
Natural Environment	Climatic Conditions	Sky (176)
	Pastoral Landscape	Natural village scenery (7)
	Flora and Fauna	Cypress forest (14), Old locust tree (3), Courtyard center tree (32)
Production Imagery	Production Activities	Villager group activities (4), Tourist visits (8), Outing/Excursion (12)
	Production Tools	Stone mill (116)

). This model is not a simple enumeration of image elements but rather reveals a complex perceptual system with inherent structures and hierarchies.

The model exhibits a prominent feature of being absolutely dominated by “settlement imagery.” As shown in Table 4, among all four image themes (settlement, nature, life, and production), elements related to settlement imagery account for over 70% of the weight, whether measured by the frequency of textual mentions (e.g., “Yaodong,” “Dikengyuan”) or the intensity of visual attention (e.g., all hotspots being buildings or architectural clusters). This once again confirms that the unique built environment, represented by the Yaodong, is the cornerstone of Baishe Village’s core attraction and local identity.

Simultaneously, the model reveals an imbalance in the image composition. Compared to the rich settlement imagery, the “life imagery” reflecting the daily lives of local residents (e.g., folklore, culinary culture) and the “production imagery” (e.g., agricultural activities, production tools) are relatively weak in tourist perceptions. This highlights the powerful attraction of the architectural heritage on the one hand; however, on the other, it implies deficiencies in the “living” presentation and experience of the village’s cultural connotation, which corroborates the shortcomings in service facilities identified in Section 3.2.

Therefore, this image theme model is not only a systematic summary of the research findings but, more importantly, serves as a “diagnostic atlas.” It clearly points out the strategic directions for Baishe Village’s future development: “consolidating strengths” (reinforcing the advantages of the settlement landscape) and “addressing weaknesses” (enriching the life and production experiences).

4.2. Implications for Land Use and Spatial Planning in Traditional Villages

The findings of this study, particularly the “cognitive tension” in tourist perceptions and the “revitalization gap” in the spatial pattern, have profound implications for the sustainable land use and refined spatial planning of traditional villages, and are closely linked to the enduring theme of “development” versus “preservation.”

First, this study provides a new “experiential effectiveness” dimension for land use benefit assessment, echoing the research by Elhosiny et al. [51], who similarly emphasized the importance of visitor experience quality in urban public space management. However, unlike their reliance on questionnaires for satisfaction scores, this study offers a more spontaneous and dynamic assessment method through sentiment analysis of User-Generated Content (UGC). For instance, Yaodong (pit-courtyard dwelling) sites serving a cultural preservation function received extremely high positive evaluations, whereas some lands serving a tourism development function (e.g., dining spots) became a major source of negative emotions. This data-driven “cognitive tension” clearly reveals the conflict between development and preservation at the perceptual level. It warns planners that when undertaking commercial development, they must prioritize the experiential quality of the core cultural landscape to avoid diminishing the destination’s overall attractiveness due to subpar supporting facilities.

Second, the “perceived space” analysis paradigm proposed in this study offers a precise basis for spatial resource allocation. By identifying the “dual-core-driven” spatial pattern and “perception blind spots,” this study clearly projects the collective behavior of tourists onto the geographical space. This provides insights into how many villages facing “shrinking” risks due to out-migration can precisely invest revitalization resources [52]. This indicates that future spatial resource investment should not be indiscriminately spread thinly but should adopt a differentiated strategy, i.e., for hotspot areas with high tourist concentration, the focus should be on enhancing spatial carrying capacity, optimizing the visitor experience, and improving micro-circulation. For the vast “coldspot” areas, a “point-based activation” approach (e.g., implanting new small-scale cultural nodes or special experience projects) should be used to precisely bridge the “revitalization gap,” guide visitor flow, and thus enhance the overall efficiency of land use and the organic connectivity of the village space.

4.3. Contribution to the Living Conservation of Cultural Landscapes

This study provides a “bottom-up,” public perception-driven new method for the value identification and living conservation of cultural landscapes, serving as a powerful supplement to the traditional expert-led model.

Conventional landscape conservation often relies on the integrity assessment of physical forms. In contrast, this study, through the analysis of texts and images, identifies the core cultural landscape DNA that truly constitutes the “Sense of Place” of Baishe Village in public cognition. This DNA includes not only macro-architectural forms like the “Dikengyuan” but also a series of specific, micro-level yet high-frequency landscape elements such as the “tree in the center of the courtyard,” the “entrance stone steps,” and the “furnishings inside the cave dwelling.” This inspires us that the conservation of cultural landscapes should not only focus on repairing the “shell” but must also emphasize the protection and representation of these key scenes and elements that can evoke tourists’ emotional resonance and constitute collective memory.

Furthermore, the revealed weakness in the perception of “life imagery” and “production imagery” also sounds a warning for the “living” aspect of cultural landscape conservation. A village devoid of the breath of life has an incomplete cultural landscape. Therefore, future conservation pathways should transcend mere architectural restoration and explore how to organically integrate intangible cultural heritage (such as folk activities and traditional crafts) with physical spaces. This can be achieved by recreating rural life scenes, thereby transforming the village from a “static architectural museum” to a “living cultural landscape experience destination”.

4.4. Integrated Discussion and Research Implications

In summary, this study effectively addresses the research gaps identified in the Introduction—the prevalent lack of systematic diagnostic tools capable of connecting public perception with spatial planning in current traditional village research—by constructing a multimodal data analysis framework. Our research paradigm, by quantifying and spatializing abstract image perceptions, provides a viable pathway to tackle the practical challenge of “disconnect between planning and perception.”

However, this study also has its shortcomings. While the “cognitive tension” and “revitalization gap” we identified reveal existing problems, they do not offer empirically tested solutions. The spatial optimization pathways proposed in this study are more-strategic recommendations based on diagnostic results, and the applicability of their results needs further verification and adjustment in future planning practices. For instance, our proposed “point-based activation” strategy might be effective in villages with a single spatial structure like Baishe Village, but its applicability in more complex villages still requires further exploration.

Despite these limitations, the core contribution of this study lies in its methodological demonstration effect. It shows that by utilizing readily available social media data, managers and planners can “listen” to public voices at low cost and high efficiency, thereby making more responsive decisions.

5. Conclusions and Prospects

5.1. Main Conclusions

Through a systematic analysis of multimodal social media data, this study has profoundly revealed the rural image perception characteristics and spatial patterns of Baishe Village, which is a traditional village. The main conclusions are as follows: (1) In terms of cognitive structure, tourists’ rural image is dominated by the built environment, presenting a “settlement-dominated” structure with Yaodong (pit-courtyard dwellings) as its core. (2) In terms of emotional tendencies, a significant “cognitive tension” exists in tourist emotions, where high recognition of architectural heritage value coexists with considerable dissatisfaction with service facilities. (3) In terms of spatial patterns, tourist activities exhibit a “dual-core-driven” clustering pattern, exposing a severe “revitalization gap.”

These findings not only provide specific diagnostic bases for the “development” and “preservation” of Baishe Village but also reveal a deep connection between physical space and human narratives. The linear distribution of tourist activity hotspots along the main road is not merely a trajectory of spatial movement; it further implies that tourists are subconsciously constructing a linear “story of place” about this location. This perspective aligns with theories by scholars such as Tuan [53] on how places are perceived through sequential experiences. The implication for planning practice is that to deepen the human experience, planners should act as “spatial storytellers,” designing an engaging narrative experience path that transforms a simple visit into a memorable cultural journey.

5.2. Research Innovations and Contributions

The theoretical contribution of this study lies in constructing and validating a “Digital Humanities + Spatial Analysis” paradigm that integrates multimodal data and connects abstract perception with concrete space, providing a new methodological tool for rural landscape perception research. Its practical contribution is to offer a replicable, data-driven “diagnostic tool” for the managers and planners of traditional villages, which is capable of translating public perceptions into tangible spatial insights, thereby providing decision support for the living conservation of cultural landscapes, the enhancement of land use efficiency, and refined spatial governance.

5.3. Limitations and Future Prospects

Despite offering valuable insights, this study also faces several limitations that need to be addressed. First, as pointed out by the reviewers, social media data cannot fully represent all tourist groups and, crucially, lack the essential perspective of local residents. Second, the spatial analysis precision of this study was limited by landmark identification rather than exact geographical coordinates. Finally, this cross-sectional study failed to capture the dynamic evolution of perceptions.

Looking ahead, these limitations also pave the way for future research. Future studies can integrate more cutting-edge Artificial Intelligence (AI) tools (such as image generation models) for visualizing scenario simulations or may combine them with traditional survey questionnaires for cross-validation. Simultaneously, conducting cross-case comparative studies and longitudinal tracking studies on the effects of planning interventions will be crucial paths to deepen understanding in this field. Through these continuous efforts, we hope to better bridge the gap between public perception and sustainable development practices.