Analysis of Tourists’ Cultural Perception in Cultural Tourism Villages Based on Online Review Data: A Case Study of Dangjia Village, Shaanxi Province, China

Abstract

Rural tourism is a significant driver of socio-economic development in rural areas. However, current offerings are often characterized by monotonous experiences, homogenized products, and a lack of cultural depth, failing to meet tourists’ growing demand for immersive engagement. While some scholars have adopted a spatial perception perspective, a visitor-centered approach remains scarce, with limited focus on methods for analyzing cultural perception. This study takes Dangjia Village, a renowned cultural tourism destination in Hancheng, China, as a case study. By scraping online reviews from travel and social platforms, we employ LDA topic modeling, textual semantic analysis, and IPA to investigate the characteristics and preferences of tourists’ cultural perception. The findings reveal that: (1) Tourists’ cultural perception of Dangjia Village includes three dimensions: History and Culture, Architecture and Culture, and Local Products and Culture. (2) Positive sentiments outweigh negative ones in tourist evaluations. (3) The History and Culture dimension received the highest levels of both attention and satisfaction. Architecture and Culture attracted the least attention but relatively high satisfaction, while Local Products and Culture garnered considerable attention yet the lowest satisfaction. Originating from a visitor perception perspective, this study explores cultural perception characteristics, providing insights for the high-quality utilization and optimization of cultural tourism resources in such villages.

1. Introduction

In recent years, the demand for rural tourism in China has grown steadily, revealing strong development potential and significant economic benefits for local communities [1]. Data from China’s Ministry of Culture and Tourism (2024) shows that both sectors continue to flourish [2]. Current forms of rural tourism mainly include natural wellness, arts-based experiences, and cultural tourism. Of these, cultural tourism villages are especially popular due to their unique cultural content and immersive experiences [3]. Integrating distinctive rural cultural resources with tourism enriches its content and diversity, thereby enhancing overall appeal and competitiveness. In the long term, the deep integration of culture and tourism can not only upgrade the rural tourism industry but also promote the preservation and innovation of traditional culture [4].

However, several significant problems have emerged during the development of rural cultural tourism:

Some villages have failed to fully explore their intrinsic cultural significance. Their tourism offerings remain at an initial, rudimentary stage characterized by basic agritourism activities like “eating farmhouse meals and staying in farmhouse lodgings”. The failure to thoroughly explore and showcase the rich cultural elements of local traditions and folk customs has resulted in a limited visitor experience [5]. Rural cultural tourism products lacking uniqueness and enduring appeal struggle to meet tourists’ growing demand for diverse and meaningful experiences.

Currently, many tourism products present culture in limited and superficial ways, hindering profound visitor experiences. This issue contributes to low revisit rates and threatens the long-term sustainability of tourism-oriented villages. For instance, while some villages preserve their vernacular architecture well, they often lack multi-layered interpretation systems and participatory activities. Consequently, visits are frequently reduced to the passive visual experience of “viewing houses”, with missed opportunities for deeper engagement with the associated historical narratives, local anecdotes, traditional customs, and folk crafts.

Moreover, the tourism offerings in some villages often fail to reflect their distinctive characteristics. This lack of uniqueness undermines their market competitiveness and makes it difficult to attract and retain visitors [6].

In summary, the current development of rural cultural tourism faces three main challenges: inadequate cultural excavation, simplistic resource transformation, and an ambiguous thematic focus. These issues result in superficial visitor experiences and limited cultural fulfillment, ultimately hindering the sustainable development of cultural tourism villages.

Adopting a visitor perception lens, this study investigates the dimensions and sentiment preferences of tourists’ cultural perception. The findings aim to provide insights for formulating precise spatial and experiential optimization strategies in cultural tourism villages.

2. Literature Review

This study reviews pertinent studies based on three key concepts: spatial perception, tourist perception, and textual semantic analysis.

2.1. Spatial Perception

Research on spatial perception examines how individuals comprehend and perceive their built environment, including its features, structures, and relationships. Common methodologies include behavioral experiments, questionnaires, cognitive mapping, and neuroimaging, which collectively help elucidate the underlying cognitive processes and psychological mechanisms. The field’s foundational framework was established by Kevin Lynch in “The Image of the City” (1960) [7], which proposed that urban spatial perception is structured by five key elements: paths, edges, districts, nodes, and landmarks. This theoretical framework has since been widely adopted and expanded upon by subsequent researchers.

Early spatial perception research primarily addressed urban, landscape, and architectural scales. The field has since broadened considerably, with recent studies becoming more empirical and methodological innovative. Current work frequently focuses on macro-scale urban contexts or meso-scale environments like streets [8] and communities, examining the fundamental elements that shape spatial perception.

Research perspectives typically address three key dimensions: the characteristics of perceiving subjects (e.g., demographics, cultural background), the attributes of perceived spaces (e.g., typology, configuration), and the effects of perception (e.g., on behavior, well-being, and place attachment) [9,10,11].

Research methodologies vary with scale. While space syntax is often employed for macro-scale urban analysis, meso- and micro-scale studies frequently utilize methods such as the PSPL (Public Space & Public Life) survey [12], online textual analysis [13,14], and semantic segmentation [15,16].

2.2. Tourist Perception

Tourist perception research, a key subfield of spatial perception, examines the experiences and subjective impressions formed from an individual’s interaction with a destination’s environment. This research significantly advances the understanding of visitor preferences, cultural resource development, and experience enhancement [17,18]. However, the multi-layered nature of these perceptions makes identifying and evaluating their specific dimensions a central yet persistent challenge in both theory and practice.

The widespread adoption of social media has established User-Generated Content (UGC) as a major open-source data resource. Unlike traditional researcher-driven data, UGC offers efficiency, dynamism, and authenticity, making it crucial for visitor perception research [19]. Analyzing the psychological perceptions and value orientations embedded in UGC allows planners to understand the tourist perspective, thereby enabling more targeted and human-centered optimization of spatial design and cultural integration.

2.3. Textual Semantic Analysis

Text semantic analysis in tourism research primarily addresses five areas: word sense understanding, sentence comprehension, sentiment analysis, topic extraction, and intent recognition. Among these, topic extraction and sentiment analysis are most prevalent in tourist review analysis [20]. In topic extraction, both traditional techniques like LDA [21], NMF [22], and the more recent BERTopic method [23] can identify latent themes from text to enable the dimensional identification of visitor perception. These methods represent distinct methodological paradigms. LDA and NMF are probabilistic or matrix factorization models that operate on word frequency statistics, offering high transparency and interpretability through explicit topic-word distributions. In contrast, BERTopic leverages pre-trained language models to generate dense document embeddings, subsequently applying clustering algorithms to discover topics. This approach excels at capturing nuanced semantic relationships but can be more computationally intensive and less transparent in its topic formation process.

Notwithstanding the advanced capabilities of neural approaches like BERTopic, this study employs LDA for its proven effectiveness, computational efficiency, and superior interpretability. The probabilistic topic-word distributions generated by LDA provide a clear and intuitive basis for identifying and labeling cultural perception dimensions, which aligns perfectly with our research objective of constructing a transparent and explainable analytical framework.

Sentiment analysis, conversely, reveals tourists’ affective evaluations by determining emotional polarity. This approach is particularly valuable given a key methodological limitation of UGC: the potential bias in numerical ratings, as some platforms may filter low scores or users may assign artificially high ratings while providing critical written feedback. By deriving sentiment directly from textual content, semantic analysis effectively mitigates this bias.

2.4. Summary

Current visitor perception research has made significant progress, supported by diverse and rigorous methodologies that provide a robust theoretical foundation for studying various spatial typologies. The use of large-scale visitor reviews further ensures data richness and reliability. Together, these advances enable a more scientific, visitor-centric analysis, effectively identifying a destination’s strengths, growth areas, and aspects needing improvement.

However, existing research on visitor perception in rural tourism contexts still exhibits the following notable shortcomings:

• Overly Generalized Themes: Studies often address overly generalized themes of visitor perception, with a notable scarcity of specialized research focusing on specific, well-defined topics.
• Lack of Integrated Analysis: There is insufficient integration between the analysis of perceived themes (what visitors focus on) and sentiment orientation (how they feel about it). This gap hinders the ability to pinpoint underlying issues and their root causes with precision.

3. Materials and Methods

3.1. Research Object

Dangjia Village is located northeast of the urban area of Hancheng City, Shaanxi Province, China, situated on the northern bank of the Bishui River Valley in a geomantically auspicious location. The ancient architectural complex of the village, a quintessential example of Guanzhong vernacular dwellings, is characterized by its well-preserved courtyard houses, intricate stone carvings, and historic alleyways. Included in the Tentative List of China’s World Cultural Heritage Sites in 2008 and designated as a National AAAA-level Tourist Attraction in 2016, Dangjia Village has established itself as a model cultural tourism destination. Here, traditional customs and cultural activities remain a vital part of the visitor experience (Figure 1).

Dangjia Village was selected as the research object for the following reasons:

• Representative Development Challenges: The current tourism development in Dangjia Village remains predominantly reliant on static exhibition. This approach faces significant bottlenecks, including limited experiential dimensions and insufficient visitor participation levels. These prevalent issues exhibit strong representativeness as challenges commonly encountered within cultural tourism villages.
• Robust Data Availability: Dangjia Village attracts a substantial volume of visitors, resulting in a relatively ample sample size of social media review data. This data possesses sufficient validity and analytical utility for robust research investigation.

3.2. Data Collection and Processing

In studies utilizing tourist reviews as data samples, crawling user-generated content from prominent, specialized tourism review platforms is a common methodological approach. Dianping (Dianping review data for Dangjia Village: https://www.dianping.com/shop/H3nMJ89hxTHWu4tI#comment (accessed on 26 May 2024)), Qunar (Qunar review data for Dangjia Village: https://travel.qunar.com/p-oi701758-dangjiacun (accessed on 26 May 2024)), Ctrip (Ctrip review data for Dangjia Village: https://you.ctrip.com/sight/hancheng120420/54588.html (accessed on 26 May 2024)), and Mafengwo (Mafengwo review data for Dangjia Village: https://www.mafengwo.cn/poi/6329062.html (accessed on 26 May 2024)), as China’s leading professional travel review platforms, are frequently utilized by scholars for collecting tourist review data. Accordingly, Therefore, this study obtained user data from the aforementioned four Chinese websites to form a user review dataset. These platforms are characterized by their substantial data volume and high credibility, fulfilling the data requirements of this study. Using Octoparse, a web scraping tool, the collection and storage of review data were implemented. This process yielded a total of 2863 review records pertaining to Dangjia Village.

Online review data is characterized by its sheer volume and unstructured nature. To ensure research quality and topic relevance, preliminary screening and cleaning of the data are essential. This study employed manual screening to remove invalid entries from the raw comment data. This included scenic area descriptions (official/promotional text), advertisements, irrelevant text and emojis/emoticons. Furthermore, the following data normalization steps were applied: Traditional Chinese characters were converted to Simplified Chinese, English text was translated into Chinese. Following this cleaning and normalization process, a preliminarily cleansed dataset of 2509 review entries was obtained, comprising approximately 174,000 Chinese characters. This refined dataset provides a suitable foundation for subsequent text mining and semantic analysis.

3.3. Research Methodology

This study analyzes visitor perception through text data from mainstream tourism review websites. After data cleaning, the Chinese text was processed using Jieba (the Jieba library is an open-source Chinese text segmentation tool. Our research employed this tool for text processing, and the source code can be accessed at: https://gitee.com/fxsjy/jieba (accessed on 26 May 2024)) library for word segmentation. We then employed word frequency statistics to identify cultural perception hotspots, revealing overall visitor focus. Subsequently, an LDA topic model (The implementation of the LDA topic modeling code was adapted from a tutorial on the CSDN blog platform. The complete resource is available at: https://blog.csdn.net/qq_59771180/article/details/138850229 (accessed on 26 May 2024)) extracted the underlying dimensions and frequency of cultural perceptions. Sentiment orientation was analyzed using SnowNLP (the SnowNLP library is available at: https://gitee.com/mirrors/snownlp (accessed on 26 May 2024)). Finally, Importance–Performance Analysis (IPA) integrated these findings—perception characteristics, cultural themes, and sentiment—to evaluate visitor attention and satisfaction levels for each cultural theme (Figure 2).

3.3.1. Word Frequency Statistics

Analyzing high-frequency words in visitor reviews reveals the most salient impressions of a destination. After segmenting the UGC data with Python 2.7’s Jieba library, we performed preliminary cleaning by removing adverbs, quantifiers, and meaningless characters. A custom stop-word list was then applied to filter out context-irrelevant terms (e.g., “ticket”, “transportation”), resulting in a refined corpus of cultural perception keywords.

3.3.2. LDA Topic Model

Clustering structurally similar perceptual elements offers clearer insights into tourists’ cultural perceptions. This study employs the LDA model to analyze unstructured UGC data, objectively identifying the structure and hierarchy of visitor perception dimensions in rural spaces. This data-driven approach mitigates the subjective bias inherent in manual analysis [24]. Specifically, we applied the LDA model to segment pre-processed word groups into thematic clusters. After determining the optimal number of topics by evaluating coherence and perplexity, we labeled each topic and analyzed its high-frequency words. This process ultimately yielded the specific dimensions and characteristic elements of tourists’ cultural perception.

To elucidate the working mechanism of the Latent Dirichlet Allocation (LDA) model, it is essential to clarify that its input consists of a bag-of-words representation derived from pre-processed and segmented text, rather than raw sentences. This model operates on a probabilistic generative framework, positing that each document (e.g., a tourist review such as “Dangjia Village has a long history, its architecture is full of historical sense, and the alleys and courtyards are very interesting”) is a mixture of a small number of latent topics, each of which is a probability distribution over words. By computationally inferring the two underlying distributions—document-topic and topic-word—LDA probabilistically clusters frequently co-occurring words (such as “history”, “architecture”, “alleys”, and “courtyards” extracted from the example) into the same latent topic. Subsequently, we, as researchers, interpret the sets of highest-probability words within each topic and assign them an interpretative label, such as “Architecture and Culture”, thereby identifying the emergent dimensions of tourists’ cultural perception.

3.3.3. Sentiment Orientation Analysis

With the rapid advancement of network technology, people increasingly use social platforms to express opinions on scenic environments, travel experiences, and other perceptions, generating vast volumes of emotionally charged textual review data. Extracting sentiment orientation from UGC data enables more accurate assessment of visitor evaluations, avoiding misjudgments caused by “high-score yet negative reviews” (strategic rating inflation). Current mainstream methods for text sentiment analysis include convolutional neural networks, long short-term memory models, bidirectional LSTM architectures, SnowNLP models, Baidu’s ERNIE text representation models, and ROST CM6.0 software [25]. Given that this study’s textual data primarily consists of unstructured Chinese short texts, the SnowNLP library on Python was employed to conduct precise sentiment orientation analysis. Sentiment scores were calculated on a 0–1 scale and categorized into three classes: positive (>0.6), negative (<0.4), and neutral (0.4–0.6).

3.3.4. IPA Method

The Importance–Performance Analysis (IPA) method evaluates items based on their importance and performance, helping to identify priorities for improvement. This study operationalizes this framework by using perception frequency as the indicator for attention (importance) and sentiment orientation as the indicator for satisfaction (performance). By plotting these two indicators, targets are classified into four distinct quadrants: high attention-low satisfaction, high attention-high satisfaction, low attention-high satisfaction, and low attention-low satisfaction. This provides a clear, visual characterization of tourists’ cultural perception evaluations [26].

4. Results and Analysis

4.1. Overall Characteristics of Visitor Perception

High-frequency words directly reflect aspects of a destination that leave the strongest impressions on visitors post-travel. Using ROST CM6.0, word frequency analysis was performed on Python-segmented UGC data, yielding high-frequency feature words for Dangjia Village (

Table 1. Frequency statistics of high-frequency feature words.

Serial №	Feature Word	Frequency	Serial №	Feature Word	Frequency
1	Hancheng	735	16	Village stockade	133
2	History	542	17	Ancient	125
3	Vernacular dwelling	532	18	Wenxing Pavilion	118
4	Architecture	497	19	Antiquity	117
5	Preservation	393	20	Village	110
6	Culture	310	21	Living fossil	109
7	Quadrangle courtyard	280	22	Pepper	109
8	Shaanxi	227	23	Oriental	101
9	Ancient architecture	187	24	China	97
10	Settlement	175	25	Atmosphere	97
11	Tradition	166	26	Unsophisticated	93
12	Ming–Qing dynasty	158	27	Xi’an	91
13	House	151	28	Protection	91
14	Treasure	150	29	Folk customs	90
15	Ancestral hall	144	30	Biyang Fort	84

). The high-frequency word list reveals that terms like “Hancheng”, “History”, “Vernacular dwellings”, and “Architecture” dominate, highlighting Dangjia Village’s core identity in visitor perception as a settlement centered on ancient architectural and courtyard culture. Notably, specific architectural names such as “quadrangle courtyards”, “Wenxing Pavilion”, and “Biyang Fort” appear prominently, indicating these spaces create lasting memories. Furthermore, words like “Well-preserved”, “Culture”, “Treasure”, “Ancient”, “Living fossil”, and “Oriental” collectively project a strongly positive image of Dangjia Village in tourists’ minds.

4.2. Characteristics of Tourists’ Cultural Perception

Tourists’ cultural perception dimensions are derived from the probabilistic clustering of co-occurring perceptual elements within the textual data, representing a thematic abstraction of shared attributes. This study utilized the LdaModel from the gensim library in Python to implement the LDA topic model. To determine the optimal number of topics, we computed both topic coherence and perplexity.

Topic coherence (C_v) quantitatively assesses the semantic interpretability of a topic by measuring the similarity between its high-probability words, with higher values indicating more human-intelligible themes [21]. The coherence score is calculated as:

$$C_v={\displaystyle \frac{1}{N}}\sum _{i=2}^N\sum _{j=1}^{i-1}{\displaystyle \frac{cosine(\overrightarrow{v}\left(w_i\right),\overrightarrow{v}\left(w_j\right))}{\mathrm{l}\mathrm{o}\mathrm{g}\left(\right(D\left(w_i,w_j\right)+1\left)\right)}}$$

where $N$ is the number of top words in a topic, $\overrightarrow{v}\left(w\right)$ is the vector representation of word $w$, and $D\left(w_i,w_j\right)$ is their document co-occurrence count.

Perplexity, in contrast, measures the model’s predictive uncertainty on unseen data. It is defined as the exponential of the negative log-likelihood per word:

$$Perplexity\left(D_{test}\right)=exp\left(-{\displaystyle \frac{{\sum }_{d=1}^{M}logp\left({\mathrm{w}}_d\right)}{\sum _{d=1}^M{N}_d}}\right)$$

where $M$ is the number of test documents, ${\mathrm{w}}_d$ represents the word sequence in document $d$, $N_d$ is the number of words in document $d$, and $p\left({\mathrm{w}}_d\right)$ is the probability of document ${\mathrm{w}}_d$ under the model. Lower perplexity values indicate better generalization performance.

In our analysis, the model was trained with 100 iterations. Owing to preliminary data cleaning that filtered out context-irrelevant terms (e.g., transportation, pricing), the derived topics showed high conceptual concentration. The perplexity curve exhibited a monotonically increasing trend with the number of topics, thus providing limited utility for model selection. Consequently, the topic coherence curve served as the primary criterion, establishing 3 as the optimal number of topics (Figure 3).

4.2.1. Cultural Perception Dimensions

After determining the optimal topic number as 3, the LDA model was used to cluster the data with the iteration count set to 100. Based on the semantic features of the top 20 high-frequency words under each divided topic, thematic word refinement was conducted, with results shown in Figure 4 and Figure 5. Topic 1 is “Local Products and Culture”, including keywords such as “pepper”, “yogurt”, and “snacks”, describing impressions of Dangjia Village’s food and local products. Topic 2 is “Architecture and Culture”, including keywords such as “quadrangle courtyard”, “vernacular dwelling”, and “ancient architecture”. Topic 3 is “History and Culture”, including keywords such as “culture”, “history”, and “ancient”, reflecting visitors’ overall perception of Dangjia Village’s historical culture.

4.2.2. Cultural Perception Frequency

As shown in

Table 2. Tourist perception dimensions and frequency in Dangjia Village.

Topic	Quantity	Frequency	High-Frequency Words
History and Culture	1163	46.4%	History, Preservation, Culture, Ming and Qing Dynasties, Villages, Ancient villages, Tradition, Ancient times
Architecture and Culture	548	21.8%	Quadrangle courtyard, Vernacular dwelling, Ancestral halls, Wenxing Pavilion, Alleys, Watchtowers, Biyang Fort
Local Products and Culture	798	31.8%	Pepper, Houses, Yogurt, Snacks, Guanzhong, Rustic, Ancient city, Eating

, tourists’ perception intensity across different dimensions of Dangjia Village varies significantly, ranked in descending order: History and Culture (46.4%) > Local Products and Culture (31.8%) > Architecture and Culture (21.8%). Visitors demonstrate high perception of the village’s overall historical-cultural ambiance, though such perceptions are predominantly impressionistic and vague—not focused on specific historical events or figures but emerging from spatial ambiance and layout. This is corroborated by high-frequency words like “history”, “architecture”, “vernacular dwelling”, and “culture” under this theme (Table 2).

Regarding architectural culture perception, spaces like “quadrangle courtyards”, “Watchtower”, “Wenxing Pavilion”, and “Biyang Fort” are frequently mentioned. This occurs because: (1) these structures are iconic within Dangjia Village, and (2) they historically housed specific functions or activities that attract visitor attention. Concurrently, under the Local Products and Culture theme, high-frequency terms such as “pepper”, “snacks”, and “yogurt” reflect visitors’ preliminary impressions of local products.

4.3. Sentiment Orientation of Cultural Perception

Sentiment orientation in visitors’ travel reviews directly reflects their satisfaction and overall impression of the space. Analysis of cleaned complete comment data using SnowNLP reveals that 73.09% of visitor sentiment is positive while 19.01% is negative, demonstrating high overall recognition and satisfaction with Dangjia Village.

4.4. Visitor Cultural Perception Evaluation

Using the IPA method, visitor cultural perception data was evaluated through a dual-quadrant matrix of attention-satisfaction, with perception frequency serving as the attention-level indicator and sentiment orientation as the satisfaction-level indicator (

Table 3. Analysis of tourist attention and satisfaction ratings across different perception dimensions.

Topic	Attention	Positive		Neutral		Negative		Satisfaction
		Quantity	Frequency	Quantity	Frequency	Quantity	Frequency
History and Culture	46.4%	932	80.14%	72	6.19%	159	13.67%	81.3%
Architecture and Culture	21.8%	415	75.73%	38	6.93%	95	17.34%	78.2%
Local Products and Culture	31.8%	487	61.03%	88	11.03%	223	27.94%	65.2%

). Attention-level data derived from the percentage distribution of each perception dimension after LDA topic modeling; satisfaction-level data originated from sentiment scores obtained by SnowNLP-based scoring of review data.

By performing standardized processing on the evaluation results of attention and satisfaction across different dimensions, a visualized scatter distribution (Figure 6) intuitively reflects the priority levels of evaluation outcomes for Dangjia Village’s perception dimensions. Concurrently, combining high-frequency word lists and specific review texts under each theme identifies areas requiring improvement in the village’s cultural tourism development.

The History and Culture dimension exhibits both the highest visitor attention level and satisfaction level, constituting a critical factor in Dangjia Village’s tourism appeal. Representative reviews, such as “This ancient village, boasting nearly 700 years of history, is an absolute paradise for history enthusiasts”, underscore how the exceptional preservation of its vernacular architectural complex immerses visitors in profound historical and cultural resonance, thereby attracting substantial interest from culturally engaged tourists.

The Architecture and Culture dimension exhibits the lowest visitor attention level yet maintains relatively high satisfaction. Integrating high-frequency words from both History-Culture and Architecture-Culture dimensions reveals that Dangjia Village’s historical and architectural culture primarily manifests through physical spaces like ancient dwellings, courtyards, public spaces, and landscapes. However, the absence of systematic interpretation prevents visitors from grasping specific cultural narratives. While tourists recognize the village’s historical significance, they remain unaware of its concrete cultural connotations. As evidenced by representative comments: “Guided tours are mandatory at scenic spots; visitors cannot navigate them independently”. This superficial understanding fails to foster strong place attachment or cultural identification, significantly diminishing revisit and recommendation intentions.

The Local Products and Culture dimension demonstrates relatively high visitor attention but notably low satisfaction. High-frequency terms like “pepper”, “yogurt”, and “snacks” under this theme reflect visitors’ concentrated impressions of local products. Within this dimension, negative reviews predominantly cite food safety concerns and substandard dining environments. Deteriorating spatial infrastructure directly compromises the experiential quality of culinary cultural offerings. Concurrently, the limited diversity of specialty products further contributes to negative evaluations.

5. Discussion

5.1. Findings

This study analyzed visitor review data to identify the characteristics of cultural perceptions among tourists in Dangjia Village and their overall emotional tendencies. The findings reveal that tourists’ cultural perceptions of Dangjia Village encompass three dimensions: architecture and culture, history and culture, and local products and culture. Visitors demonstrated high overall recognition and satisfaction with Dangjia Village. Among these dimensions, historical and cultural aspects garnered the highest visitor attention and satisfaction. Architectural and cultural elements received the least attention but still achieved relatively high satisfaction. Local specialty products and cultural elements attracted considerable attention but yielded the lowest satisfaction levels. By categorizing perceptions through these dimensions and integrating emotional sentiment analysis, this approach enables precise identification of issues in converting Dangjia Village’s cultural resources into tourism assets, thereby proposing corresponding optimization strategies.

5.1.1. Sources of Visitor Dissatisfaction

By analyzing negative reviews, the current spatial conditions of the village and flaws in cultural resource utilization approaches can be precisely identified. Cross-referencing specific comments reveals three primary causes of visitor dissatisfaction:

1.

Inadequate Cultural Experience Activities

Spatial homogeneity and static displays—devoid of participatory activities—result in weak visitor engagement. The absence of interactive experiences diminishes visitors’ sense of agency and immersion.

2.

Monolithic Cultural Interpretation

Overreliance on textual/pictorial displays fails to convey cultural narratives effectively. Visitors receive fragmented information, reducing cultural ambiance to vague, conceptual impressions rather than tangible understanding.

3.

Deficient Spatial Planning

Outdated infrastructure (e.g., inadequate parking, restroom facilities, and poorly maintained pathways) fails to accommodate visitor needs, directly degrading tourism experiences.

5.1.2. Optimization Strategies

To address the aforementioned issues, optimization suggestions for Dangjia Village’s cultural tourism development are proposed from three perspectives:

• Clarify the three core cultural tourism themes of Dangjia Village and optimize the utilization of related cultural tourism resources. The History and Culture dimension exhibits the highest visitor attention and satisfaction levels, constituting a key factor in Dangjia Village’s appeal. As historical-cultural heritage represents a developmental strength, prioritize holistic preservation of the village’s authentic character and historical scene reenactments, comprehensively documenting and showcasing its rich historical-cultural narratives through multi-perspective displays. The Architecture and Culture dimension shows the lowest attention but relatively high satisfaction; further tap its potential to enhance tourism appeal by increasing the quantity and diversity of spatial interpretation and guided tours, disseminating knowledge of traditional architectural culture to deepen visitors’ cultural understanding. The Local Products and Culture dimension attracts high attention but the lowest satisfaction; prioritize improvements such as developing specialty product brands to elevate visitor experiences.
• Enrich cultural interpretation methods and enhance interaction between the destination and visitors. For Dangjia Village’s unique architectural cultural resources, expand interpretive formats and interactive elements to facilitate intuitive cultural learning—for example, using 3D projections to recreate historical spatial functions, and employing video presentations, role-playing, and interactive games to intensify cultural immersion during visits.
• Plan cultural tourism routes and upgrade supporting infrastructure. Deeply explore the village’s cultural uniqueness by auditing cultural resources and spaces, focusing on core cultural themes including patriarchal ritual systems, agrarian-literati traditions, merchant culture, folk customs, and historic building craftsmanship. Develop a distinctive “Guanzhong Vernacular Dwelling Living Museum” IP to avoid homogeneous agritourism. Implement scenario-based micro-renovations of core spaces—transforming ancestral halls and steles into AR-assisted ancestral instruction experiences and immersive ritual theaters—while creating cohesive narrative itineraries to enhance engagement. Concurrently, upgrade infrastructure such as parking facilities, restrooms, and barrier-free access to ensure foundational visitation quality.

5.2. Comparison with Other Studies

In previous research on visitor perception, some scholars employed methods similar to this study, analyzing online reviews to investigate tourists’ perceptions of destinations. Xu Xiwei (2024) [19] utilized machine learning to examine visitor perception in Xi’an’s Beiyuanmen Historical-Cultural District, categorizing perceptions into six dimensions: consumption experiences, local cuisine, landscape architecture, environmental ambiance, artisanal crafts, and intangible cultural heritage. Yuan Jiahui (2024) [27] integrated web semantics and questionnaires to collect negative evaluations of Suzhou classical gardens, constructing a framework for assessing adverse feedback. However, most existing studies focus on holistic destination perception, with limited specialized research on singular perceptual themes.

This study further clarifies its contribution by engaging with recent methodological advances. Compared to the macroscopic decision-making framework of Hao Yingyi (2025), which integrates LDA with FCE and QFD for high-level planning [28], our study provides a complementary, micro-level analysis by deeply examining the specific structure of cultural perception. Similarly, while Da Yeon Kim (2025)’s attribute extraction method identifies a broad set of service features [29], our focused approach reveals the intrinsic dimensions of cultural experience in heritage tourism. Crucially, our integration of IPA with LDA moves beyond identification to offer a diagnostic tool, pinpointing the critical gaps between tourist attention and satisfaction for strategic prioritization. Thus, this study contributes by offering a nuanced, thematic deep-dive into cultural perception and a framework that directly translates findings into actionable insights for cultural tourism management.

Regarding prior studies on Dangjia Village, Li Chang (2018) [30] explored tourism development issues from a cultural landscape conservation perspective, concluding—consistent with this study—that certain cultural resources suffer from superficial utilization, lack deep experiential engagement, and generate ambiguous cultural perception. Concurrently, Zhao Yang (2024) [31] adopted a methodology akin to this research, analyzing UGC data to map Dangjia Village’s tourism image perception, identifying problematic prioritization of sightseeing over cultural experiences.

In summary, the inadequate transformation and utilization of cultural resources, which diminishes the depth and richness of visitor perception, constitutes the foremost challenge in Dangjia Village’s cultural tourism development. This study employs machine learning (LDA topic modeling) [21], textual semantic analysis [20], and IPA methodology [26] to concentrate specifically on cultural perception themes. By delineating cultural perception characteristics and preferences, it diagnoses deficiencies in Dangjia Village’s tourism development and examines spatial planning issues, establishing critical connections between visitor cultural perception evaluations and tangible destination management challenges.

5.3. Limitations and Future Research

This study explores cultural perception in cultural tourism villages through topic mining and sentiment orientation analysis of visitor reviews from Dangjia Village Via big data scraping. However, limitations persist:

• Review texts are influenced by visitors’ age, gender, education level, and other factors; comment data alone cannot fully reconstruct authentic cultural perception experiences.
• During data cleaning, manual removal of culturally irrelevant terms introduces subjectivity, potentially compromising objectivity and authenticity.
• Focusing solely on Dangjia Village, the data reflects perceptions of a single settlement, limiting conclusion generalizability and strategy applicability.

Future research should integrate visitor movement trajectories with survey interviews to comprehensively explore hierarchical cultural perception evaluations. Additionally, comparative studies of diverse cultural tourism villages could further investigate space-emotion-behavior interaction mechanisms.

6. Conclusions

This study selects a representative cultural tourism village in Guanzhong region as the research object. From a visitor perception perspective, textual semantic analysis was applied to UGC data for latent topic mining and sentiment analysis. The findings reveal:

• Tourists’ cultural perception of Dangjia Village comprises three dimensions: History and Culture, Architecture and Culture, and Local Products and Culture, with History and Culture exhibiting the highest perception frequency.
• Positive sentiment (73.09%) significantly outweighs negative sentiment in tourist evaluations.
• Within cultural perception: The History and Culture dimension shows the highest attention and satisfaction levels. The Architecture and Culture dimension has the lowest attention but relatively high satisfaction. The Local products and Culture dimension attracts high attention but the lowest satisfaction.

By investigating cultural perception characteristics and preferences in cultural tourism villages, this research provides insights and references for high-quality utilization and optimization of cultural tourism resources. Beyond the specific case of Dangjia Village, this study offers broader implications for cultural tourism development. The methodology integrating LDA topic modeling with IPA presents a transferable framework for other heritage destinations to systematically assess tourist perceptions from UGC.