AI Analysis of ACM Journal on Computing and Cultural Heritage
Quantitative Evaluation of Cultural Heritage Development in Ancient Towns from the Perspective of Tourist Experience
Authors: LEI YU, YUXIANG MA, GUOQIANG CHEN
Published: March 2026
The cultural heritage of ancient towns serves as a testament to history and a living carrier of culture. The preservation and development of this heritage not only contribute to cultural prosperity but also promote economic and social progress. However, the current approach to developing ancient towns lacks a universal set of evaluation criteria to meet tourist needs. This study introduces an automatic classification method based on machine learning. We selected 10 popular and successfully developed ancient towns in China as case studies. Using online tourist reviews as our data source, we first applied the Latent Dirichlet Allocation (LDA) topic model to identify the core elements of cultural heritage. Following this, we used the Naive Bayes classification method to distinguish between positive and negative tourist comments. From the perspective of the tourist experience, our results show that the evaluation of ancient town cultural heritage development depends on 8 key indicators, which consist of 20 specific aspects. Among these, Differentiated Landscape Experience was identified as the most important factor, followed by Personal Needs. Personal Needs also represented the most positive indicator, while Consumption Satisfaction was the most negative. The research proposes and validates an evaluation framework that deeply integrates machine learning with expert knowledge, overcoming the disadvantages of manual coding, which is labor-intensive and highly subjective. Additionally, this is the first study to investigate the common elements across multiple ancient towns. The findings provide guidance for the preservation and development of cultural heritage in ancient towns and offer a valuable reference for future research.
Enhancing Cultural Heritage Tourism with AI-Driven Insights
This study pioneers a robust, AI-powered framework for evaluating ancient town cultural heritage, addressing the critical need for objective, tourist-centric development criteria. By integrating advanced machine learning with expert validation, it offers a transferable model for sustainable tourism.
8
Key Evaluation Indicators
20
Specific Development Aspects
29,656
Tourist Reviews Analyzed
88.03%
Overall Positive Sentiment
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Differential Landscape Experience
This is the most salient factor (20.8%). Tourists highly value distinctive regional landscapes, local charms, and the town's reputation. However, a negative sentiment around 'typical' highlights dissatisfaction with over-commercialization and price gouging that compromise authenticity.
20.8%
Salience
-1.1%
Valence
Individual Needs (Recreational Fun)
The most positive factor (10.4% Valence) and second most salient (13.5%). This dimension focuses on attractive entertainment activities, cost-effectiveness, and the balance between price and value, reflecting tourists' desire for a rich and interesting experience within an affordable range. Negative sentiment around 'beautiful environment' points to service issues.
13.5%
Salience
10.4%
Valence
Consumption Satisfaction
This is the most negative factor (-1.6% Valence) and highlights challenges related to reasonable ticket pricing, appropriate control of visitor capacity, and quality of scenic area services. Widespread discontent with high prices and 'too crowded' conditions are key issues.
11.7%
Salience
-1.6%
Valence
Architectural Atmosphere Experience
This factor has the lowest salience (9.0%) but a negative valence (-1.2%). It emphasizes maintaining original architecture, preventing excessive commercialization, and creating a comfortable ambiance. Keywords like 'relaxation,' 'pandemic,' 'commercialization,' and 'comfortable' (when negated) reflect unmet expectations.
9.0%
Salience
-1.2%
Valence
Diversity of Experience
With a salience of 12.0% and a valence of -1.0%, this dimension covers the variety of activities, culinary experiences, and nocturnal events. While 'fun,' 'delicious,' and 'quaint' are positive, 'night scene' and 'evening' often carry negative connotations due to associated issues like over-commercialization and lack of local character at night.
12.0%
Salience
-1.0%
Valence
Night-time Experience
This theme has a salience of 10.0% and a positive valence of 0.5%. It focuses on the aesthetic beauty, suitable activities, and order during the night. Although 'night scene' and 'beautiful scenery' sometimes appear with negative sentiment due to other problems, overall, tourists appreciate the nocturnal landscape. However, 'lively' can indicate dissatisfaction with chaos.
10.0%
Salience
0.5%
Valence
Cultural Atmosphere Experience
This factor has a salience of 12.8% and a valence of -1.6%. It pertains to the deep historical and cultural substance of the ancient town and the quality of cultural communication services. Negative sentiments for 'explanation,' 'tour guide,' 'commerce,' and 'architecture' highlight issues with professional interpretation and resistance to over-commercialization that obscures historical charm.
12.8%
Salience
-1.6%
Valence
Spatial Atmosphere Experience
With a salience of 10.0% and a neutral valence of 0.0%, this dimension covers the natural environment (scenery, beautiful) and leisure-oriented ambiance (satisfaction, leisure, vacation, service). While the natural environment is appealing, 'satisfaction' can be negative when used in combination with 'not', indicating unmet overall expectations.
10.0%
Salience
0.0%
Valence
20.8%
Most Salient Factor: Differential Landscape Experience
This highlights the paramount importance tourists place on unique regional identity, local charm, and brand recognition of ancient towns. Over-commercialization and perceived inauthenticity are significant deterrents.
Enterprise Process Flow
Phase 1: Machine Learning Mining
→
LDA Model Topic Extraction
→
Naive Bayes Sentiment Analysis
→
Phase 2: Evidence Integration
→
Topic Keywords & Lexical Valence
→
Phase 3: Qualitative Decision
→
Delphi Expert Panel Consensus
→
Final Evaluation Criteria
| Feature | LDA & Naive Bayes (Current Study) | Advanced Models (e.g., BERTopic, SVM, LSTM) |
|---|---|---|
| Topic Modeling |
|
|
| Sentiment Analysis |
|
|
| Integration & Interpretation |
|
|
Insights from China's Ancient Towns
Subtitle: Analysis of 10 top-rated ancient towns from Ctrip.com
This study leveraged online reviews from 10 popular and successfully developed ancient towns in China, identified from Ctrip.com’s 2022 list, to provide a highly representative dataset for tourist perceptions.
- Data Source: Over 29,000 tourist reviews scraped from Ctrip.com, China’s largest online travel agency, ensuring a comprehensive and authentic dataset.
- Geographic Scope: Case studies include diverse regions such as Nanxun, Ping Yao, Lijiang, Fenghuang, Dali, Wuzhen, Zhouzhuang, Taierzhuang, Xi Jiang Qianhu Miao Village, and Hongcun.
- Relevance: Focuses on common elements across multiple ancient towns, making the findings widely applicable to similar cultural heritage sites globally.
- Preprocessing: Rigorous data cleaning, including stop-word removal and filtering of short/identical reviews, to enhance data quality and model accuracy. Specific place names were removed to ensure generalizable insights, preventing location-centric clustering.
Calculate Your Potential AI Impact
Estimate the efficiency gains and cost savings your enterprise could realize by implementing AI-driven insights from research like this.
Estimated Annual Cost Savings
$-
Annual Hours Reclaimed
-- hrs
Your AI Implementation Roadmap
Understand the phased approach to integrate AI-driven insights from this research into your enterprise strategy.
Phase 1: Machine Learning Mining
Raw tourist reviews were processed using the LDA model for latent topic extraction and the Naive Bayes classifier for sentiment polarity. This initial phase established the raw computational outputs.
Phase 2: Evidence Integration
The computational outputs—Topic Keywords and Lexical Valence (sentiment)—were converted into objective metrics. This allowed for a quantifiable understanding of 'what tourists are talking about' and 'how they feel about it'.
Phase 3: Qualitative Decision (Delphi Method)
A multidisciplinary Delphi expert panel utilized these objective metrics to reach a consensus, formalizing the final 8 evaluation criteria and 20 key points of meaning. This ensured the indicators were deeply grounded in actual tourist experiences and expert knowledge.
Ready to Transform Your Enterprise with AI?
Leverage cutting-edge research and AI strategies to drive innovation and efficiency. Schedule a free 30-minute consultation with our expert team to discuss how these insights can be applied to your specific business challenges.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat