Topic Modeling for Hiking Trail Online Reviews: Analysis of the Mutianyu Great Wall

Abstract

Hiking is now one of the most popular activities amongst adventure travelers. Although recent studies have highlighted the differences between Chinese adventure tourists and their international counterparts, few studies have comprehensively explored the differences in hikers’ interests and concerns for experience elements between these two groups. Topic modeling is adopted for an analysis of the online reviews of the Mutianyu Great Wall to identify attributes influencing hikers’ experiences and behavior. Using a large-scale review dataset, the latent Dirichlet allocation (LDA) technique was applied to construct a comprehensive list of the topics posted by hikers. The findings revealed that Chinese and non-Chinese hikers have common concerns regarding the degree of challenges, scenery, tour services and crowding during hiking. Nevertheless, their perceptions of cultural resources are presented in a different way. These findings are beneficial for understanding the similarities and differences between Chinese and non-Chinese hikers’ perceptions, in addition to improving domestic and international markets’ management and marketing strategies.

1. Introduction

Hiking is now one of the most popular ways to experience a destination and benefits the sustainable development of the respective destination. Hiking tourism has been popular in many countries and regions in recent years. For example, the number of people who participated in hiking in the United States increased from 29.86 million in 2006 to 47.86 million in 2018 [1]. In 2019, the participation rate of hiking participants in South Korea of different age groups over 19 ranged from 29% to 67% [2]. From 2010 to 2019, over 11 million visitors traveled to country parks in Hong Kong. The most common activities are leisure walking, hiking, fitness exercises and camping activities [3,4]. Hiking tourism can improve tourists’/residents’ physical and mental health, quality of life and wellbeing [5,6,7,8], promote local economic growth [9,10] and be an essential part of rural and cultural tourism [11,12]. Therefore, a destination could consider developing hiking tourism to benefit the local community and promote integration between visitors and residents. [13].

Previous studies that focused on exploring hiking tourists’ benefits sought motivations [14], determinants of experience quality [15,16,17], satisfaction and behavior intention [18,19]. Despite various common experience elements that have been unveiled in prior studies, the preferences of hikers from different cultural backgrounds seem to be underexplored. Significantly, recent studies have proposed that adventure tourism shows a high degree of homogeneity worldwide; however, some have highlighted the differences between Chinese adventure tourists and their international counterparts [20]. While some findings have implied the importance of tourists’ cultural backgrounds in tourism marketing and management [21], few studies have comprehensively explored the difference between Chinese and non-Chinese hikers’ interests and concerns for experience elements.

Previous studies have widely adopted qualitative and quantitative research methods to analyze hikers’ behavior; however, these studies must compromise between the cost of data collection and the issue of sample representativeness [22]. Social media is an excellent platform for people to express opinions and information regarding a particular attraction or destination [23]. These online comments or reviews convey visitors’ recommendations of tourism products, experiences, feelings, sentiments and moods. Online reviews can further reduce asymmetric information [22,23,24]. Therefore, online reviews can provide valuable insights into the hospitality industry and destination marketing [22]. Significantly, these user-generated contents allow for the exploration of the preferences of experiences across users from various backgrounds, such as their cultural backgrounds and related preferred experience elements. In order to explore the differences in experience elements between Chinese and non-Chinese hikers, this study adopts an approach to examine hikers’ experiences and behavior by utilizing online textual data. This study applies one of the topic modeling techniques, latent Dirichlet allocation (LDA), which has been employed in various topics in the tourism and hospitality research field, such as hotels [22] and theme parks [24].

Hikers want to be relaxed, by being close to nature along hiking trails, and experience the local culture through hiking [14]. Therefore, hiking trails’ natural and cultural resources may influence hikers’ decision making and experiences. The Great Wall is one of the Seven Wonders of the World and one of China’s best hiking trails [25]. In 2018, part of the Great Wall alone received over 9.9 million hikers [26]. The length of the Great Wall is over 20,000 km and it spans over 15 provinces, with conditions ranging from excellent to ruined [27]. The Mutianyu section of the Great Wall is one of the most popular hiking trails [28] because it retains most of its historical structures and, most importantly, is available for hiking [25]. The purpose of this study is to answer the following research questions: (1) What are the most important topics for hikers in online reviews? (2) What attributes that may influence hikers’ experiences can be summarized from these topics? (3) What are the similarities and differences in the topics mentioned by Chinese and non-Chinese hikers? The relevant topics of visitors’ experiences of the Mutianyu Great Wall are revealed based on online reviews. By utilizing online review data, this study will help researchers and managers understand hikers’ behavior. The contributions of this study are manifold: First, by adopting unprompted user-generated data and reducing the intrusive influence of the researchers, this study attempts to reveal and verify the important experience elements in hiking activity. Second, based on the topic modeling method, this study further explores hikers’ relative interests and concerns for different experience elements, providing empirical evidence for the marketing and management of hiking trails. Third, in line with the focuses of prior studies [20], this study unveils the differences in experience elements between Chinese hikers and their international counterparts. The findings of this study provide a further understanding of the domestic and international markets.

2. Literature Review

2.1. Hiking Experience

Hiking trails are not solely a way to arrive at a destination. Through hiking activities, tourists can obtain mobility experiences related to self-perception and personal growth [29]. Chhetri, Arrowsmith and Jackson [15] found that the hiking experience has four dimensions: desirable, impelling, apprehension and social interaction experiences. The desirable experience is desirable feelings, such as excitement, relaxation, etc. The impelling experience is the driving force, such as motivation and stimulation. The apprehension experience is hikers’ fear of the unknown. The social interaction experience is associated with crowding and challenging. Hiking is a multidimensional experience, and hikers usually have various motivations. Previous researchers have discovered various motivations, such as relaxing and escaping from daily routines [30], challenges and achievement [31], health and exercise [32], building friendships and being with family [30,32], as well as experiencing regional cultures [14].

Previous studies discovered that various factors might influence hikers’ experiences and satisfaction with hiking activities. These factors include the accessibility of the trails, the associated services and facilities [16,33,34], the degree of crowding [35], other visitors’ presence and behavior [36] in addition to the environment of hiking trails, including the natural and human features [33,37] that may influence visitors’ experiences and satisfaction. To illustrate,

Table 1. Category summary.

Category	Attribute	Keyword	Reference
Management	Facility	Pre-hiking facilities Trail facilities Accommodation facilities	[33,34,36,38]
	Service	Information Answer	[34,36]
Degree of difficulty		Risk Challenge	[34,38]
Accessibility		Public transportation Location Trail accessibility	[34]
Environment	Natural features	Landscape: animals, vegetation, water and geologic features	[16,17,33,36]
	Human (cultural) features	Constructed and marked paths Other hikers/local inhabitants Forestry/farm/district roads Pollution/erosion Industries and town agriculture Houses or other structures Power/telephone lines	[16,19,33,34,35,36,37]

provides a summary of influence factors.

The management, degree of difficulty, accessibility and environmental factors have been identified as essential elements to promote physical and mental rewards as well as benefit the local community [33]. For instance, in terms of management factors, to satisfy hikers’ needs and promote retention, it is essential to ensure operators’ service quality, such as providing sufficient information and education, as well as ensuring their safety [34]. The trail facilities, such as the design of trail steps and the quality of trail signs, are also important management issues. Well-designed trail facilities could ensure that hikers keep track of progress and form a pleasant experience [36]. Regarding the degree of difficulty, a prior study has identified that the appropriate difficulty could provide hikers with physical challenges, improve their health and energize their bodies [38]. The importance of the accessibility of hiking trails has drawn wide attention in prior studies. A study has found that a sound public transportation system and accessibility to hiking locations may positively enhance hikers’ satisfaction, and therefore promote their intention to revisit a location [34]. The natural and human elements encountered during hiking, such as environmental factors, could influence hikers’ experiences. For instance, the aesthetic of a trail is one of the vital elements during the hiking experience, consisting of various natural features that could also be damaged by these features [17]. Moore, Leung, Matisoff, Dorwart and Parker [16] proposed that various trail conditions, such as walking through a muddy trail segment, may influence the visual and aesthetic qualities of hikers’ experiences. Apart from these natural features, hikers could also encounter other hikers and, therefore, influence their experience positively or negatively. The influence of other hikers may be determined by an individual’s preference and other hikers’ behavior. Some hikers may form a positive impression from individuals’ behavior, while other hikers may prefer to stay away from crowding [36]. Though prior studies have unveiled these common experience elements, the preferences of hikers across different backgrounds seem to be underexplored. While some findings have implied the importance of hikers’ cultural backgrounds in hiking tourism [39], few studies have provided a comprehensive understanding of hikers’ interests and concerns for experience elements.

2.2. Comparison between Chinese and Non-Chinese Tourists in Outdoor Recreation

Culture is defined as the customs, values, beliefs, habits, traditions, expectations and lifestyle patterns that people or societies share [40]. The meanings and implications of Chinese culture for tourists’ attitudes [41], behavior, preferences and related marketing as well as strategies have drawn attention in prior tourism studies [42]. The cultural differences between Chinese and non-Chinese tourists could be reflected in various behaviors, such as decision making [43], complaint behavior [44] and the spatial distribution of destinations [45]. In outdoor recreation activity, many studies have paid attention to the different characteristics between Chinese and non-Chinese tourists. For example, Buckley, McDonald, Duan, Sun and Chen [20] explored the cultural heterogeneity of river-based adventure tourism between Chinese and non-Chinese tourists. Findings revealed that the Chinese domestic river-based adventure activity model is very different from international models. The results imply that the features of rafting products that could satisfy Chinese tourists may be disappointing to international inbound tourists. In a different context, Du, Buckley and Tang [39] further verified that Chinese and international inbound hiking tourists differ in their interests, motivations, decision making and factors affecting satisfaction. Gardiner and Kwek [46] showcased the concerns of Chinese Generation Y when they participate in adventure tourism. Compared with their Western counterparts, Chinese tourists emphasized familial obligation and reflected distinctive cultural beliefs. Compared with Western culture, the values of traditional Chinese culture also induce a conceptual difference in ecotourism in outdoor recreation, such as promoting health and a preference for enhancing nature [47].

Although prior findings have revealed differences between Chinese and non-Chinese tourists, the focus of these studies was limited to needs and decision making; few studies have explored the differences in the interests and concerns of experience components during travel. The usage of a large-scale online dataset provides a new way to investigate the effect of tourists’ cultural backgrounds on their experiences and behavior. Based on a large-scale online review dataset, the current study applies topic modeling analysis to investigate the differences and similarities in experiences amongst Chinese and non-Chinese hikers on the Mutianyu Great Wall.

2.3. Online Reviews and LDA

Online reviews have become increasingly crucial to studying consumer attitudes and purchasing intention [48]. Online reviews may change recipients’ perceptions of tourism-related products [49] and influence their decision-making behavior [50]. The availability of data, speed, low cost and simplicity of data collection are the advantages of using online reviews as a data source [22,51]. Online reviews convey the feelings, sentiments and moods of tourists [23], highlight the product/service attributes that tourists care about and reflect tourists’ detailed perceptions through an open-structure form [52]. Moreover, online reviews are a nonintrusive source due to minimizing the complications associated with direct interactions with human subjects [51]. Therefore, online reviews can be a significant, major information source for researchers to understand consumer preferences and demands [22].

Given that online reviews are usually text-based and unstructured, the methods used to analyze online reviews differ from traditional economic and statistical methods [22]. Data collection, preprocessing and pattern discovery are the three steps in analyzing online textual data [23]. Previous studies usually developed web crawlers to collect online textual data. Data preprocessing involved different techniques, such as data cleaning, tokenization, stemming and part-of-speech tagging. Pattern discovery aims to explore relevant information from online reviews, a crucial stage in data mining [23]; one of the techniques utilized for this is a topic model. A topic model is a probability model that aims to extract abstract “topics” from the text [22]. Topic modeling can be conducted via different mathematical frameworks, in which LDA is the most common method. Ultimately, LDA assumes that documents contain many random latent topics, where each topic has a topic–word distribution. LDA assumes that each document word is generated by a randomly chosen topic drawn from a document–topic distribution [53]. The multiple classifications of LDA for each textual content provide researchers with more detailed information. Moreover, the LDA method is an unsupervised machine learning process. It does not need to employ a dictionary, which is especially advantageous when the textual data contains misspelled words or reverts to colloquial language [54].

In the tourism research field, previous studies have adopted LDA for various topics, such as hotels [22,55,56,57,58], restaurants [59], heritage sites [60], destination images [61], theme parks [24], national parks [62], shopping experiences [63], travel itineraries [64] and tourism industry risk exposures [65]. This study uses LDA to explore visitors’ behavior and experiences in hiking tourism. The results can help managers manage the destinations more effectively and provide corresponding improvements.

3. Materials and Methods

3.1. Data Collection and Preprocessing

Various online platforms are available for online textual data collection. The current study collects English and Chinese online reviews of the Mutianyu Great Wall from different online platforms. In the tourism research field, English online reviews are collected from TripAdvisor, one of the largest social media platforms for travel consumers [22,23]. Many tourism studies collected Chinese online reviews from three online review platforms: Dianping, Ctrip and Qunar [23,48]. The current study developed web crawlers to extract the online textual reviews of online platforms. Other associated data, such as a user’s ID/name, review time and rating, are collected in addition to the online reviews.

Given that the online reviews are in textual form, successive techniques of text preprocessing should be adopted before LDA analysis. The most commonly used operations are tokenization, data cleaning, part-of-speech tagging (POST) and word stemming [23]. Specifically, in the tokenization step some special words, such as “Burger King”, are added to the dictionary to achieve better tokenization results. The data-cleaning step includes deleting repeated comments and destination names in addition to eliminating stop words, non-English characters or non-Chinese characters in English- and Chinese-based text. The tokens are then tagged by parts of speech in the POST step and filtered, except for nouns and adjectives. The word-stemming step reduces inflected words to their word stems. The current study incorporates two libraries in the Python programming environment to implement the preprocessing step for the English and Chinese datasets, respectively. These libraries are NLTK (https://www.nltk.org/) and Jieba (https://pypi.org/project/jieba/). Many studies have adopted these libraries to analyze various topics [22,61,63,66].

Table 2. Preprocessing example.

Step	Result
Original review	I’ve been fortunate to see this part of the wall three times and it never disappoints, especially with first-time visitors. It’s an easy visit, and well worth a visit. If you go in the winter, it’s incredible. Cold, but pretty.
Tokenization	'I', '’', have', 'been', 'fortunate', 'to', 'see', 'this', 'part', 'of', 'the', 'wall', 'three', 'times', 'and', 'it', 'never', 'disappoints', ',', 'especially', 'with', 'first', 'time', 'visitors', '.', 'It', '’', 's', 'an', 'easy', 'visit', ',', 'and', 'well', 'worth', 'a', 'visit', '.', 'If', 'you', 'go', 'in', 'the', 'winter', ',', 'it', '’', 's', 'incredible', '.', 'Cold', ',', 'but', 'pretty', '.'
POST	('I', 'PRP'), ('’', 'VBP'), ( have', 'RB'), ('been', 'VBN'), ('fortunate', 'JJ'), ('to', 'TO'), ('see', 'VB'), ('this', 'DT'), ('part', 'NN'), ('of', 'IN'), ('the', 'DT'), ('wall', 'NN'), ('three', 'CD'), ('times', 'NNS'), ('and', 'CC'), ('it', 'PRP'), ('never', 'RB'), ('disappoints', 'VBZ'), (',', ','), ('especially', 'RB'), ('with', 'IN'), ('first', 'JJ'), ('time', 'NN'), ('visitors', 'NNS'), ('.', '.'), ('It', 'PRP'), ('’', 'VBD'), ('s', 'PRP'), ('an', 'DT'), ('easy', 'JJ'), ('visit', 'NN'), (',', ','), ('and', 'CC'), ('well', 'RB'), ('worth', 'IN'), ('a', 'DT'), ('visit', 'NN'), ('.', '.'), ('If', 'IN'), ('you', 'PRP'), ('go', 'VBP'), ('in', 'IN'), ('the', 'DT'), ('winter', 'NN'), (',', ','), ('it', 'PRP'), ('’', 'VBZ'), ('s', 'JJ'), ('incredible', 'JJ'), ('.', '.'), ('Cold', 'NNP'), (',', ','), ('but', 'CC'), ('pretty', 'RB'), ('.', '.')
Removing stop words	('fortunate', 'JJ'), ('see', 'VB'), ('part', 'NN'), ('wall', 'NN'), ('times', 'NNS'), ('never', 'RB'), ('disappoints', 'VBZ'), ('especially', 'RB'), ('first', 'JJ'), ('time', 'NN'), ('visitors', 'NNS'), ('easy', 'JJ'), ('visit', 'NN'), ('well', 'RB'), ('visit', 'NN'), ('go', 'VBP'), ('winter', 'NN'), ('incredible', 'JJ'), ('cold', 'NNP'), ('pretty', 'RB')
Remaining nouns and adjectives	('fortunate', 'JJ'), ('part', 'NN'), ('wall', 'NN'), ('times', 'NNS'), ('first', 'JJ'), ('time', 'NN'), ('visitors', 'NNS'), ('easy', 'JJ'), ('visit', 'NN'), ('visit', 'NN'), ('winter', 'NN'), ('incredible', 'JJ'), ('cold', 'NNP')
Stemming	'fortunate', 'part', 'wall', 'time', 'first', 'time', 'visitor', 'easy', 'visit', 'visit', 'winter', 'incredible', 'cold'

provides a preprocessing example:

3.2. Topic Modeling: Latent Dirichlet Allocation (LDA)

The current study applies the LDA algorithm to extract hidden topics in online reviews. The Genism library (https://pypi.org/project/gensim/) and the mallet package (http://mallet.cs.umass.edu) is used in the Python programming environment to accomplish this task. The library and package have been used in prior studies [67,68]. Prior studies usually adopted perplexity and coherence indices to find the optimal number of topics [24,68]. Given that the perplexity minimization index may return with a very large number of topics [68], this may lead to poorly interpretable topics [54]. Hence, this study utilizes the topic coherence index and combines it with a manual method to find a small number of interpretable topics [68]. Topic coherence measures the degree of semantic similarity among words with high scores in the topic. The underlying idea is that words with similar meanings tend to occur in similar contexts [69]. The topic coherence will improve rapidly when the number of topics is small, and the improvement will gradually decelerate when the number of topics is large. The elbow method can help locate the optimal number of topics. The final number of topics can be determined through the interpretability of topics.

Various factors may influence the LDA’s performance. In addition to the number of total online reviews and length of each online review, alpha and beta, the Dirichlet distribution’s hyperparameter also influences the results [70]. Therefore, we adopted the mallet library’s hyperparameter optimization option to set the proper value, which allows the model to fit the data better (http://mallet.cs.umass.edu/topics.php).

3.3. Data Collection and Preprocessing

This study collected Mutianyu Great Wall online reviews and associated username/id, review time and ratings from October 2005 to August 2020. We extracted the online reviews of domestic hikers in Chinese from three online platforms: Dianping, Ctrip and Qunar. We extracted the online reviews of foreign hikers, mainly in English, from TripAdvisor due to the language barrier. One of the current study objectives is to compare the topics’ similarities and differences between Chinese and foreign hikers. Therefore, the time at which the reviews were posted should be limited to the same period. Given that the earliest post time of Chinese reviews was in 2007, we deleted the English online reviews from 2005–2006 from the dataset.

Table 3. Preprocessing example 2.

Platform	Number of Reviews	Reviews in English/Chinese
TripAdvisor	16,426	16,426
Dianping	7235	14,765
Qunar	4959
Ctrip	2571
Total		31,191

shows that we collected 31,191 online reviews of the Mutianyu Great Wall, including 16,425 in English and 14,765 in Chinese. Before the LDA analysis, we applied numerous preprocessing steps, as described above. We filtered some online reviews through this process. The final numbers of online reviews were 16,419 in English and 13,555 in Chinese.

After the preprocessing steps, we calculated the rating scores of the reviews. According to the findings of [68], compared with high-rating reviews datasets, the performance of the LDA analysis is significantly lower when online reviews datasets contain a large number of low ratings. In the current study, the proportion of online reviews with a 4–5-star rating in the Chinese and the English datasets was 97% and 98%, respectively. Therefore, the datasets are suitable for LDA analysis.

4. Results

4.1. Topic Discovery

Two important parameters exist in topic discovery and LDA: alpha and beta, also known as hyperparameters. In many studies, the default value of the alpha is 50/number of topics, and the beta is set to 0.01 [71,72,73]. However, some studies also adjusted the parameters to obtain more interpretable topics. For example, Jacobi, van Atteveldt and Welbers [71] compared the performance of LDA when the alpha was set to 50/number of topics and five/number of topics, and found that the latter alpha value can yield better outcomes. We adopted the mallet library’s hyperparameter optimization option to set a proper value in the current study. Given these optimal hyperparameters, the LDA analysis process and the topic coherence calculation are repeated 10 times for each topic. The final topic coherence curves of the English and Chinese datasets are shown in Figure 1. According to the topic coherence curves, the improvements in the topic coherence of the English and Chinese datasets tend to slow down when the number of topics is five. Therefore, we qualitatively interpreted the five topics in the corresponding dataset. Then, we gradually increased the number of topics to find the proper number of topics. Finally, the optimal number of topics was set at 10 for the English dataset and nine for the Chinese dataset. This study adopts the same methodology of naming topics as that used in Guo, Barnes and Jia [22] as well as Taecharungroj and Mathayomchan [74]. A researcher identified a topic according to the logical connection between the top 15 words in each topic–word distribution, and another researcher confirmed and refined the topic [22,74].

Table 4. English online review topics.

Topic	Top Words
Difficulty and challenge	steep, step, tower, climb, top, high, stair, hard, difficult, fitness, uneven, challenge, hill, path, young
Tour services	tour, guide, trip, hotel, private, driver, lunch, English, excellent, group, company, Chinese, service, review, booking
Recreation elements	toboggan, fun, lift, great, chair, ride, slide, ski, mountain, kid, luge, child, hill, family, safe
Service facilities	area, shop, restaurant, small, price, food, vendor, souvenir, drink, Chinese, toilet, local, park, good, expensive
Crowding perception	time, hour, people, crowd, early, holiday, morning, place, busy, location, drive, quiet, group, minute, traffic
Cultural experiences	amazing, experience, year, history, world, site, wall, incredible, mile, sight, structure, impressive, magnificent, feel, real
Essentials and related conditions	lot, water, shoe, comfortable, summer, hat, wear, worth, walk, hot, plenty, top, cool, stair, place
Hiking options	car, cable, option, walk, top, easy, access, short, long, open, uphill, strenuous, point, part, left
Transportation	bus, ticket, station, entrance, driver, shuttle, stop, taxi, hour, money, trip, yuan, RMB, cost, return
Climate and scenery	day, view, experience, beautiful, weather, fantastic, wonderful, clear, cold, winter, perfect, sky, sunny, awesome, scenery

and

Table 5. Chinese online review topics.

Topic	Top Words
Hiking options	cable car, rope way, children, old man, suggestion, downhill, less, time, feeling, physical strength, ferry car, uphill, good, hour, queue, beacon tower
Tour services	guide, Mubus, enthusiasm, ZanBus, time, great, itinerary, driver, experience, patience, pleasure, get on the car, whole journey, neat, detail
Social identity	wisdom, ancient, history, people, wall, miracle, the world, architecture, majestic, magnificence, motherland, culture, pride, engineering, hero
Ticketing services	QR code, cheap, on-site, mobile phone, electronic, change ticket, queue up, scan, fast, booking, enter, ticket, SMS, fast, speed
Difficulty and challenge	step, feeling, route, brave, leg, place, high, steep, hand, difficult, stair, tired, downhill, tower, less
Climate and scenery	weather, autumn, red maple leaves, beautiful, scenery, season, view, winter, snow, spring, cold, rain, air, summer, spectacular
Service facilities	souvenir, stuff, price, water, expensive, food, car park, Burger King, ice-cream, drink, clean, restaurant, Subway, snack, worn
Crowd perception	less, tourist, foreigner, feeling, crowd, traffic, suggestion, friend, long, experience, good, spot location, companion, fellow
Transportation	Huairou district, Dongzhimen, bus, transportation, North Street, fast, car, road, driver, time, subway, hour, express, special line, roundabout

provide the final topic names and top 15 words in each topic. For interpretation, we further grouped these topics (see

Table 6. Topic grouping.

Category	Attribute	Topics
Management	Facility	Service facilities (En, Ch) Recreation elements (En)
	Service	Tour services (En, Ch) Ticketing services (Ch)
Degree of difficulty		Difficulty and challenge (En, Ch) Essentials and related conditions (En)
Accessibility		Transportation (En, Ch) Hiking options (En, Ch)
Environment	Natural features	Climate and scenery (En, Ch)
	Human (cultural) features	Crowding perception (En, Ch) Social identity (Ch) Cultural experience (En)

) based on the summarized attributes as described in the “Hiking Experience” section and Table 1.

The topics extracted from the online reviews are consistent with prior research. The management, degree of difficulty, accessibility and environment of hiking trails may influence hikers’ experiences, behavior and satisfaction [33,34,36,38]. The management of hiking trails involves facilities and services. Two topics, “service facilities” and “recreation elements”, are extracted from hikers’ online reviews. Recreation elements are rarely mentioned in prior research. Oh, Kim, Choi and Pratt [14] found that a hiking trail’s recreation facilities are one of three factors that may determine hiking tourism demand. As shown in the top words in this topic, recreation elements include some facilities and services provided by management organizations, such as a “toboggan” or “ski”, which may influence hikers’ recreational experiences. This study discovered two topics within the management category: “tour services” and “ticketing services”, which are related to the service attributes. Service is an integral part of hiking tourism [33]. Mohd Taher, Jamal, Sumarjan and Aminudin [34] proposed that an organized company’s service is one of the pull factors that may influence hikers’ satisfaction and intention to revisit a location. The most frequently mentioned words in “tour services” are related to tour guides and drivers, whereas the most frequently mentioned words in “ticketing services” are price, speed and paying methods, such as electronic tickets and online booking.

Hiking trail difficulty also influences hikers’ experience [17]. One of the benefits sought by hikers is physical challenges [38]. Two topics are identified in this category, namely “difficulty and challenge” and “essentials and related conditions”. The former topic means the steepness of hiking trails as well as hikers’ physical and mental feelings, whereas the latter topic means the temperature during the hiking activities and related essentials, such as water and shoes. Mohd Taher, Jamal, Sumarjan and Aminudin [34] uncovered that hiking trail difficulty may influence hikers’ satisfaction and intention to revisit a location. However, hikers may have different preferences for trail difficulty [33]. The current study implies that hiking trail managers must pay attention to trail difficulty and related essentials.

This study extracts two topics related to the accessibility of hiking trails, namely “transportation” and “hiking options”. The meaning of the former topic is the methods used to travel to hiking sites in general, whereas the latter topic means the routes that hikers can choose when going up and down the mountain through a cable car. Moreover, specific groups related to hiking options and the price of public transportation are also discovered in the most frequently mentioned words. The accessibility of hiking trails positively influences hikers’ satisfaction [34]. However, “hiking options”, which highlight the different demands for hikers’ segments, have only been studied to a limited extent.

In terms of the environmental attributes, previous studies focused on the natural and human/cultural features of hiking trails. Natural resources, such as natural landscapes [34,36], may influence hikers’ experiences. The topic “climate and scenery” is extracted in the current study, which involves season, weather, temperature, visibility and related scenery. The role of climate has rarely been discussed in prior hiking experience studies. Martínez-Ibarra, et al. [75] found that hikers have specific preferences towards climate conditions. Specifically, climate conditions involve various aspects, such as thermal comfort, season and visibility.

Three topics related to human/cultural features were extracted: “crowding perception”, “personal identification” and “cultural experiences”. The influence of crowding perception on hikers’ experience and satisfaction has been explored in prior research [35,36]. Dorwart, Moore and Leung [36] revealed that the presence of other hikers positively and negatively affects hikers’ experiences. In this study, crowding is related to peak times and the presence of other hikers. Cultural experiences have been found to motivate hikers [14]. Vidal-González and Sánchez [11] unveiled that hiking trails are essential for cultural tourism. According to the most frequently mentioned words in the topics, hikers’ cultural experiences and social identity are usually elicited by the Mutianyu Great Wall’s structure and history.

4.2. Topic Distribution Analysis

The probability distribution of topics may reflect the popularity of topics [24] amongst hikers. Figure 2 shows that hiking options are the most mentioned topics amongst Chinese hikers, followed by crowding perception and tour services. The most frequently mentioned words indicate that Chinese hikers pay more attention to their family members’ hiking capability, especially those who require care, such as children and the elderly. Contrarily, the most popular topic of foreign hikers is recreation elements, followed by “transportation” and “climate and scenery”. According to the most frequently mentioned words in the most popular topic, recreation elements, such as toboggans, are related to young family members, which indicates that foreign hikers pay more attention to family entertainment and recreation when hiking. Furthermore, the most popular topics in Chinese and foreign hikers also highlight their motives to seek social interaction. To increase intimacy with family members is one of the motivations of hiking tourists, which positively influences hikers’ subjective wellbeing and intention to revisit a location [32].

Moreover, our results also identified some differences between Chinese and foreign tourists. For example, Chinese hikers seem to pay more attention to ticketing services than foreign hikers. This finding may be relevant to different crowding perceptions between both groups. As mentioned above, Chinese hikers’ second most popular topic is “crowding perception”. Therefore, a fast and convenient ticketing service may positively influence their hiking experiences. In addition, the topic “essentials and related conditions” is only extracted from the English dataset. Given that Chinese hikers can obtain more detailed and accurate information about local hiking trails’ conditions than foreign hikers, this topic may be more frequently mentioned by foreign hikers. Chinese and foreign hikers have different perceptions of the cultural elements of the Great Wall. The long history and gorgeous culture of China, manifested by the Great Wall, elicits Chinese hikers’ social identity, which can be inferred from the top words in the topic, such as “motherland”, “pride” and “hero”.

5. Discussion

This study provides some theoretical contributions. The current study applied the topic modeling technique to discover the experiences and behavior of hikers, as well as to reveal the influence of hikers’ cultural backgrounds, which have only been studied to a limited extent. We adopted a topic coherence index, combined with a manual method, to extract more interpretable and nonredundant topics from online reviews [68]. The results are consistent with findings from previous studies and provide detailed information. A comprehensive list of hikers’ topics was constructed on the basis of previous studies, and these topics were further classified into six attributes under four categories. For example, prior studies have widely discussed the management category as an influencing factor of hikers’ experiences and behavior [33,34,36,38]. In the current study, specific topics, such as recreation elements and tour services, can be classified into the management category. Furthermore, some findings in this study could also complement prior research. For instance, the role of climate elements, such as temperature, weather, season and visibility, in hikers’ experiences, as was discovered in this study, has rarely been discussed in prior research; this can be confirmed from the findings of Martínez-Ibarra, Gómez-Martín, Armesto-López and Pardo-Martínez [75]. Some review examples that may reflect these climate elements and hikers’ feelings are presented in

Table 7. Review examples.

Aspect	Top Words	Review Example
Visibility	clear, air, sky	the air was clean, the sky clear and perfect for the 21/2 h walk up and down… (En39)
		…it is very clear to see the surrounding mountains, which is very beautiful… (Ch273)
Thermal comfort	cold	We were not quite prepared for the cold, foggy, damp day at Mutianyu. Wearing only light jackets over short sleeves. We needed to buy sweatshirts for extra warmth… (En861)
		…It’s cold. If you need to hold on to the iron railing, remember to wear gloves… (Ch2827)
Weather	snow, rain, sunny	… I could imagine that walking on the Great Wall in the rain is not completely safe. (review En1250)
		It’s so cool to hike in the rain… (Ch3176)
Season	spring, summer, autumn, winter	…Loved the view because we went in October where autumn foliage were simply everywhere… (En1417)
		It has always been my wish to see the forests all over the mountains in autumn… (Ch3387)

. As shown from these examples, climate elements are associated with various aspects of hiking activity, such as senses, natural scenery, preferences and safety concerns, which call for the further study of the influence of climate. The accessibility of hiking trails has been discovered as an influencing factor of hikers’ experiences [34]. The cost of transportation topic extracted from the English dataset reveals more specific concerns in this category. The degree of hiking trail difficulty might influence hikers’ satisfaction [34]. The current study reveals that the essentials and conditions related to difficulty during hiking activities may also influence hikers’ experiences. In terms of cultural differences, the topics of the two datasets reflect the similarities and differences between Chinese and foreign hikers. On the one hand, family members seem to be a very important topic to Chinese and foreign hikers. Bichler and Peters [30], Kim, Lee, Uysal, Kim and Ahn [32] and Li, et al. [76] found that social interaction is one of the motivations of hiking tourism in Asian and Western contexts. For example, Li, Yang, Wei and Zhang [76] discovered that social interaction motivation is of primary importance to Chinese hikers. Similarly, this motivation has been identified in German, Austrian and Italian contexts [30]. On the other hand, foreign hikers pay more attention to the topic related to recreation elements, whereas Chinese hikers focus on ticketing services. The difference may be explained by cultural influences. For example, Tsang and Ap [77] discovered that Asian and Western tourists evaluated relational quality attributes differently.

Similar and differential topics could contribute to hiking trail management. In addition to the basic attributes identified in this study, such as tour services and scenery, our results show that Chinese and foreign hikers seek social interaction in hiking. This result indicates that managers should emphasize marketing strategies’ social interaction benefits to attract hikers. For example, some family activities or recreational facilities may enhance family members’ social interaction. Moreover, managers should accommodate the differences between Chinese and foreign hikers. For foreign hikers, managers should pay close attention to recreation elements and design more satisfying family activities. Moreover, foreign hikers seem to require more information, such as transportation costs and essentials related to specific hiking trail conditions. Therefore, more convenient payment methods as well as more accurate and detailed information related to these attributes should be provided to foreign hikers. Contrastingly, Chinese hikers pay more attention to their family members’ physical capability during hiking and are more sensitive to crowding perception. Thus, managers should focus on the needs of the elderly and children, and provide them with related services, such as hiking gear and safety measures. Managers should rate the difficulty level of different path sections to provide more reliable decision-making information for tourists with different skill levels. Furthermore, managers should provide timely feedback on congestion information and perform good crowding management. Both Chinese and international hikers are concerned about climate and scenery elements. Trail managers could provide more detailed information regarding hiking gears corresponding to specific weather conditions. In addition, the scenery of hiking trails varies by season, providing various natural environments for promoting different sightseeing themes to satisfy niche markets. The salient difference between Chinese and foreign hikers is their perceptions of cultural resources. Since cultural resources are important elements for attracting hikers, this finding indicates that hiking trail managers should pay attention to different marketing orientations. Trails managers should emphasize the cultural meanings of hiking trails. For example, more information should relate to domestic hikers’ social identity and emphasize the ethnic history in terms of Chinese hikers. For foreign hikers, the marketing information could highlight the uniqueness of the cultural resources to satisfy their novelty motivations and authenticity perceptions.

6. Conclusions

The development of hiking tourism can be viewed as a means of welfare policy, due to various benefits [13], and an essential part of rural and cultural tourism. As a non-intrusive source, online reviews have emerged as useful data sources. However, studies that focus on the potential of online reviews in exploring hiking tourism experiences and the comparison between Chinese and international hikers remain scant. Thus, this study aimed at filling this research gap by applying unprompted user-generated content. The findings of this case study were consistent with prior research and provided useful as well as practical implications. Furthermore, the findings attempted to reveal hikers’ relative interests and concerns for different experience elements by taking advantage of the topic modeling method. Following the focuses of prior studies, this study paid attention to the differences between Chinese hikers and their international counterparts. The findings revealed that both Chinese and non-Chinese hikers have common concerns in terms of the degree of challenges, tour services, crowding, service facilities, ticketing services and climate as well as scenery elements. In contrast, their perceptions of cultural resources, entertainment elements and hiking gears were presented differently. The findings of this study shed light on the crucial role of hikers’ cultural backgrounds in their outdoor recreation experiences.

Several limitations should be discussed, despite utilizing a large-scale dataset. Given that the essential part of the Great Wall is a cultural heritage, the findings of this study may be inapplicable to other types of hiking trails. Given that the current study only focused on the Mutianyu section of the Great Wall, the results and recommendations may not be generalized to other sections of the Great Wall, due to variations in preservation degrees, facilities, difficulties and other characteristics. We only collected Chinese and English reviews due to language barriers; therefore, the topics extracted from this study may not cover all possible concerns and attributes. Moreover, the amount of data may influence topic extraction and comparison; therefore, a larger dataset may improve the performance of the LDA analysis. In addition, various national cultural backgrounds of international hikers may also differ in their preferences; therefore, future research could further adopt the multilingual topic modeling approach to provide a more detailed comparison among hikers from various cultural backgrounds. In terms of online reviews, future studies can collect online reviews from more platforms to reduce data bias and improve the performance of the LDA analysis. Moreover, an analysis of reviewers’ demographics, such as age and gender, combined with topic distribution could provide more detailed findings. Additionally, other analytical methods, such as sentiment analysis, could complement the LDA analysis and provide more comprehensive findings.