Understanding the contents of smart tourism comprehensively can not only provide abundant guidance and reference for the practices of smart tourism, but also help to identify and overcome the shortcomings of smart tourism research. Smart tourism, as a new form of tourism, is applied in tourism experience, industry development, administration, and other aspects by Internet of things, cloud computing, smart data mining, and other technologies. The development process of smart tourism matches the core ideas of creative computing indicating the combination of knowledge from various disciplines and the integration between creativity and technology, which can provide a new way of thinking for computer-aided innovation [1]. It is believed that creative computing may be a good way of grasping the context, hotspots, and trends of the development of smart tourism by analyzing domestic research on smart tourism.

By extracting data from smart tourism literature, it is found that most of the smart tourism research is descriptive and lacks in-depth exploration [2]. Seven papers and 69 reports on smart tourism at www.ckni.net concluded that smart tourism research remains at a stage where practice facilitates theoretical development [3]. Smart tourism research currently remains immature, lacking both depth and width [4]. Most academic smart tourism research remains at the level of concepts [5-6], connotations, frameworks, values, applications and development trends related to smart tourism [7], as well as the relation between smart tourism and tourism informatization [8-9]. However, a problem of the existing reviews of the smart tourism literature is the lack of adequate references and analysis of all data on smart tourism literature. As a result, the comprehensiveness of the research conclusions cannot be guaranteed. Most of the research adopts qualitative methods. Quantitative research methods have not been paid much attention in smart tourism literature review.

The quantitative research methods used to identify the main research contents in a field are mostly interdisciplinary knowledge combination, which is a typical case of creative computing. By using the literature research method for knowledge discovery, Gao et al. [10] designed a knowledge discovery model based on document data visualization and conducted cluster analysis and a semantic network with a co-occurrence matrix of high-frequency keywords. Aiming at the research status of Scientometrics in the context of visualization, starting from the spatial and temporal dimensions, Chen et al. [11] summarized the main contents of current research, using word frequency analysis and co-word analysis to identify, for example, subject topics and research fronts. Using the analysis integration software Ucinet [12], VOSViewer [13], CiteSpace [14], and other tools, some scholars conducted clustered analyses or drew visual network diagrams after constructing high-frequency keyword co-occurrence matrices [15]. All existing quantitative analyses on research hotspots screen high-frequency keywords, construct a co-occurrence matrix, and then perform cluster analysis or draw a knowledge map to identify hotspots in relevant fields based on keyword analysis. To an extent, such practices may lose important information implied in low-frequency keywords and result in one-sided research conclusions. Furthermore, the cluster results based on the co-occurrence matrix are subjective because they require manual identification of different categories of keywords for interpretation, which involves an enormous workload in situations involving large amounts of data and high-frequency keywords.

Covering all the journal papers dealing with smart tourism research on www.ckni.net, this paper analyzes the cluster based on the word frequency matrix of the keywords and conducts an in-depth discovery on the cluster results by selecting highly relevant keywords based on the chi-squared statistics. In this way, the paper presents the characteristics of clustering results comprehensively and objectively and identifies the main research ideas of each category.

Keywords are important attributes of academic research that can reflect the essence of papers and play a critical role in research contents. Keywords therefore provide an important basis for research on the most popular research topics within an academic field.

The CNKI database is used as the data source of this paper. “Smart Tourism” is used as the keyword to search the CNKI journal database. The selected time point is before 2018. A total of 1219 related articles were retrieved, and the key words of these articles constitute the data used in this study.

The data set is imported into Excel to construct a word frequency matrix for the keywords, and this matrix serves as the data source for the cluster analysis. The partial word frequency matrix is shown in Table 1 below.

As shown in Table 1, each line in the word frequency matrix represents a paper, while each column represents a keyword, where 1 indicates the keyword is among those of the corresponding paper and 0 indicates it is not among the keywords of the paper.

The keyword word frequency matrix is imported into SPSS software for cluster analysis, and the intra-group connection method with better cluster effect is used for system cluster analysis. Because the word frequency matrix built in this paper is a 0-1 binary matrix, the simple matching coefficient is adopted to measure the distance or similarity between the papers. As for the keyword word frequency matrix, the simple matching coefficient is the frequency at which all keywords share the same values for the two papers.

The simple matching coefficient S may be expressed as

As shown in Table 2, “a” represents the number of keywords that are included in both papers, d represents the number of keywords included in neither paper, and b andc represent the numbers of keywords included only in each one of the two papers, respectively. “a+b+c+d” represents the total number of keywords, i.e., the number of variables in the word frequency matrix.

The pedigree of the system cluster is shown in Figure 1. Based on the research purpose and classification effect, the current smart tourism research can be assigned among 18 categories according to the distance standard shown by the red line. The cluster results are shown in Table 3.

Table 3 shows a gap in the distribution of papers among the categories. Category 10 is the largest category, containing a total of 516 papers, and is followed by Categories 2, 12, and 13, each with 80-100 papers. The smallest categories are Categories 17 and 18, each of which contains just one paper, and because both these papers involve technological information they are ignored here. Each of the remaining categories contains only a small number of papers.

The large number of papers in each category and the resultant large number of keywords create certain difficulties in the description of the cluster results. In this paper, the relevance between keywords and categories is calculated using chi-squared statistics, and the cluster results are analyzed and described by selecting keywords with high relevance. The calculation formula is as follows:

In the above formula, “n” represents the total number of papers sampled from the relevant literature. “a” represents the number of papers that belong to Category “c_i”and include keyword “t”, “b” represents the number that do not belong to Category “c_i” but include keyword “t”, “c” represents the number that belong to Category “c_i” but exclude keyword “t”, and “d” represents the number of papers that neither belong to Category “c_i” nor include the keyword “t”.

The chi-squared value CHI between the keyword and the category is obtained based on chi-squared statistics, and a relevance ranking of the keywords and categories can be obtained by sorting keywords and the CHI value of each category in descending order. The higher the relevance, the stronger the ability to distinguish between Category “c_i” and other categories. This paper also describes the ten most relevant characteristic words for each Category c_i. The extracted high-relevance keywords are shown in Table 4.

Table 4 shows a clear difference between categories in terms of highly relevant keywords and degrees of similarity among keywords in the same category, which can better show the main creative content of each category.

The high relevance of the keywords in Category 1 indicates that the category includes research on smart tourism talent cultivation, which involves, for example, specialized settings and training patterns. The research papers in Category 2 focus on the current state of tourism and propose corresponding development strategies, models, and strategic responses, and so this category can be defined as the status quo and strategic responses category. The high relevance keywords in Category 3 indicate that this category deals with smart tourism big data research, covering, for example, databases, data mining, data centers, and big data platform. Category 4 contains the largest number of papers, and high relevance keywords in this category involve information, informatization, smart city, data, and tourism information, indicating that this is the tourism informatization category. Category 5 focuses on tourism service, service supply chain, and tourism public service, and it can be described as the tourism service supply chain category. The high relevance keywords in Category 6 fully reflect the Internet, mobile Internet, and similar contents, so this category can be defined as the Internet + tourism research category. Category 7 includes keywords such as city construction, smart city, and pilot demonstration, and hence it can be considered the smart tourism city construction research category. Category 8 contains keywords related to tourism public service system, so it represents the smart tourism public service system category. The high relevance keywords in Category 9 include tourism experience, tourist experience, and user experience, making this the tourism experience research category. Category 10 comprises the smart tourism public information service research category. The high relevance keywords for Category 11 include smart tourism city, regional coordinating, urban agglomeration, and integration, so this is the smart tourism city cluster research category. The high relevance keywords in Category 12 include terms such as tourism informatization, information system, information technology, and wisdom, indicating that this is the tourism information technology research category. The high relevance keywords for Category 13 include terms such as smart scenic area, scenic area management, and tourist, indicating that this is the smart tourism scenic area management research category. Category 14 can be defined as the smart tourism public service platform research category. Category 15 contains keywords such as tourism enterprises, enterprise management, business transformation, and wisdom of tourism enterprises, indicating that this is the tourism enterprise wisdom category. Finally, the high relevance keywords for Category 16 include tourism industry, industrial chain, modern tertiary industry, and thinking form, making this the smart tourism industrial chain research category. For definitions of each category, please see Table 5 below.

Changes in the numbers of papers published in each creative content category over time are shown in Table 6.

In terms of the literature amount, research on tourism informatization ranks first. There are 516 references about tourism informatization, accounting for 42.4% of the total. Tourism informatization research grew rapidly before 2015. Although there was a downward trend in 2016 and 2017, the number is still large, and it is a hotspot in smart tourism research.

There are also many studies on the status quo and counter measures of smart tourism, tourism informatization technology, and smart tourism scenic area management. There are over 88 articles about each issue. The number of articles studying the status quo and counter measures has shown an upward trend in recent years while research on the smart tourism scenic area management has shown a downward trend; research on tourism informatization technology has shown volatile fluctuations and dramatic changes over time.

The total amount of research on tourism-related big data, tourism service supply chain, Internet + tourism, and tourism experience is not large. These issues were rarely studied before 2013. However, it has shown a clear upward trend since 2014, which is very worthy of attention.

There are few studies on talent cultivation, smart tourism city construction, smart tourism public information service, smart tourism city cluster, and smart tourism industrial chain. Among them, studies on smart tourism public information service are quite unstable. In addition, studies on smart tourism city construction, smart tourism city cluster, and smart tourism industrial chain all show a downward trend.

Studies on smart tourism public service system, smart tourism public service platform, and smart tourism enterprises are the most limited.

According to the literature time, articles related to smart tourism began to appear after 2011.In 2014, there were over 200 journal articles about smart tourism research, and it reached itspeak in 2015 and developed stably in 2016 and 2017. The reason is that research on smart tourism is obviously influenced by industry policies: in 2010, China proposed the construction of smart cities, and smart tourism began to appear as a part of the smart city. The China National Tourism Administration proposed that the year 2014 should be regarded as the “Smart Tourism Year”. The research results in 2014 and 2015 have reached a new level. In 2018, the Ministry of Culture and Tourism of the People’s Republic of China was established. Research on culture and tourism is bound to be popular in the future.

As for the literature content, smart tourism research highlights multi-disciplinary integration and integrates the development of tourism industry creativity into the process of information technology development. Combining the tourism big data developed in 2014, the Internet + tourism proposed in 2015, the subsequent tourism service supply chain, and the tourism experience and tourism informatization technology, there were a total of 795 articles on tourism information research accounting for more than 65% of the total. The application of the emerging informatization technology in tourism will be an important creative content of smart tourism research.

By selecting journal articles about smart tourism research in China, this paper constructed a word frequency matrix of keywords and carried out the cluster analysis. It also introduced the chi-square statistic and selected keywords with high correlations to interpret the clustering results. The number of articles ensures the comprehensiveness of the analysis results. High-frequency keywords and low-frequency keywords are not distinguished and screened to avoid the omission of valuable and important information. Keywords with high correlations provide convenient methods for the description and deep analysis of the clustering results.

Based on the analysis of the literature amount, time, and creative content, it can be found that there is much research on tourism informatization, status quo and counter measures of smart tourism, tourism information technology, and smart tourism scenic area management. Among them, tourism informatization, the status quo, and counter measures have been hotspots in recent years. Research on tourism big data, tourism service supply chain, Internet + tourism, and tourism experience has attracted more attention andmay be the research hotspots in the future. The impact of tourism industry policy and the application of new information technology have always been an important source of creativity in smart tourism research.