Keywords

1 Introduction

This Chapter presents the results of research activities conducted to develop a digital application for the strategic development of evidence and nature-based health tourism (NHT) destinations. The Health Tourism Assessment and Benchmarking (HTAB) tool aims to provide customized recommendations for developing health tourism products in Alpine destinations in order to further exploit the NHT market.

Tourism is a major force for economic growth, job creation, and sustainable development in the Alpine area. Additionally, nature-based experiences and products are increasingly growing due to their known benefits on people’s physical and mental health. This trend brings substantial opportunities for developing innovative NHT knowledge and implementation tools in the Alpine regions.

Health tourism is a type of tourism in which people travel to specific destinations in order to receive particular medical treatments or enhance their physical/mental health and general well-being [1, 2]. Health tourism is divided into three different yet overlapping components: medical tourism, wellness tourism, and spa tourism [3]. While medical tourism foresees traveling to a particular destination to access medical treatment, wellness tourism is associated with traveling to maintain, strengthen, and rejuvenate the tourists’ physical or mental health [4]. Spa tourism, however, is the intersection of medical and wellness tourism to combine the medical treatment and healing process together.

Wellness tourism is a broad concept, and its meaning depends on the geography and culture of each tourism destination. While in southern Europe, wellness tourism is associated with the seaside and Mediterranean culture, it is more connected to swimming and hiking in northern Europe. Alpine regions appear to be among the most popular health travel destinations for rewarding elements such as indulgence, leisure, and regeneration, combined with more challenging and stimulating elements, including outdoor activities and sports (hiking, mountain biking, and golf) [5]. The unique Alpine natural resources such as healing waters, forests, and waterfalls play a significant role in the exploitation and sustainability of health tourism in Alpine destinations. Identifying these potential assets and resources in different regions within the Alpine area can lead the regional stakeholders and policy-makers to further cooperate and reap the benefits of a new value chain [6].

In this context, this Chapter investigates the development of the HTAB tool based on the ontology implemented within the Decision Support System (DSS) described in Chap. 4 [7] to identify the natural resources of different Alpine destinations and provide customized recommendations accordingly. Starting from an online survey filled by the destination, the HTAB tool cleans, processes, and translates the data into a semantic representation of the domain knowledge to reason and infer the output results. The HTAB tool results are then sent to the destination managers to be used by tourism stakeholders to better understand the potential of their region and the specific resources and products to capitalize on.

The remainder of this Chapter is organized as follows: Sect. 2 underlines some of the remarkable studies in the context of health tourism and digital applications; Sect. 3 highlights the process of domain knowledge elicitation and formalization; Sect. 4 discusses the detailed architecture and pipeline of the HTAB tool; and finally, Sect. 5 concludes the main outcomes of the project.

2 Related Work

Health tourism has been around throughout history; however, the research and literature study on this topic has grown extensively over the last few decades [3]. Many cities around Europe have started to invest in their available resources to position themselves as health destinations and attract health-conscious tourists [1]. Study shows that some destinations utilize natural resources such as healing waters and clean air to promote health tourism and the local economy [8, 9]. Natural resources may include the existence of the physical natural features (such as lakes, mountains, waterfalls, and forests), the derived products (like local food and sports classes), and regional cultural heritage [10].

As a result of this growing opportunity in health tourism, the adoption of new technologies such as DSS, semantic knowledge implementation, and digital tools and applications to exploit the local resources are also rising. The implementation of semantic technologies and ontologies for a medical tourism recommender system has been investigated in the work of Frikha et al. [11]. Moreover, a DSS was developed in another study [12] to evaluate the key factors such as climate, tourism development, and attractions to propose a hierarchical structure for rating the destinations.

Exploiting the semantic reasoning and ontologies, Moreno et al. [13] developed a web-based platform to provide personalized recommendations for touristic and leisure activities based on demographics, motivations, user actions, and ratings. In another example of an ontology-based application [14], researchers created a novel health tourism ontology and built a semantic web-based search application for health tourism in Thailand.

This work presents a digital web-based application to provide customized recommendations to destination stakeholders and tourism policy-makers in order to offer insights and suggestions on how to further exploit the health tourism industry by investing in their natural resources. This digital application relies on an ontology-based DSS (described in Chap. 4 [7]), which formalizes the representation of health tourism destinations, their natural resources, services, and activities based on the natural resources.

3 Knowledge Implementation

Prior to the development of the digital tool, adopting a collaborative approach to implement and elicit the domain knowledge is a vital step. As described in the previous Chapter, within the HEALPS2 project, the process of formalizing the information is performed through the collaboration among various stakeholders from six countries to capture their ideas, knowledge, and opinions in the process of knowledge elicitation. Data collected from the research studies, unstructured interviews, and stakeholders’ brainstorming was then transformed into quantifiable Key Performance Indicators (KPIs), which later formed the basic roadmap toward semantic model formalization.

3.1 Defining KPIs

In order to provide a set of tailored recommendations, some general KPIs have been identified to further emphasize the importance of the natural resources of each destination [15]. The existence of one or more natural resources, which are known to have a positive impact on particular health conditions, can lead to a more precise approach toward improving health tourism in a given destination. The KPIs are defined considering the existence of the natural resources and general framework of tourism in every destination in order to receive a set of customized suggestions on the target group, products, and natural resources to exploit. The process of selecting the KPIs in Alpine health tourism is discussed in detail in Chap. 3 [15].

3.2 Semantic Representation

The information gathered from the KPIs is then further formalized in a semantic data model representing the knowledge regarding the tourist destinations, the existence of the natural resources, target groups, and the relationship between all these factors [16]. A semantic approach can support data integration by annotating and enriching the unstructured and semi-structured data coming from the KPIs and online surveys. The semantic model is expressed as a set of ontologies, i.e., “formal, explicit specifications of a shared conceptualization” [17], modeled with Resource Description Framework (RDF) [18] and Web Ontology Language (OWL) [19], while reasoning and data inference are accomplished with the Semantic Web Rule Language (SWRL) [20] and SPARQL Protocol and RDF Query Language (SPARQL) [21].

Together with the SWRL, the ontology is then stored on Stardog [22]—a commercial RDF database with fast SPARQL query, transactions, and OWL reasoning support. Once the ontology is available on the Stardog server, the HTAB digital tool is able to exploit the semantic repository, reason the knowledge domain, and infer new data. The structure of the semantic model and the ontology defined to be used within the HTAB tool is discussed in detail in Chap. 4 [7].

4 The HTAB Tool Pipeline

The HTAB tool is a web-based digital application that triggers a pipeline behind the web platform. The synergy between a web-based application and ontology-based approaches is widely documented in the literature, and some examples can also be traced in the tourism industry [23, 24]. The pipeline receives data, preprocesses and cleans the data, exchanges and translates the data into a semantic representation of the HEALPS2 data model, and produces the output results with recommendations and suggestions for the destination (Fig. 1). The HTAB tool is implemented on the Paracelsus Medical University (PMU) server to run automatically every day and verify if there is any new answer to the online questionnaire on the Lime Survey platform [25]. If a new questionnaire answer is found, the HTAB tool starts the pipeline by gathering the input data, reasoning and querying the semantic knowledge, and finally dispatching the output results to the destination. These steps of the HTAB digital application and its pipeline are discussed in detail in the following subsections.

Fig. 1
A diagram exhibits a sequence behind the web platform from the H E A L P S 2 project website to recommendations and suggestions for the destination.

The pipeline behind the HTAB web-based application to receive and preprocess input data, query and infer reasoned data, and dispatch the output data

Full size image

4.1 Gathering Input Data

The HTAB tool runs based on the input data coming from the online questionnaire, to which a destination must answer in order to receive the recommendations. These input data should then be collected, cleaned, and translated to add value to the DSS to be reasoned and infer the output results.

Online Survey. The online questionnaire is hosted on Lime Survey available in five languages: English, French, German, Italian, and Slovenian. Each Alpine destination can fill the questionnaire directly on the Lime Survey platform and provide their email address to which the output results will be sent later. The survey consists of 25 questions divided into the subtopics as follows:

  • General information on the region such as number of inhabitants, number of tourists, tourists’ countries of origin;

  • Presence and use of some natural health resources such as waterfalls, forests, healing waters;

  • Presence of some health tourism services such as health checks, physiotherapy, bicycle availability; and

  • Cooperation with stakeholders and membership in tourism networks and tourism attractions.

The questionnaire’s answers are retrieved and downloaded once the application is launched through the HTTP POST method of the REST API. This process is implemented and automated via a Python script to fetch the questionnaire’s answers from the Lime Survey platform and then decode and translate the collected data into CSV format. Consequently, the CSV data is easily imported and stored in the database.

Data Preprocessing. Since the data coming from the online questionnaire is not machine-readable and therefore not in the required format to be used in the semantic model, it is necessary to preprocess and encode the data before sending it to the semantic database. Firstly, the data from the questionnaire must be mapped to the associated KPIs defined earlier, considering that each question within the online survey may refer to one or more KPIs. The data downloaded from the Lime Survey and stored in the database are identified with coded column names, while the KPIs are defined with another coding system. Therefore, a mapping table is implemented in Python to match the questions with the database column names and associated KPIs.

Moreover, the given answers to the questionnaire must be decoded according to the expected machine-readable answers. As a result, another mapping table is implemented to define and specify all the real answers and their associated machine-readable codes. Also, some questions may have more complicated textual or imagery answers that must be translated to numerical values.

Input JSON File Generating for Semantic Reasoning. Once the questionnaire’s results are decoded, cleaned, and in a correct machine-readable format, the data regarding the destination must be written in a JSON file to be readable for the semantic model. Therefore, a JSON file is generated via Python, which includes all the necessary data in a format that is consistent with the semantic model and the ontology. The JSON file contains the information about a destination together with the necessary datatype properties and the descriptors defined in the ontology (Fig. 2). All these JSON attributes must be filled in via Python dictionary by extracting the required information from the preprocessed survey answers. The JSON file generated at this stage will be saved on the server to be used by the semantic repository in the next phase.

Fig. 2
A 16-line code of data for details of destination named Valposchiavo, with key numbers included.

An excerpt of the JSON file generated from the data coming from the questionnaire that is served as input to insert the new destination “Valposchiavo” into the semantic model

Full size image

4.2 Reasoning and Querying Inferred Output Data (Java Middleware)

In order to receive the customized recommendation and suggestions regarding each destination and target group, HTAB is capable of connecting to the semantic repository, running the reasoner, and retrieving the inferred data. Moreover, the scalability of the digital tool relies on the possibility of adding new destinations to the system in the future. However, there are some limitations imposed by the Open-World Assumption (OWA) of the monotonic nature of Description Logic (DL). OWL exploits the OWA reasoning technique in which it cannot assume something does not exist unless it is explicitly stated [26]. Therefore, it is impossible for the deductive reasoner to infer the existence of a new instance unless it is already modeled in the semantic knowledge base. Hence, inserting a new piece of information into the semantic repository is not a task supported by DL-based technologies [27, 28]. As a result, the HTAB tool relies on a Java application acting as a middleware [29, 30] between the digital tool and the semantic repository to clear these barriers.

This Java application runs the Stardog server, creates a semantic database, and uploads the ontology files on the server. The Java application also defines the reasoning type needed to be exploited on this particular knowledge base according to the DL rules. At this point, the knowledge base is up and running on the Stardog repository, the SWRL rules are translated and consistent with the Stardog to infer data, and the reasoning logic to infer new pieces of information is set.

The Java application is now ready to receive the information about the new destination produced from the online survey. All the necessary information about a new destination is generated, cleaned, and incorporated into a JSON file (Sect. 4.1). The application fetches the JSON file to extract the information and generates a proper SPARQL query to insert the new destination into the knowledge base. The query must conform according to the data types defined in the ontology regarding each attribute while including all the necessary descriptors characterizing each destination in order not to cause any inconsistency within the knowledge base. The Java middleware utilizes a mapping table to associate each destination’s descriptor to the corresponding KPIs defined within the semantic database and the ontology to keep the data consistency. The application translates the JSON information into a correct INSERT SPARQL query and runs the query against the Stardog query system to append the new destination data into the knowledge base (Fig. 3).

Fig. 3
A 20-line code of data for H E A L P S 2, an excerpt of INSERT S P A R Q L query.

An excerpt of the SPARQL INSERT query to insert the new destination “Valposchiavo” into the semantic model together with its attributes

Full size image

The ontology on the Stardog semantic repository is now updated with the new destination information; hence, the knowledge can be reasoned according to the predefined rules and reasoning techniques. The Java application then generates the proper SELECT SPARQL query to retrieve the inferred data regarding the new destination recommendations. The query then runs on the Stardog server with DL reasoning techniques to infer data based on each target group. In fact, the reasoner is run once for each target group to calculate the correct optimal value for the specific health condition the target group entails. Each target group represents a group of tourists suffering from physical limitations or chronic conditions such as respiratory diseases or allergies in which it is associated with an optimum value to demonstrate how much the health condition described within the target group can be enhanced by a specific destination natural resources and products. The application retrieves the reasoned results from the semantic repository through the reasoning process, which illustrates the optimum values for each target group for the new destination. The process of choosing different target groups and calculating the optimal values is discussed in detail in Chap. 4 [7].

Once the reasoning for every target group is done, the application generates the output JSON files integrating the results (Fig. 4). The results are divided into separate JSON files for each target group, providing reasoned information based on each target group's health conditions and their respective optimal values.

Fig. 4
A diagram exhibits processes in the Online Survey Platform and Java Middleware Application with a semantic repository, connected through J S O N.

Conceptual architecture of the HTAB digital pipeline and the interactions between the semantic middleware with semantic repository and the input/output data

Full size image

4.3 Dispatching the Output Results

Once the inferred and reasoned results are generated in JSON files, the system is ready to calculate the target group ranking and dispatch the final output for the destination. In order to calculate the ranking for the destination, the values related to the KPIs are extracted from the output JSON files and multiplied by the predefined weights associated with KPIs. These weighted values are then aggregated and compared with an optimal value specific for each target group. A ranking among target groups is then created, sorting their final scores in descending order.

The final output is a report, provided as a PDF file generated dynamically via Python, which consists of an introduction, medical evidence, recommendations, natural resources and services, key figures on the importance of tourism, cooperation and networks, and additional illustrative figures. The PDF file is the final output of the HTAB pipeline that provides necessary information and insights for the destination to improve their tourism resources, products, and policies to better exploit the health tourism within their region.

This PDF report is automatically sent to the email address the destination provided on the Lime survey to fill the questionnaire.

5 Conclusion

This Chapter presented the architecture of the HTAB digital tool to support the Alpine destinations in further exploiting their natural resources in health tourism. Starting from literature to understand the impacts of the Alpine natural resources on tourists’ physical and mental health [31], the project defined a comprehensive list of success factors through collaboration among various stakeholders. This Chapter describes how the HTAB digital tool is developed to identify the destination’s tourism characteristics and utilize this data in order to provide tailored and customized suggestions to the destination to exploit the NHT industry of the destination. The digital application retrieves the data about each Alpine destination through an online survey; cleans, preprocesses, and encodes the data; reasons and queries the data against the domain knowledge base; and finally provides a set of customized suggestions and recommendations to the destination manager.