Open Access

Ontology construction and application in practice case study of health tourism in Thailand

SpringerPlus20165:2106

https://doi.org/10.1186/s40064-016-3747-3

Received: 27 August 2015

Accepted: 25 November 2016

Published: 20 December 2016

Abstract

Ontology is one of the key components in semantic webs. It contains the core knowledge for an effective search. However, building ontology requires the carefully-collected knowledge which is very domain-sensitive. In this work, we present the practice of ontology construction for a case study of health tourism in Thailand. The whole process follows the METHONTOLOGY approach, which consists of phases: information gathering, corpus study, ontology engineering, evaluation, publishing, and the application construction. Different sources of data such as structure web documents like HTML and other documents are acquired in the information gathering process. The tourism corpora from various tourism texts and standards are explored. The ontology is evaluated in two aspects: automatic reasoning using Pellet, and RacerPro, and the questionnaires, used to evaluate by experts of the domains: tourism domain experts and ontology experts. The ontology usability is demonstrated via the semantic web application and via example axioms. The developed ontology is actually the first health tourism ontology in Thailand with the published application.

Keywords

Ontology engineeringHealth tourismHua-Hin tourismKnowledge management

Introduction

Ontology is an important element underlying the semantic web technology. It contains the domain-specific knowledge to increase the powerfulness of a particular search engine. The ontology design is very important since it affects the search effectiveness and efficiency. With the good property characteristic design, the inference engine can infer new knowledge from existing ones, returning more related results.

Designing ontology and evaluating it are challenging tasks. Ontology contains several concepts in the object-oriented style. The relationships are the properties which map from a domain to a range and can be viewed as functions or relations. The property characteristics are needed to increase the possibility of the knowledge inference.

Designing particular ontology requires domain knowledge. Moreover, the designer must know about the object concept, the hierarchy of subclasses, the property mapping, and property characteristics. These are required for the inference engine. After designing and inputting instances or individuals, it is questionable whether the design is complete according to the specification.

Tourism is one of the interesting application domains since tourism industry can attract tourists to a country or region which can increase the local or domestic income. With a good information system and Internet infrastructure, the search can facilitate tourists to find the right travel information and accommodation.

Semantic web has been applied in many applications including tourism. It has been used in Morocco, Hong Kong, China, Germany, and etc. Though there are many existing tourism ontologies, each of them has different focuses. In this research, we are interested in developing a semantic web for health tourism in Thailand. The prototyped ontology is based on the health tourism in Hua Hin district. We discuss the design process and experiences focusing on the health tourism ontology engineering, starting from information gathering, ontology conceptualization and evaluation, until the application deployment.

Our development has two unique characteristics.
  1. 1.

    Current ontology around is about a general tourism concept which focuses on attractions, hotels, and etc. The ontology example on health tourism, to our knowledge, is not available.

     
  2. 2.

    This work is a pioneer work in gathering extensive Hua Hin health tourism information, and classifying it using ontology. We focus on experiences in gathering such information in practice and building the ontology as well as the prototype application.

     

The next section presents a brief background on ontology, Resource Description Framework (RDF), and Web Ontology Language (OWL). Also the definition of the health tourism domain is presented as well as related studies. Third section presents the overall methodology. Fourth section presents the ontology design and the evaluation process. Fifth section demonstrates the sample application and discussion in sixth section. Conclusion is presented in final section.

Background

This section presents some background related to the research. It includes backgrounds in ontology, RDF, and OWL. Next, we present the definition of health tourism, and its context of Thailand. At last, we discuss the literature area in the field.

Ontology, RDF, OWL

Tim Berners-Lee presented the future web concepts and published them in Scientific American 2001 (Berners-Lee et al. 2001) known as “Semantic Web”. The purpose of this concept is to enable machines to comprehend semantic documents and data that are enriched by the convention. Three components are identified: ontologies, knowledge representation and agent as essential to function. The semantic web development has been done in many domains such as tourism, languages, organization.

The core standard is RDF for knowledge representation (Auer et al. 2007). RDF is a data model consisting of a triple (i.e. subject, predicate, object), containing information about web resources. OWL used along with RDF is an ontology language for a semantic web with formally-defined meaning.

Since OWL became a W3C standard (Bechhofer et al. 2004), there has been a notable increment in the number of ontologies. In OWL1 species, there are three official sublanguages: OWL Full, OWL DL and OWL Lite. The study of OWL DL has become an important aspect of ontology representation for any inference problem in the semantic web. The OWL DL is essentially the description logic (DL), SHOIN(D). It offers a high level of expressivity. For example, it provides full negation, disjunction, and a restricted form of universal and existential quantification of variables (Motik et al. 2005; Beek and Horrocks 2005).

OWL DL language constructs can be represented using classes, roles, individuals, class membership and role instances, owl:Thing and owl:Nothing, class equivalence, class conclusion, negation of classes, disjointness, conjunction, disjunction, role restriction using owl:allValuesFrom and owl:someValuesFrom, rdfs:domain and rdfs:range.

Health tourism

According to the World Health Organization (WHO), health is a state of complete physical, mental and social well-being and not merely absence of disease. Health tourism is the used of medical services away from home or the travel for the purpose of healing, obtaining medical services of health improvement, and etc.

Health tourists are those who travel for the above purposes. In Thailand, health tourism is divided in two kinds based on purposes.
  1. 1.

    Health healing: it is the travel whose purpose is to restore health, and cure diseases, including cosmetic surgery, and dental services. This is called medical tourism.

     
  2. 2.

    Health promotion: it is the travel whose purpose is to increase health strength in the tourist attraction area. Examples are spa, aroma therapy, Thai herb sauna or stream, and body massage. Supplementary services may be body detox, meditation, yoga, and nutrition consulting. This is also called wellness tourism. In some texts, it is divided into spa and wellness tourism.

     

The spa business is a major business in Thailand’s wellness tourism which is included in the term “health business” according to the Act of legislation of Public Health Ministry (B.E. 2509).

The business provides a place for health or beauty maintenance. The health business is categorized as:
  1. 1.

    Spa business for health. It is a type of business that uses water and massage for treatment. There can also be facilities such as nutrition advice, herb sauna and stream, meditation, yoga and alternative medicines.

     
  2. 2.

    Massage business for health. The business focuses on massaging for health, and for relaxing according to massage science. It does not provide a bathing area.

     
  3. 3.

    Massage business for beauty. Examples of this type of business are hair salons or beauty salons. It contains different massage type according to the massage science, which can increase personal beauty. However, it does not provide a bathing area.

     

Hua Hin is one of the popular districts in Thailand that is close to Bangkok. It comprises many interesting attractions especially beautiful beaches. There are many local and international tourists. There are many famous resorts and spas as well as destination spas. It is expected to be the next health tourism area in addition to Bangkok.

In Hua Hin, there are lots of medical tourism and wellness or spa tourism businesses. For medical tourism, there are also many famous hospitals and clinics, dental clinics, beauty clinics, pharmacy, and etc. It is one of the destinations for wellness spas as promoted by the Thai Government with the policy “Medical Hub of Asia” (Thai Board of Investment 2012). In Hua Hin, there are four kinds of wellness spas:
  1. 1.

    Hotel and resort spa It is a kind of the spa situated in a hotel. The main business is the hotel and resort while the spa is a facility of the hotel.

     
  2. 2.

    Destination spa It is the spa which provides packages for tourists who intend to take a spa course. The tourists must attend a course which may require 2 or more night stays at the resort. A very famous one in Thailand is Chiva-Som (http://www.chivasom.com) which is located in Hua Hin.

     
  3. 3.

    Day spa It is a kind of spa which provides various services. The treatment or service can be applied and there is no need for an overnight stay. On the contrary, the hotel and resort spa can be a day spa since an outside hotel guest can take a visit.

     
  4. 4.

    Medical spa It is the facility that lies between the medical clinic and the day spa. It must be operated under a qualified medical team in many related fields including cosmetic surgery, nutrition, and etc.

     

Related work

First, we discuss the work in information or knowledge engineering in tourism. Secondly, the methods to evaluate the ontology are studied, and then the applications of tourism that use the ontology are discussed.

Ontology extraction in tourism

A lot of work studied ontology extraction approaches. They differed in the application domain, the techniques, the target documents, automatic or semi-automatic approach.

Meersman et al. (2009) presented a way to write the ontology documents (Meersman et al. 2009; Ruiz-Martínez et al. 2011). They focused on how ontology requirement was gathered and the ontology specification was written formally. The key activities were the search and reuse of existing knowledge resources and ontological resources. Also, the verification and validation were considered.

Karoui et al. (2004) proposed the automatic method for the ontology discovery for the tourism application. They proposed the ontology building process from HTML documents which is a complement to the Aussenac-Gilles approach. It used similarity and clustering techniques to group words to define the hierarchy. The processes were corpus preparation, Aussenac-Gilles methodology for ontology building, discovery approach based on clustering considering the HTML structure.

Ogata (2001) described a framework to construct the formal ontology based on web documents. The method was based on logics, and web technology, XML, and NLP. Mouhim et al. (2011) presented the knowledge management approach based on ontology. They used Morocco tourism ontology. The approach considered Mondeca tourism ontology, OnTour ontology (Siorpaes et al. 2004), etc. Then, the vocabulary was constructed from thesaurus by the United Nation World Tourism Organisation (UNWTO). The category was established and social platforms were examined. Next, the ontology was built using the tool and verified.

Tang and Cai (2010) presented the domain ontology construction from unstructured texts. The approach started from pre-processing of the text to extract keywords or preserved terms and composed simple and compound statements. The descriptive logic (DL) was used to represent the knowledge. At last, the ontology was generated.

Sigala et al. (2007) presented an approach for the creation of the e-tourism domain (Sigala et al. 2007). The process contained four steps: NLP and corpus processing, named entity recognition, ontology population, and consistency checking stages. The first stage used POS Tagger, and syntactic parser, while the second stage used Gazetter and Transducer. The last stage used OWL2 reasoner.

Alani et al. (2003) proposed an automatic extraction of knowledge from the web documents. The domain of impressionist artists and their painting was selected as a prototype for the process. The knowledge extraction was done based on the HTML structure, pattern rule extraction, or machine learning. The corpora such as WordNet and lexical database, and GATE were used to extract name entities and relationships.

Daramola, Adigun, and Avo built ontology for a tourism recommendation application (Daramola et al. 2009). The developed ontologies were Destination Context Ontology and Accommodation Ontology. The usability evaluation was collected from 15 users of the recommendation service.

Gouveia and Cardoso (2007) presented an integration of the tourism information in ontology. The architecture consisted of five layers: semantic layer, mapping layer, syntactic layer, and external data sources. The tourism ontology was at the semantic layer.

Our ontology is different from all the ontologies above. We combine a spa tourism concept and a general tourism concept. Moreover, our concept drills down to the detail of each activity of spa tourism that is appropriate in Thailand.

Ontology evaluation

Supekar (2005) presented the peer-review approach for reviewing the ontology. The approach provided the qualitative ratings of the ontology content. It is the qualitative research that evaluated the content of an organizational ontology, developed within a large Brazilian energy utility company. The evaluation process consisted of a set of questionnaires, based on a multi-disciplinary approach, of a prototype system. The methodology contained four phases: determining the research tools, collecting data for the design of the ontology, building the ontology, and evaluating the ontology.

Almeida (2009) proposed a way to evaluate ontology using questionnaires. The questionnaires contained three kinds of questions: competency questions, information quality, and educational objectives.

Mugellini et al. (2011) presented the quality assurance framework for ontology construction. The approach was based on Hozo reasoner for consistency verification. The prototype ontology was a sustainability science and clinical ontology. For the content, they evaluated the relationships among concepts using concept maps.

Tankeleviciene and Damasevicius (2009) described the characteristics of domain ontology. They proposed two methods for evaluation: (1) a method for an expert-based evaluation of the ontology content, (2) a model and a collection of technical metrics to evaluate using the structural complexity of ontology. Their application was web-based learning. The complexity was analyzed using 7DO models. The expert evaluation was based on completeness, consistency, conciseness, preciseness and clarity.

Lehmann et al. (2011) presented the evaluation of class expression. They created OntoWiki plugin for DL-Learner functionality and tested ontologies using DL expressivity.

Kehagias et al. (2008) presented a method for evaluating the ontology. The conditions used to check the completeness had the following criteria: concept and property hierarchy, module subtraction, documentation and visualization, definition of ranges for property values, disjointness restrictions and adherence to naming conventions.

Five basic internal layers were considered for evaluation: lexical/vocabulary layer, structural/architectural layer, representational/semantic layer, data/application layer, and philosophical layer. Basic external dimensions were (1) user dependence: how many users depend on the ontology? (i.e., what is the impact of changes to the ontology? should this be avoided or is it simple to implement?) (2) is the ontology used as a medium of information exchange across distinct communities? (3) is it documented? If so, in which form? (natural language, UML, logical spec, and etc.) (4) is it a national or international standard? (5) Is it a de facto working standard for some community? (6) usability layer. The approach was applied to ASK-IT (TourismAndLeisureOntology).

According to the above study, we apply two kinds of evaluation: the evaluation by automated reasoning and the evaluation by the experts. For the first type, Pellet and RacerPro reasoners which are plug-into Protégé ontology editor are used. The evaluation focuses on inferring the concept from the created ontology assertion. For the second type, the proposed ontology is evaluated by experts’ questionnaires where the focus of the questions is completeness and documentation.

Related work in semantic web and tourism semantic web

Currently, several tourism projects are based on ontology. Harmonized project (aka Harmonize project) (Dell’Erba et al. 2002, 2005; Foder and Werther 2005) is one of the semantic platform which provides a shared ontology and facilitates the semantic cooperation between the tourism business sectors in European countries.

SATINE (Dogac et al. 2004) is a the famous framework which extends Global Distribution System (GDS) connecting Online Travel Agent (OTA) between semantic webs for distributed web service platforms. Each tourism service needs the registry to be included for the automatic search.

Bottari (Balduini et al. 2012) combined the social network such as Twitter to the ontology which can recommend about the rated restaurants for a given time period. Jakkilinki, Sigala et al. (2007) developed an application for a tour planner with an intelligent approach using the designed ontology. Cardoso (2006) proposed to generate the semantic web process dynamically. It uses web services to gather information. Many services were composed. E-tourism is used as an application for the approach.

EIFFEL (2006) by Mondeca is a tourism search engine developed by French National Software Technology (Mondeca 2006). The goal is to promote regional tourism. It uses semantic-oriented widgets to construct a semantic web portal.

Developing health tourism ontology may be related to many existing ones. For example, in Table 2 is a sample list of ontology we have surveyed. Most of them are general tourism concepts with different focuses. Column “DL expressivity” shows the reasoning of each ontology.

Consider existing health tourism web sites in Thailand, such as: http://www.mymedholiday.com/country/thailand/hua-hin, http://www.thailandmedtourism.com/DestinationDetail/99/2579/Hua-Hin, http://www.thaimedtour.com/content/163/Hua-Hin-Hospital-/, http://www.health-tourism.com/, etc.

The first three are local sites while the last one is not. The first two sites list famous hotels with branded spas, well-known hospitals and clinics. They have a good information presentation but the variety of data is limited. The third one presents only information about hospitals and clinics. It contains hospitals and clinics all over the world. Only paid registered businesses are presented. Thus, our work is the pioneer work to gather health tourism information and build the ontology for Thailand. We aim to publish these data as open linked data in the future.

Methods

The two popular approaches in building ontology are METHONTOLOGY (Fernández-López et al. 1999; Gómez-Pérez 1996), and NeOn (Suárez-Figueroa et al. 2008). The methodology of METHONTOLOGY is more appropriate to us since it focuses on a method to build ontology from scratch, and partly reuses other ontologies while NeOn presents nine scenarios for building the ontology network where most of them consider the scenarios for ontology reuse.

Recall that METHONTOLOGY consists of five steps: specification, knowledge acquisition, conceptualization, integration, and implementation. In Fig. 1, these phases are located. For part (1) in the figure, knowledge acquisition, two kinds of information gathering are used: the semi-automatic approach and the manual approach. The semi-automatic approach is to explore the existing tourism data on the Internet. Typical websites for Hua-Hin tourism are gathered and the information provided by each site is studied. We build a parallel program to extract information from the websites. The extracted information is the spa information provided by hotels and resorts. For the specific details of the day spas, the medical places, the destination spa, and the spa packages, we need a field trip to collect the information and some personnel to search particularly at their official websites. Also, the information about the spa business registered at the Ministry of Public Health is requested. These are the information in hardcopies and in Excel files.
Fig. 1

Overview process of ontology building

Part (2) is relevant to specification, where a corpus of health tourism definition and the related terms, categories including natural products, such as Mueller and Kaufmann (2001), Caballero-Danell and Mugomba (2007), ISPA (Cunningham et al. 2014; Põld; Constantinides 2011; Smith and Puczko 2009) etc. are studied. Several existing tourism ontology or related ontology like Dell’Erba et al. (2002, 2005), Foder and Werther (2005), Knublauch, Siorpaes et al. (2004), Sigala et al. (2007), Ou (2008) are studied. The goal is to extract name entities to derive classes, subclasses, and individuals for the ontology design. The classification about health tourism business is also reviewed. In part (3), we start to extract major attributes or keywords from the previous collected data for conceptualization of ontology. Some grouping is created and synonyms are gathered. The property may be grouped into subclasses such as the product of spa, and spa area.

Lastly, in part (4), the ontology is evaluated by both experts and reasoners. After that, in part (5), the documents of ontology are prepared for publishing.

Ontology conceptualization

Previous tourism ontology in the literature has different focuses. Some focuses on accommodation while others focus on cultural objects, packages, and etc. Also, the tourism in each country has its own unique characteristics.

In particular, there are some special tourism attractions and lifestyle (Mili et al. 2011) for some regional area. There are several kinds of tourism attractions in Thailand. The cultural styles embed traditional Thai lifestyles in the past. For example, the housing and decoration styles are preserved as the traditional Thai housing style. This is shown as the architecture style of the building. Thai massage is also a unique massage type. It includes compress massage, reflexology massage, Ayurveda massage, and Chaleoysak massage.

From the information gathering approach described in “Appendix 1”, we start from designing classes and hierarchy which are subclasses (Gouveia and Cardoso 2007). Also, relations must be described, to define the interaction among classes or properties.

Equivalent classes may be defined. It implies that both classes must contain the same set of individuals. Disjoint class can imply that a member of one class cannot be a member of the other class at the same time. Then, the complex class may be given from connectives such as complement, intersection, and union. A class can be defined as an enumeration of individuals.

A property maps from a domain of individuals (instances of a class) to a range of individuals. The domain and range may be from the same class. The property characteristics may be defined later. For example, it may be an equivalent property where “engage” and “play” mean the same thing in sentences such as “A person plays sports.” or “A person engages in sports”.

Inverse property and functional property are taken into consideration. Functional property means the property has no more than one output value. For example, a spa shop has one unique ID. Similarly, it is an inverse functional property. Thus, we can infer that two spa shops cannot have the same ID. The other characteristics are symmetric property. Transitive property creates the inferences between two properties. For instance, Chiva-Som is in Hua Hin and Hua in is in Prachuap Kirikun province. Thus, Chiva-Som is in Prachuap Kirikun province.

The property restrictions are also defined. The restrictions specify the condition of instances of the class. For example, it is the universal quantification (forall), existential quantification (for some), or specific values as well as cardinality restriction, where we can define max, min, and the number of exact individuals (in Table 1) (Cardoso 2006). The design guideline adopted is described in “Appendix 2”.
Table 1

List of ontology studied

Ontology name

Owner

DL Expressivity

travel.owl (Knublauch)

HolgerKnublauch

SOIN(D)

Etp-tourism (Mili et al. 2011)

PetkoValtchev

SHOIN(D)

e-tourism.owl (Siorpaes et al. 2004)

DERI (STI Innsbruck 2009)

ALCHIF(D)

Qall-me (Ou 2008)

STREP project (Oct 2006–Sep 2009) by FBK [Trento, Italy]

SHOIN(D)

Travel guides (Cunningham et al. 2014)

Travel Guides ontology

SHIF(D)

Protégé 4.3 is used for the ontology design and Hermit 1.3.7 and Pellet 2.3.1 reasoners are applied to check reasoning as we will describe in the next section.

Figure 2 presents the overview of the ontology. It contains only important concepts. All classes are displayed in “Appendix 3”. Our main concept “TourismSite” is as highlighted in the center circle. Full documents are available at http://health-tourism.cpe.ku.ac.th/huahinonto/index.html.
Fig. 2

Main concepts

In Fig. 2, the solid circles represent class and subclasses while the dashed circles show external concepts. Solid-headed arrows show property relation and lighter-head arrows show an is-a relationship (subclasses). The dashed line shows an example of equivalent classes.

Individuals of “TourismSite” are spa shops, Hua Hin, hotel name, and etc. Wellness spa is one class of TourismSite. WellnessSpa is divided into 4 subclasses: hotel and resort spa, day spa, medical spa, and destination spa. The spa shop may also be registered to the Ministry of Public Health. The registration can be a new application for the first year and extended status for the following year. Various kinds of concepts related to spas are shown such as services, facility, treatment, packages, and location. There are subclasses of services such as massage, and beauty service, and subclasses of facilities such as free services, room, and dry heat facility. For example, a massage can be divided into many types: eastern massage, and western massage where Thai massage is in the category of eastern massage. Beauty massage is such as slimming massage, firming massage, and collagen massage.

External concepts can also be linked such as linking GPS coordinate to geonames (http://www.geonames.org/ontology/documentation.html), Hua Hin individual to Hua Hin District in dbpedia (http://dbpedia.org/page/Hua_Hin_District), or HotelandResortSpa to an accommodation concept (http://ontologies.sti-innsbruck.at/acco/ns.html).

Consider the dashed box. This part presents the class of spa businesses based on the Act as in “Background” section which has three types: health spa, massage for health, and massage for beauty. Health spa is the focus where we will infer to equivalent class as HydrotherapySite. We can put axioms to define the valid health spa by considering individual services and facilities. Then valid health spas inferred will be a member of HydrotherapySite. Similarly, we can do the same thing to define a valid massage for beauty business or massage for health business. Note that one business can belong to more than one kind.

The above ontology is first evaluated by reasoners, namely Pellet and RacerPro, to check the correctness of the inferred results from defined axioms, and the usefulness of characteristics. Table 2 presents the measurement from various aspects. Metric row shows the summary of number of axioms, classes, properties and individuals. The DL expressivity is SRIOQ (Dentler et al. 2011).
Table 2

Ontology metrics

 

DL expressivity

SRIOQ(D)

Overall metrics

Axiom

1799

Logical axiom count

935

Class count

80

Object property count

92

Data property count

16

Individual count

250

Class axioms

SubClassOf axioms count

55

EquivalentClasses axioms count

19

DisjointClasses axioms count

3

CGI count

0

Hidden CGI count

19

Object property axioms

SubObjectPropertyOf axioms count

37

EquivalentObjectProperties axioms count

3

InverseObjectProperties axioms count

33

DisjointObjectProperties axioms count

2

FunctionalObjectProperty axioms count

10

InverseFunctionalObjectProperty axioms count

1

TransitiveFunctionalObjectProperty axioms count

2

SymmestricObjectProperty axioms count

1

AsymmestricObjectProperty axioms count

2

ReflextiveObjectProperty axioms count

1

IrreflextiveObjectProperty axioms count

3

ObjectPropertyDomain axioms count

82

ObjectPropertyRange axioms count

82

SubPropertyChainOf axioms count

2

Data property axioms

SubDataPropertyOf axioms count

7

EquivalentDataProperties axioms count

1

DisjointDataProperties axioms count

1

FunctionalDataProperty axioms count

3

DataPropertyDomain axioms count

14

DataPropertyRange axioms count

12

Individual axioms

ClassAssertion axioms count

274

ObjectPropertyAssertion axioms count

196

DataPropertyAssertion axioms count

64

NegativeObjectPropertyAssertion axioms count

1

NegativeDataPropertyAssertion axioms count

0

SameIndividual axioms count

24

DifferentIndividuals axioms count

0

Annotation axioms

AnnotationAssertion axioms count

417

AnnotationPropertyDomain axioms count

4

AnnotationPropertyRangeOf axioms count

4

We ran Pellet, and RacerPro to verify it which took 1671, and 17,630 ms respectively. The results contain approximately 32, 354 unsatisfiability class inference for Pellet and RacerPro respectively and there are 1, and 84 unsatisfiability object property inferences.

The domain concept is also verified by the domain experts who evaluated based on the completeness and accuracy of the terms. At last, the structure of ontology is commented by the ontology experts.

The questionnaires for the ontology experts are based on the above internal layers and external dimensions in the following section as well as the ontology dimension. We build ontology documents for experts to read. The documents are categorized in two kinds. First, it is the concept of the ontology which we hand to the domain knowledge expert to evaluate the completeness, and correctness of the terminology. Secondly, we construct the technical ontology document for the ontology expert to verify the architecture and structure of the ontology. The technical document is in the HTML form located at the site http://health-tourism.cpe.ku.ac.th/huahinonto/index.html.

For the domain experts, the following aspects are demanded: the completeness of each class, and the correctness in terms of naming, category, and description. The score is given in five levels [1, …, 5], where value 5 is the most proper value. The domain experts agreed to our domain concepts. The average score of the main classes from the two experts is 4.6. The overall correctness of the terminology and description is 4.59.

From the ontology experts, the comments are all satisfactory. One expert agreed that the comment can be used via “rdfs:comment” but there are other choices of putting annotation such as using multi-language using “rdfs:lang”. To improve the usability, the comments can be translated in many languages. The depth of classes is about 2–3 levels from roots with is-a hierarchy. One good comment is the difference between “equivalent class” and “synonym”. In some cases, only a synonym is sufficient, such as “Sauna” and “Dry Heat”. It is better to put as two labels rather than two classes. Also, the similar problem is with “sameAs” individual. Attention should be paid to distinguishing between individuals and classes. There should not be lots of equivalent classes; the use of labels may be better. Besides, the object properties/characteristics are comprehensive. The ontology is a good starting point for the standard depending on its usability in the future. Figure 3 shows the ontology dimension score.
Fig. 3

Ontology dimension score

The ontology data set is linked to Link Open Data Cloud (Cyganiak and Jentzsch 2011; HPI 2011). We need to adjust several points for this application such as all the URIs in the data set must be resolved. Linked Open Data Cloud must be in RDF data format such as (RDFa, RDF/XML, Turtle, N-Triples). The data set should contain at least 1000 triples and must be connected via RDF links that are already in the diagram. The data set may be accessed using SPARQL endpoint, RDF crawling, or RDF dump. After the data set meets these criteria, it is added to the Data Hub located at http://datahub.io/organization/https-www-facebook-com-healthtourismmanagement as shown in “Appendix 4”.

Hua-Hin health tourism application

The web application that exploits the health tourism ontology is developed. For the user interface, we gather requirements from stakeholders: administrative personnel who maintain spa shops’ web site, spa business managers who provide us their spa services, menus, and tourists.

The example of a web site is in Fig. 4 where a user is asked to search for Health Spa.
Fig. 4

Example of user interface

Let us consider the case of knowledge inference. According to the Act of Public Health Ministry mentioned in “Background” section, for the main concept, TourismSite, there are three types according to the Act: HealthSpa, MassageForBeauty, and MassageForHealth.

A health spa must have the use of water treatment, or advanced water therapy. In particular, the water treatment can be pool, tub, jacuzzi relaxation, bath, or foot bath. The advanced water therapy can be such as water exercise or aquatic exercise whose individuals can be Watsu, Aichi, and water relaxation. A health spa must contain at least one bath service and one pressure shower whose individuals are such as Effusion shower, Swiss shower, Vichy shower, Jet Blitz, and Experience shower.

The following axiom (1) describes the concept of health spa. Note that the property “hasRegisteredTerm” implies that the spa shop must be registered to the Ministry of Public Health. The registration can be the new registration or extended application for subsequent years.
$$\begin{aligned} & \text{`}\text{`}{\text{TourismSite}} \\ & {\text{and}}\;(({\text{hasAdvancedWaterTherapy}}\,\hbox{min} \,1\,{\text{AdvancedWaterTherapy}}) \\ & {\text{or}}\;({\text{hasWaterTreatment}}\,\hbox{min} \,1\,{\text{WaterTreatment}})) \\ & {\text{and}}\;({\text{hasBathType}}\;\hbox{min} \,1\;{\text{Bath}}) \\ & {\text{and}}\;({\text{hasPressureShower}}\;\hbox{min} \,1\;{\text{PressureShower}}) \\ & {\text{and}}\;\left({\text{hasRegisteredTerm}}\,\hbox{min} \,1\,\left(\left\{ {{\text{Extended}},{\text{NewApplication}}} \right\}\right)\right) \text{''}\\ \end{aligned}$$
(1)
Figure 5 shows the related classes, subclasses, and properties used by axiom (1). The dashed arrows are subproperties from properties. For example, hasBathType is a subproperty of hasFacility since Bath is a subclass of Facility.
Fig. 5

Related classes and properties for axiom (1)

For a beauty massage shop, the spa’s purpose is to massage for beauty improvement. The shop must provide beauty massage service including individuals such as firming massage, detox massage, fat massage, and collagen massage and the business shop has beauty services which are classified as nail services, hair services, and eye services. The equivalent class axiom is defined as axiom (2). Figure 6 displays the classes, subclasses, and properties for this axiom.
$$\begin{aligned} & \text{`}\text{`}{\text{TourismSite}} \\ & {\text{and}}\;({\text{hasBeautyMassage}}\;\hbox{min} \,1\;{\text{BeautyMassage}}) \\ & {\text{and}}\;({\text{hasBeautyService}}\;\hbox{min} \,1\;{\text{BeautyService}}) \\ & {\text{and}}\;\left({\text{hasRegisteredTerm}}\;\hbox{min} \,1\;\left(\left\{ {{\text{Extended}},{\text{NewApplication}}} \right\}\right)\right) \text{''}\\ \end{aligned}$$
(2)
Fig. 6

Related classes and properties for axiom (2)

We define HydrotherapySite which is equivalent class to HealthSpa as in the dashed box in Fig. 2, which can be inferred by axiom (1) and, thus, it is a subclass of TypeOfBusinessFromAct. Therefore, HydrotherapySite is also a subclass of WellnessSpa. The inferred individuals can be derived accordingly.

Consider Chiva-Som as an individual. Chiva-Som is the kind of desitnation spas in Hua Hin. It is registered as extended application, and property assertion and data properties are shown in Fig. 7. Thus, evaluating axiom (1) can result in “Chiva-Som is a HydrotherapySite”.
Fig. 7

Chiva-Som properties and axiom evaluation

In another example, THE BARAI individual and property assertions are similarly declared in Fig. 8. THE BARAI is a kind of hotel and resort spa. Figure 9 shows the example where properties are inferred from subproperties.
Fig. 8

THE BARAI property inferred

Fig. 9

isLocationOf and similar property

Figure 9 shows an individual, the Hyatt Regency Resort and Spa. To demonstrate the transitivity, consider the isLocationOf property. THE BARAI is a TourismSite located at the Hyatt Regency Resort and Spa in Hua Hin. AnantaraSpa is another spa in Hua Hin and Hyatt Regency is Hua Hin; therefore, is the location of Hua Hin, where isLocationOf has transitivity. The implication is “THE BARAI is located in this hotel, therefore, the address of THE BARAI is inferred from “LocationAddress” of this hotel by the object property “isLocationOf”. Similarly on the right side, Hua Hin is the location of Chiva-Som, Anatara Resort and Spa, Hyatt Regency, and etc. Consequently, Hua Hin is the location of THE BARAI, Anantara Spa, and etc.

Similar property has reflexivity; thus, we derive HyattRegency is similar to HyattRegency and Hua Hin is similar to Hua Hin.

Lesson learned and discussion

Although the way we gather the information and construct the ontology is conventional, we believe that building semantic web application in each domain knowledge has specific characteristics and challenges. We have gained many interesting experiences regarding to this ontology construction, and the application.
  • The study of domain knowledge is important. Without the correct knowledge, the information cannot be classified correctly in the first place.

    With the information collected, we learned that the health tourism concept in Thailand is very vague. Most people do not know the right meaning of spa. Medical tourism and health tourism are interchangeably used in many occasions.

    Most business shops are registered with the Ministry of Commerce. Only a few in the area are registered with the Ministry of Public Health as the business for Health (health spa, massage for beauty, and massages for health). The business may use the word “spa” in the title but it is not one of the three kinds.

  • The ontology designed should be optimal. Individuals should not just be inserted in there and classified as subclasses. Object properties and characteristics are important to expand the knowledge by inference. The ontology should contain axioms, and equivalent classes to exhibit the rules for generating knowledge.

  • The ontology design must have a goal. The axioms for classes can be used properly. For example, we may use the axiom to validate whether or not a business shop is a valid hydrotherapy site from their given services.

  • Information gathering is a very important phase. If we cannot collect information properly, thoroughly, we cannot design the ontology to cover all of them.

  • There is always room for improvement. Once we have more new data, the ontology may need to be revised. We can always find the wrong axiom or wrong inference results, due to wrong object properties/characteristics.

  • Data cleansing is a huge task and it is never-ending. The naming is a major problem in every area. Even though with the standard dictionary, a business shop itself may use wrong words due to insufficient knowledge. Also, the local language is another problem. Thus, it is the problem of different ways of calling, spelling, and language translation.

  • Besides, there are always new websites, web portals and new updates. The information gathering and maintenance is, therefore, the continuing process.

Comparison of our ontology to other tourism ontologies in Table 3 is presented in Table 2. Of all, tgproton, or PROTON (PROTo ONtology) contains the highest number of individuals. It covers upper-level ontology providing coverage of general concepts necessary for a wide range of tasks. The traveling domain can use http://goodoldai.org.yu/ns/tgproton.owl# prefix to gather the resources as proton.owl, travel_wkb.owl, and upproton.owl. The structure of these classes in owl file is about the traveler, attraction, destination, and user profile. Our ontology characteristics are shown in column “HT”. We start our design from the first version by adding more properties characteristics and axiom as well as individuals. The current version is evolved from travel.owl which is also a general tourism concept. Our HT can be integrated with existing general tourism ontology such as dbpedia (Hua Hin) i.e.:or the latitude or longitude can be mapped to geoname, i.e.,Table 4 summarizes the reuse of standard prefixes: foaf, owl, rdf, rdfs, xsd in our ontology.
Table 3

Metric comparison of our ontology and other tourism ontologies

Metrics

E-tourism

Tgproton

Travel

ETP-tourism

HT

DL expressivity

ALCHIF(D)

SHIF(D)

SHOIN(D)

SHOIN(D)

SROIQ(D)

Total axiom counts

1535

12,306

145

907

1799

Total logical

1320

8135

93

533

935

Total classes

20

393

35

194

80

Total object properties

12

116

6

41

92

Total data properties

78

44

4

46

16

Total individuals

172

2855

14

19

250

Table 4

Ontology reuse prefixes

Table 5 shows the features’ comparison between the semantic web developed in this paper in the last column and existing systems of tourism semantic web. Our system provides a semantic web search engine concerning Hua Hin district and spa tourism domain. We used web crawler to gather information from various Hua Hin web portals. However, our system does not serve a commercial purpose like SATINE, and has not provided web services yet (which is an ongoing work), and Augmented Reality (AR) library has not support.
Table 5

The features comparison between the presented system in this paper and existing systems of tourism semantic web

Features

EIFFEL

Harmonize

SATINE

Bottari

Jakkilinki, Sigala et al.

Cardoso (2006)

Our work

Semantic search engine

   

Semantic web

Find relationship among the website

       

Indexing

       

Sorting

      

Ontology base

 

    

Business

  

    

Web services

  

  

 

Crawling from social networks–website to ontology

   

  

Augmented reality (AR)

   

   

Tour planner

   

  

Conclusions

We present the ontology construction experience of Hua Hin Health Tourism, Thailand. The methodology follows the standard approach which starts from gathering Hua Hin heath tourism information. Two gathering approaches are used. The first one is an automatic information extraction from HTML documents which focus on the agencies like Agoda, TripAdvisor, HeyHuaHin and AtSiam. These sites contain mostly hotel information with spa facilities. The information from website is usually not sufficient. For other kinds of shops such as day spa shops, hospital, and clinics without official websites or not hosted at a large agency like Agoda or TripAdvisor, a field trip is also necessary. These shops may have only Facebook or FourSquare pages. Data from both gathering approaches are cleaned up, unified in many aspects, and merged automatically.

The study of corpus of health tourism is also regarded as a specification phase. Keywords, categories, and local laws related to it are investigated. The existing tourism ontology design and construction are also studied. Protégé’ is used as a tool to construct the ontology concept. Its plugin inference engine is used to check for the reasoning correctness. We also evaluate the ontology using the domain expert and ontology expert based on existing metrics both internal and external metrics.

At last, the application of the ontology is built. We are developing the semantic web on Health tourism for Hua Hin district. The example use of axioms is also shown for inferencing new knowledge.

Declarations

Authors’ contributions

Chantana Chantrapornchai: Perform ontology design, perform writeup and analysis. Chidchankok Choksuchat: Evaluate existing research/ontologies, development of ontology, information collection and import. Both authors read and approved the final manuscript.

Acknowledgements

This work is supported in part by Royal Golden Jubilee Ph.D. Program under Grant PHD.005/2554, Faculty of Engineering, Kasetsart University Funding Contract No. 57/12/MATE, and The Thailand Research Fund, Tourism and Hospitality Management Program (TRF-THM) Grant Contract No. 5850042.

Competing interests

The authors declare that they have no competing interests.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors’ Affiliations

(1)
Department of Computer Engineering, Faculty of Engineering, Kasetsart University
(2)
Department of Computing, Faculty of Science, Silpakorn University

References

  1. Alani H, Sanghee K et al (2003) Automatic ontology-based knowledge extraction from web documents. IEEE Intell Syst 18(1):14–21View ArticleGoogle Scholar
  2. Almeida MB (2009) A proposal to evaluate ontology content. Appl Ontol 4(3–4):245–265MathSciNetGoogle Scholar
  3. Auer S, Bizer C et al (2007) DBpedia: a nucleus for a web of open data. Semant Web 4825:722–735View ArticleGoogle Scholar
  4. Balduini M, Celino I et al (2012) BOTTARI: an augmented reality mobile application to deliver personalized and location-based recommendations by continuous analysis of social media streams. Web Semant Sci Serv Agents World Wide Web 16:33–41View ArticleGoogle Scholar
  5. Bechhofer S, van Harmelen F et al (2004) OWL web ontology language reference. Retrieved 24 May 2014, from http://www.w3.org/TR/owl-ref/
  6. Beek P, Horrocks I (2005) OWL: a description logic based ontology language. Principles and practice of constraint programming—CP 2005, vol 3709. Springer, Berlin, pp 5–8Google Scholar
  7. Berners-Lee T, Hendler J et al (2001) The semantic web. Scientific American, New YorkGoogle Scholar
  8. Caballero-Danell S, Mugomba C (2007) Medical tourism and its entrepreneurial opportunities—a conceptual framework for entry into the industry. Tourism and Hospitality Management Goteborg University. Master Degree: 111Google Scholar
  9. Cardoso J (2006) Developing an owl ontology for e-tourism. In: Cardoso J, Sheth AP (eds) Semantic web services, processes and applications, vol 3. Springer, New York, pp 247–282View ArticleGoogle Scholar
  10. Cunningham H et al (2014) GATE: general architecture for text engineering. Retrieved 20 May 2014, from http://gate.ac.uk/sale/talks/gate-course-july09/hands-on-resources/questio/ontology/offline/tgproton.owl
  11. Choksuchat C, Ngamphak S et al (2014) Parallel health tourism information extraction and ontology storage. In: The international joint conference on computer science and software engineering (JCSSE), Pattaya, ThailandGoogle Scholar
  12. Constantinides C (2011) Wellness tourism and medical tourism…different concepts? Retrieved 20 May 2014, from http://www.imtj.com/articles/2011/spas-and-the-wellness-tourism-sector-30110/
  13. Cyganiak R, Jentzsch A (2011) The linking open data cloud diagram. Retrieved 25 May 2014, from http://lod-cloud.net/
  14. Daramola O, Adigun M et al (2009) Building an ontology-based framework for tourism recommendation services. In: Höpken W, Gretzel U, Law R (eds) Information and communication technologies in tourism 2009. Springer, Vienna, pp 135–147View ArticleGoogle Scholar
  15. Dell’Erba M, Fodor O et al (2002) Harmonise: a solution for data interoperability. I3E 233:433–445Google Scholar
  16. Dell’Erba M, Fodor O et al (2005) Exploiting semantic web technologies for harmonizing e-markets. J Inf Technol Tour 7:201–209View ArticleGoogle Scholar
  17. Dentler K, Cornet R et al (2011) Comparison of reasoners for large ontologies in the OWL 2 EL profile. Semant Web 2(2):71–87Google Scholar
  18. Dogac A, Kabak Y et al (2004) Semantically enriched web services for the travel industry. SIGMOD Rec 33(3):21–27View ArticleGoogle Scholar
  19. Fernández-López M, Gómez-Pérez A, Sierra JP, Sierra AP (1999) Building a chemical ontology using methontology and the ontology design environment. Intell Syst 14(1):37–46View ArticleGoogle Scholar
  20. Foder O, Werther H (2005) Harmonise: a step toward an interoperable e-tourism marketplace. Int J Electron Commer 9:11–39Google Scholar
  21. Gómez-Pérez A (1996) A framework to verify knowledge sharing technology. Expert Syst Appl 11(4):519–529View ArticleGoogle Scholar
  22. Gouveia M, Cardoso J (2007) Tourism information aggregation using an ontology based approach. In: ICEISGoogle Scholar
  23. HPI (2011) Data Hub LOD validator. Retrieved 25 May 2014, from http://validator.lod-cloud.net/levels.html
  24. Karoui L, Aufaure M-A et al (2004) Ontology discovery from web pages: application to tourism. In: Workshop of knowledge discovery and ontologiesGoogle Scholar
  25. Kehagias DD, Papadimitriou I et al (2008) A methodological approach for ontology evaluation and refinement. In: ASK-IT international conferenceGoogle Scholar
  26. Knublauch H. Retrieved 20 April 2014, from protege.cim3.net/file/pub/ontologies/travel/travel.owl‎Google Scholar
  27. Lehmann J, Auer S et al (2011) Class expression learning for ontology engineering. Web Semant Sci Serv Agents World Wide Web 9(1):71–81View ArticleGoogle Scholar
  28. Meersman R, Dillon T et al (2009) How to write and use the ontology requirements specification document. In: Meersman R, Dillon T, Herrero P (eds) On the move to meaningful internet systems: OTM 2009, vol 5871. Springer, Berlin, pp 966–982Google Scholar
  29. Mili H et al (2011) E-Tourism portal: a case study in ontology-driven development. In: Babin G, Stanoevska-Slabeva K, Kropf P (eds) E-technologies: transformation in a connected world. 5th international conference, MCETECH 2011, Les Diablerets, Switzerland, January 23–26, 2011, revised relected papers. Lecture notes in business information processing, vol 78. Springer, HeidelbergGoogle Scholar
  30. Mondeca (2006) Eiffel—French National Software Technology Network Project ANR—2006–2009. Retrieved 20 April 2014, from http://www.mondeca.com/Research/Projects/EIFFEL-French-National-Software-Technology-Network-Project-ANR-2006-2009
  31. Motik B, Sattler U, Studer R (2005) Query answering for OWL-DL with rules. Web Semant Sci Serv Agents World Wide Web 3(1):41–60View ArticleGoogle Scholar
  32. Mouhim S, Aoufi AE et al (2011) A knowledge management approach based on ontologies: the case of tourism. Int J Comput Sci Emerg Technol 4(3):362–369Google Scholar
  33. Mueller H, Kaufmann EL (2001) Wellness tourism: market analysis of a special health tourism segment and implications for the hotel industry. J Vacat Mark 7(1):5–17View ArticleGoogle Scholar
  34. Mugellini E, Szczepaniak P et al (2011) A quality assurance framework for ontology construction and refinement. In: Proceedings of 7th atlantic web intelligence conference (AWIC2011), Fribourg, Switzerland, January 26–28, 2011, pp 207-216Google Scholar
  35. Ogata N (2001) A formal ontology discovery from web documents. In: Web intelligence: research and development, vol 2198. Springer, Heidelberg, pp 514–519Google Scholar
  36. Ou S (2008) QALL-ME ontology. http://qallme.fbk.eu. Retreived 1 Dec 2016
  37. Patel-Schneider PF, Horrocks I (2006) OWL 1.1 web ontology language. Retrieved 24 May 2014, from http://www.w3.org/Submission/2006/SUBM-owl11-overview-20061219/#2.3
  38. Põld H. Health tourism. Retrieved 20 May 2014, from http://www.eas.ee/en/for-the-entrepreneur/tourism/tourism-product-development/health-tourism
  39. Ruiz-Martínez JM, Miñarro-Giménez JA et al (2011) Ontology population: an application for the e-tourism domain. Int J Innov Comput Inf Control 7(11):6115–6133Google Scholar
  40. Sigala M, Mich L et al (2007) Connecting destinations with an ontology-based e-tourism planner. In: Sigala M, Mich L, Murphy J (eds) Information and communication technologies in tourism 2007. Springer, Vienna, pp 21–32View ArticleGoogle Scholar
  41. Siorpaes K, Prantner K et al (2004) OnTour documentation. Retrieved 20 May 2014, from http://e-tourism.deri.at/ont/
  42. Smith M, Puczko L (2009) Health and wellness tourism. Elsvier, AmsterdamGoogle Scholar
  43. Suárez-Figueroa MC, Aguado de Cea G, Buil C, Dellschaft K, Fernández-López M, García A, Gómez-Pérez A, Herrero G, Montiel-Ponsoda E, Sabou M, Villazon-Terrazas B, Yufei Z (2008) Neon methodology for building contextualized ontology networks. Technical report, NeOn Deliverable D5.4.1Google Scholar
  44. Supekar K (2005) A peer-review approach for ontology evaluation. In: 8th international Protégé conference, Madrid, SpainGoogle Scholar
  45. Tang S, Cai Z (2010) Tourism domain ontology construction from the unstructured text documents. In: 2010 9th IEEE international conference on cognitive informatics (ICCI)Google Scholar
  46. Tankeleviciene L, Damasevicius R (2009) Characteristics of domain ontologies for web based learning and their application for quality evaluation. Inf Educ 8(1):131–152Google Scholar
  47. Thai Board of Investment (2012) Thailand medical hub of Asia. http://www.thinkasiainvestthailand.com/download/Medical.pdf. Accessed 18 July 2016

Copyright

© The Author(s) 2016