semantic web empowered e-tourism · 2020-04-24 · semantic web empowered e-tourism kevin angele,...

Semantic Web Empowered E-Tourism

Kevin Angele, Dieter Fensel, Elwin Huaman, Elias Kärle,Oleksandra Panasiuk, Umutcan Simsek, Ioan Toma,and Alexander Wahler

Contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2Semantic Web in a Nutshell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3Semantic Web of Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Schema Definition Languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Annotation Languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Ontologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

Semantic Web of Data: Linked Open Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13Semantic Web of Services: Semantic Web Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Heavyweight Semantic Web Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16Lightweight Semantic Web Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17Illustration: A Hotel Booking Chatbot Based on Schema.org Actions . . . . . . . . . . . . . . . . . 18

Knowledge Graph Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19Knowledge Creation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22Knowledge Hosting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25Knowledge Curation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26Knowledge Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Use Cases in E-Tourism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Touristic Chatbots and Intelligent Personal Assistants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Open Touristic Knowledge Graph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

K. AngeleSemantic Technology Institute, University of Innsbruck, Innsbruck, Austriae-mail: [email protected] GmbH, Telfs, Austria

D. Fensel (�) · E. Huaman · E. Kärle · O. Panasiuk · U. SimsekSemantic Technology Institute, University of Innsbruck, Innsbruck, Austriae-mail: [email protected]; [email protected]; [email protected];[email protected]; [email protected]

I. Toma · A. WahlerOnlim GmbH, Telfs, Austriae-mail: [email protected]; [email protected]

© Springer Nature Switzerland AG 2020Z. Xiang et al. (eds.), Handbook of e-Tourism,https://doi.org/10.1007/978-3-030-05324-6_22-1

1

http://crossmark.crossref.org/dialog/?doi=10.1007/978-3-030-05324-6_22-1&domain=pdf

mailto:[email protected]








https://doi.org/10.1007/978-3-030-05324-6_22-1

2 D. Fensel et al.

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Expected Future Developments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Abstract

Smart speakers such as Alexa and later Google Home have introduced ArtificialIntelligence (AI) into millions, soon to be billions of households, making AIan everyday experience. These new communication channels present a newchallenge for successful e-Marketing and e-Commerce providers. Data, content,and services are becoming semantically annotated, allowing software agents, so-called bots, to search through the web and understand its content. Nowadays,users typically consult their bot to find, aggregate, and personalize informationand to reserve, book, or buy products and services. As a consequence, it isbecoming increasingly important for touristic providers of information, products,and services to be prominently visible in these new online channels to ensuretheir future economic maturity. In our chapter, we survey the methods and toolshelping to achieve these goals. The core aim is the development and applicationof machine-processable (semantic) annotations of content, data, and services,as well as their aggregation in large Knowledge Graphs. It is only throughthese methods bots are able to answer a question in a knowledgeable way andorganize a useful dialogue (Knowledge Graphs in Use A significantly extendedand generalized version of this article will appear as D. Fensel, K. Angele,E. Huaman, E. Kärle, O. Panasiuk, U. Simsek, I. Toma, J. Umbrich, andA. Wahler: Knowledge Graphs: Methodology, Tools and Selected Use Cases.Springer Nature, 2020.).

Keywords

Smart speakers · Artificial Intelligence (AI) · e-Marketing · e-Commerce ·Knowledge graphs · Semantic web · Semantic technologies

Introduction

In Berners-Lee et al. (2001), the authors envisaged a web where now it is nolonger humans but bots who are accessing information on the web, and thesebots are supporting humans in the fulfilment of their tasks. Content, data, andservices must be enriched with machine-processable semantics to be accessibleby these bots. We are now on the brink of seeing this vision becoming reality.Meanwhile, a large body of work has been done to provide rich frameworks forthe semantic description of content, data, and services and large, heterogeneous,dynamic, and distributed environments. Furthermore, industrial de-facto standardssuch as schema.org1 provide the necessary impact for such approaches. Just as it hasbecome a must for an e-Marketing and e-Commerce provider to have a web site, itis now state of the art to add semantic description to web presences.

Semantic Web Empowered E-Tourism 3

As a result of billions of semantic statements on the web, a new area ofresearch has been evolving: so-called knowledge graphs; see Bonatti et al. (2019).They integrate these statements with other data and information sources and buildknowledge-based systems using billions and soon trillions of facts. This introducesnew requirements for the scalability of traditional knowledge-based technologiesand provides significant new opportunities for online interaction with potentialcustomers. Meanwhile, all core players in e-Tourism are using this technologyto enrich their service offers with customized information on points of interests,events, and further information of potential interest for their customers. In the end,a customer may not be simply wanting to book a hotel room but is instead searchingfor an exciting holiday experience with a variety of entertaining and informativeaspects and therefore may also look for a room to sleep in. This significantlyincreases the requirements for scalable and dynamic information integration on aworldwide scale. Ignoring this may lead to a provider being invisible to potentialguests.

In our chapter, we provide a survey of the available methods and technologies tosemantically enable e-Marketing and e-Commerce, with a focus on e-Tourism. Weintroduce the core and essence of semantic web technology in section “SemanticWeb in a Nutshell.” Sections “Semantic Web of Content,” “Semantic Web of Data:Linked Open Data,” and “Semantic Web of Services: Semantic Web Services”elaborate on the different means for annotating content, data, and services, whilesection “Knowledge Graph Technology” focuses on Knowledge Graphs and dis-cusses their essence, creation, hosting, curation, and deployment. Section “UseCases in E-Tourism” describes use cases and pilots in the area of e-Tourism, andfinal conclusions are provided in section “Conclusions.”

Semantic Web in a Nutshell

We are currently at the beginning of a major paradigm shift in accessing and sharinginformation on the Internet. In fact, this is not the first time that the Internet hasdrastically changed the way we cooperate and communicate. The Internet beganin the 1960s as a local network of four computers in the USA and evolved overthe next 20 years into a worldwide computer network. An early paradigm shiftfor human communication based on it was email, which has provided an instant-based messaging service to a fast-growing number of people. A complementaryinteraction paradigm began in 1989 based on the work of Tim Berners-Lee. Insteadof messaging, the World Wide Web (hereafter simply referred to as the web) isbased on publishing information to a large number of potential readers. The webis an information space where documents and other web resources are describedby hypertext markup, interlinked by hypertext links, identified by URIs, and canbe accessed via the Internet. This combination of hypertext with the Internetwas Berners-Lee’s actual innovation. Soon this information space began growingdramatically and crowded out all the competing approaches. Research on theSemantic Web began in 1996 for two reasons. First, the aim has been to support

4 D. Fensel et al.

<html><body><a onto="page:Researcher"><h2>Welcome on my homepage </h2>My name is <a onto="[name=body]"> Richard Benjamins </a>.

</a></body>

</html>

Object[].

Concept Hierarchy Attribute Definitions Rules

Person[ FORALL Person1, Publicaiton1Publication1:Publication[author ->> Person1]<->Person1:Person[publication ->> Publication1].

firstName =>> STRING;lastName =>> STRING;eMail =>> STRING;

publication =>> Publication].Employee[

affiliation =>> Organization:...].

...

Person :: Object.Employee :: Person.AcademicStaff :: Employee.

Researcher :: AcademicStaff.Publication :: Object.

Fig. 1 An early example of a schema and annotation language

the web in its nearly infinite scale. As more information is added, more machinesupport is needed to access relevant information pieces. In Fensel et al. (1997, 2000),we described a semantic web system based on a schema (Ontology), annotations ofcontent (based on an annotation language called HTML-a), and reasoning enginesand crawlers to access and process the available information (see also Fig. 1). Thesecond goal was to solve the knowledge acquisition bottleneck and create a brain forhumanity (cf. Fensel and Musen 2001). Billions of humans put data, information,and knowledge on this global network for free. Through this, the web mirrorslarge fractions of the human knowledge, and a new brain of humanity based onthe knowledge of mankind is generated. Empowered by semantics, computers canaccess and understand this knowledge. This vision of the Semantic Web has been tobuild a brain of/for human kind.

Unfortunately, in the period around the millennium, web search engines arosethat chose a different approach toward information access on the web. They basedtheir operations on syntax and statistical analysis, and, in fact, some of themperformed quite amazingly in retrieving a proper list of links to follow a givenkeyword as an input. Obviously, the statistical analysis of web resources is enoughto provide a fast and excellent index system for the web. The situation changeddrastically around 2011 when search engines such as Google tried to reach theirnext level of service. Originally, the business model was quite simple. Ads on theGoogle site brought revenue because increasing numbers of users used it as thestarting point for their web surfing. After they found an interesting link, they leftthe Google site and manually extracted information from the websites they visited.This search engine business model was extremely successful but ultimately limited.As quickly as users entered the Google site, as soon did they leave it. Therefore,step by step, Google is aiming to turn from a search into a query answeringengine (see Guha et al. 2003; Harth et al. 2007). Why point the visitor to otherwebsites? Why not provide the answer to his query directly at the Google website,thus keeping him there and opening new opportunities with him for commercial


cooperation. However, this requires more intelligence on the Google side. It mustbe able to extract exact information from a website based on machine-processablesemantics of content and data. In fact, achieving this goal requires more elaboratedapproaches. As a consequence, around 2011, a coalition of leading search enginesbegan the schema.org1.initiative that allows for the injection of semantic annotationsin HTML code based on JSON-LD, Microdata, and RDFa syntax. Meanwhile, amature corpus of types, properties, range restrictions, and enumeration values havebeen developed, and the uptake is significant. Most important websites are usingit. In complementary fashion, Google has developed its Google Knowledge Graph,a knowledge base already containing around 100 billion facts about more than 5billion entities. These figures offer substantial proof that the knowledge acquisitionbottleneck is being bypassed.

Furthermore, Google is certainly not the only player in this game. The currenthype is focused around chatbots and Intelligent Personal Assistants, which aretargeting this new access layer on top of the web. Alexa, Bixby, Cortana, Facebookmessenger, Google assistant, Siri, and others provide personalized and (spoken)message-based access to information. Clearly this generates new challenges forentities that need to make their content, data, and services visible to potentialcustomers. Just as it was a must 20 years ago to communicate via email and bevisible on the web, it is now key for economic success to be firmly present inthis new era of dialog-based information access. In the following sections, we willintroduce some core technologies underlying these efforts and illustrate their usagein e-Tourism.

Semantic Web of Content

This section introduces the schema definition languages, annotation languages, andpredefined vocabularies (Ontologies) used in the Semantic Web. Figure 2 presentsthe language stack of semantic web languages. On top of the stack is the application,and the bottom of the stack is the Semantic Web of Linked Data. In the first part ofthis section, we describe the languages used in the Dictionaries layer. The secondpart describes the annotation languages itself, which are part of Document Typeslayer. The third part is again focused on the Dictionaries layer, but in contrast to thefirst part, we describe predefined Ontologies, i.e., vocabularies.

Schema Definition Languages

In this section, we present schema definition languages. We first start with RDFand continue with RDFS, a schema definition language for RDF. Subsequently, wepresent OWL2, SKOS, and RIF.

Resource Description Framework (RDF)RDF (Manola et al. 2004) is a standard language to interchange semantic data onthe web. RDF is a foundation for describing such metadata and can be used in

6 D. Fensel et al.

Smart (Cognitive) Applications & Services

Trust

Proof

Unifying Logic

Dictionaries

First-Order Logic (FOL)

SWRL, SPIN, R2RML, SHACL(Ontologies)

RDF, RDFS, OWL,SKOS, Schema.org

Security(Crypto)

SPARQL, SPASQL

HTTP IRIs & URIs

RDF-NTriples, RDF-Turtle, RDF-XML,RDF-JSON,JSON-LD, others

RDF Subject->Predicate->Object Sentences

TransmissionRules

Query

Abstract Language

Sentence Part Identifiers

Document Types

Semantic Web of Linked Data

Fig. 2 Language st ack of the Semantic Web. (Cited from Language stack of semantic-webhttps://cdn-images-1.medium.com/max/1600/1*YQ04iyBrbq-VrQwkzCMkkA.png, accessed: 02Jun 2019)

many application areas. RDF defines triples which have the following structure.They consist of a subject (URI), a property (predicate), and an object (blank node,literal, URI). An example of such a triple is as follows:

Listing 1 Example of a RDF tripleOra Lassila is the creator of the resource http://www.w3.org/Home/Lassila

Subject: "http://www.w3.org/Home/Lassila"Predicate: "creator"Object: "Ora Lassila"

As the object can be an URI, RDF defines a graph as a data model. The normativeXML syntax of RDF is not very easy to use. Therefore, there are many syntaxes ontop of RDF, e.g., Turtle (Beckett et al. 2014). This makes it easier to use RDF as aninterchange format.

RDF-SchemaRDF-Schema (Brickley et al. 2014) is an extension of the basic RDF vocabulary. Itis used as data-modeling vocabulary to represent RDF data. It provides mechanisms

https://cdn-images-1.medium.com/max/1600/1*YQ04iyBrbq-VrQwkzCMkkA.png


Fig. 3 RDFS exampledefinition. (Cited from https://en.wikipedia.org/wiki/RDF_Schema (CC BY-SA3.0))

ex:dog1rdf:type

rdf:type

zoo:host

RDF special terms RDFS special terms

rdfs:

rang

erdfs:sub Classof

ex:cat1 ex:cat

ex:zoo1

ex:animal

ex:cat2

describing groups of related resources and the relationships between resources.These definitions are written again in RDF. The following example in Fig. 3 showsa definition of two types and some instances. It shows the special RDFS termsrdfs:subClassOf and rdfs:range. RDFS defines a simple Knowledge RepresentationFormalism for the Semantic Web.

Web Ontology Language (OWL2)OWL2 (Beckett et al. 2014) is the recommended language used to describe ontolo-gies on the web. The ontologies are exchanged as RDF documents. OWL2 is anextension of OWL, which was published by the W3C Web Ontology Working Groupin 2004. The goal of OWL2 is, as with OWL, to make web content better accessibleand understandable for machines by applying and customizing Description Logic(Baader et al. 2003). OWL2 is a very expressive language, and, therefore, it isdifficult to implement and work with it. Furthermore, OWL2 also defines differentprofiles (DL, Full, EL, QL) which makes it hard for all those profiles to be supported.In particular, the syntax of OWL2 is very complex.

Simple Knowledge Organization System (SKOS)SKOS (Miles and Bechhofer 2009) is a recommendation used for data sharing. Thismeans that an existing knowledge organization system can be expressed in a waythat it is machine-readable. The idea is to exchange data between different computerapplications and also publish the description on the web. The data model of SKOSis defined as an OWL Full ontology. Concept schemas for knowledge representationsystems are represented by SKOS. The following example presents SKOS dataexpressed as RDF triples. It describes some of the terms from a thesaurus2:

Listing 2 SKOS data expressed as RDF triples<A> rdf:type skos:Concept;

skos:prefLabel "love"@en;skos:altLabel "adoration"@en;

skos:broader <B>;skos:inScheme <S>.

<B> rdf:type skos:Concept ;skos:prefLabel "emotion"@en;skos:altLabel "feeling"@en;skos:topConceptOf <S>.

<S> rdf:type skos:ConceptScheme ;dct:title "My First Thesaurus";skos:hasTopConcept <B> .

https://en.wikipedia.org/wiki/RDF_Schema



8 D. Fensel et al.

Annotation Languages

In this section, we present the languages used for annotating information on the web.Annotating means describing the available information in a way that it can be readand interpreted by machines.

Microformats“Humans first, machines second” is the slogan of microformats (Khare and Çelik2006). The principles of microformats are that they should solve a specific problem,start as simply as possible, reuse building blocks from existing standards, and mustbe modular embeddable. Microformats take advantage of existing XHTML facilitiesto reuse existing web pages for new services and applications. The followingexample shows a calendar event represented in microformat XHTML:

Listing 3 Example calendar event represented in microformat XHTML. (Cited from Khare andÇelik 2006)<div class="vcalendar-vevent">

<span class="summary">Microformats:What the Hell Are They and Why Should I Care?

</span><p class="description">

Ryan King will explain why microformats are important and how youcan markup specific kinds of content in ways that make it easierfor the right people to find your stuff.

</p><abbr class="dtstart" title="20050926T050000-0700">

September 25th, 2005, 5</abbr><abbr class="dtend" title="20050926T060000-0700">

6PM</abbr>in the<span class="location">

Balder Room</span>

</div>

Limitations include the fact that not all things can be represented using HTMLtags. For example, the address tag is used for the author of the web site; thereforeit cannot be used to represent a location of an event. In that case, you need to useother mechanisms such as vCard to represent an event. This means in general thatyou have to combine different standards to describe your information correctly.The latest version is Microformats2,3which combines the lessons learned fromMicrodata and RDFa. It was published in 2012.

MicrodataMicrodata4 is a specification for new HTML attributes that offers the possibilityof embedding machine-readable data in HTML documents. The approach is verysimilar to RDFa, but Microdata is not as expressive. Furthermore, RDFa andMicrodata do not have the same level of internationalization. Essentially, Microdataadds labels to content in a document. Therefore, the content can then be interpreted


as name-value-pairs. The following example shows a definition of the propertyname:

Listing 4 Microdata definition of name in HTML document. (Cited from Microdata https://www.w3.org/TR/microdata/, accessed: 14 Jun 2019)<div itemscope>

<p>My name is<span itemprop="name">

Elizabeth</span>.

</p></div>

With itemscope, you can start a new definition, and itemprop then defines the nameof the property. Limitations arise if you need internationalization to support differentlanguages. To give internationalization information, they need to be encoded asMicrodata. The downside is that it does not follow an established standard, so itmay not be understood by users of the information.

Resource Description Framework in Attributes (RDFa)RDFa5 adds structured data to HTML pages directly. It provides a set of markupattributes for machine-readable hints. RDF 1.0 was specified only for XHTML. Thenew version 1.1. is specified for both XHTML and HTML5. The idea is to reusesome HTML attributes. Furthermore, with the use of simple HTML tags, the websiteis made understandable by machines that crawl web pages. As already mentioned,RDFa is based on existing HTML tags and also adds new attributes. The RDFadesign choices replicate an early Semantic Web proposal called HTML-a (Fenselet al. 2000). The following example shows a HTML page with annotations made inRDFa. The property attributes are the RDFa annotations.

Listing 5 HTML page with annotations made in RDFa. (Cited from RDFa https://www.w3.org/TR/xhtml-rdfa-primer/, accessed: 14 Jun 2019)<html>

<body>...<h2 property="http://purl.org/dc/terms/title">

The Trouble with Bob</h2><p>

Date:<span property="http://purl.org/dc/terms/created">

2011-09-10</span>

</p>...

</body></html>

This example annotates the title and the date when it was created so that machinescan extract this information.

https://www.w3.org/TR/microdata/


https://www.w3.org/TR/xhtml-rdfa-primer/


10 D. Fensel et al.

JavaScript Object Notation in Linked Data (JSON-LD)JSON-LD6 is a format to serialize Linked Data based on JSON.7 With minimalchanges, existing JSON can be extended to be interpreted as Linked Data. Thedesign goals of JSON-LD are simplicity, compatibility, expressiveness, and as fewedits as possible to enrich a normal JSON. With JSON-LD, it is possible to embedlinks to other pieces of information saved on different sites across the web. AsJSON-LD is 100% compatible with JSON, it is well supported. Nowadays, nearlyevery programming language has a JSON parser; therefore JSON-LD can also beparsed by nearly all of them. The following example8 shows a definition of a personwith name, url, and image. The @id keyword means that the given value should beinterpreted as IRI (Internationalized Resource Identifier).

Listing 6 Example definition of person with name, url and image{"http://schema.org/name": "Manu Sporny","http://schema.org/url": {"@id": "http://manu.sporny.org/"

},"http://schema.org/image": {"@id": "http://manu.sporny.org/images/manu.png"

}}

The benefit of JSON-LD is that it is not only machine-readable and machine-understandable, but also humans can easily read JSON-LD definitions. Currently,JSON-LD 1.1 is in development. The latest editor’s draft for the new version can befound here9:

Ontologies

In the previous sections, we defined schema and annotation languages. This sectionis about a predefined set of vocabularies. These vocabularies can be used whenannotating information on the web. The benefit of predefined vocabularies is thatthey provide a common understanding. This means that everyone who uses themhas the same understanding of what the annotation means.

Dublin CoreDublin Core10 is a lightweight RDFS vocabulary describing generic metadata. A setof attributes are provided to define a term. A term can consist of a name, label, URI,definition, and the type, and additional attributes can be used to describe metadata.As an example, we give the definition of a format, taken from (Schema.org Airporthttps://schema.org/Airport, accessed: 14 Jun 2019.):

– URI: http://purl.org/dc/elements/1.1/format– Label: Format– Definition: The file format, physical medium, or dimensions of the resource.

https://schema.org/Airport

http://purl.org/dc/elements/1.1/format


– Comment: Examples of dimensions include size and duration. Recommendedbest practice is to use a controlled vocabulary such as the list of Internet MediaTypes [MIME].

– References: [MIME] http://www.iana.org/assignments/media-types/– Type of Term: http://www.w3.org/1999/02/22-rdf-syntax-ns#Property– Note: A second property with the same name as this property has been declared

in the dcterms: namespace (http://purl.org/dc/terms/). See the introduction to thedocument “DCMI Metadata Terms” (/specifications/dublin-core/dcmi-terms/)for an explanation.

Friend of a Friend (FOAF)FOAF11 is an ontology to link people and information. Social networks, representa-tional networks, and information networks are integrated in FOAF. The descriptionsof FOAF are published on the web. We present here an example of a persondescribed with FOAF. The example is taken from11.:

Listing 7 Friend of a friend example of a person<foaf:Person rdf:about="#danbri" xmlns:foaf="http://xmlns.com/foaf/0.1/">

<foaf:name>Dan Brickley</foaf:name><foaf:homepage rdf:resource="http://danbri.org/" /><foaf:openid rdf:resource="http://danbri.org/" /><foaf:img rdf:resource="/images/me.jpg" />

</foaf:Person>

This example defines the name, homepage, the openid, and an image for the person.

GoodRelationsGoodRelations12 presents a data structure for e-Commerce that is industry-neutral,valid across different stages of the value chain, and syntax-neutral. Four entities rep-resent the e-Commerce scenarios: an agent (e.g., person), object (e.g., camcorder),promise (offer), and a location (e.g., a store). GoodRelations is a fully fledgedMicrodata vocabulary. In GoodRelations, only the local part of a property identifieris used, in contrast to Microdata and RDFa. See the following example, taken fromGoodRelations12.:

Listing 8 GoodRelations example of a Product<div itemscope itemtype=http://purl.org/goodrelations/v1#Individualitemid="#product">

Weight: <div itemprop="weight" itemscopeitemtype="http://purl.org/goodrelations/v1#QuantitativeValue"><span itemprop="hasValue">50</span> kg<meta itemprop="hasUnitOfMeasurement" content="KGM" >

</div>

Schema.orgSchemas for structured data on the Internet are provided by schema.org.1.They canbe used in many different encodings such as RDFa, Microdata, and JSON-LD.The vocabularies cover entities and relationships between entities and actions.Schema.org provides a mechanism to extend the given set of schemas with its

http://www.iana.org/assignments/media-types/

http://www.w3.org/1999/02/22-rdf-syntax-ns#Property

http://purl.org/dc/terms/

12 D. Fensel et al.

Fig. 4 Screenshot from schema.org (Schema.org Airport https://schema.org/Airport, accessed: 14Jun 2019.)

own schemas. The following screenshot in Fig. 4 shows the schema for an airport.The schema defines the possible properties and their ranges. Schema.org providesschemas for many domains and topics. Given its industrial support, it seems tohave become a de-facto standard, but it is also very complex, provides numerousalternatives to be used, and is still fairly incomplete for the specifics of mostdomains. This makes the usage of schema.org reasonably difficult and is thereason why we have developed Domain Specifications (Simsek et al. 2018b), seesection “Knowledge Graph Technology” for more details, as a means to restrict,constrain, and domain specifically extend schema.org.

https://schema.org/Airport


Further TypesIn the tourism area, there are no globally accepted ontologies. Nevertheless, thereare some initiatives trying to establish an ontology to describe touristic information.One initiative is the Harmonise project (DellErba et al. 2003) that proposes anontology-based mediation and harmonization tool to allow touristic organizationsexchange data while keeping their own data format. Another initiative in thisdirection is the DACH-KG13 that is currently working on a common schema toexchange tourism information in the German-speaking area (Germany, Austria,Switzerland, and South Tirol (Italy)). The current state of the schema.org extensionsis available on github.14 Many more ontologies are available on the web as part ofthe Linked Open Data cloud (see the following section). Rather than being too fewof them, there are actually too many, comparable to the human language zoo ofmore than 6000 languages which makes interaction and cooperation beyond culturalboundaries so cumbersome.

Semantic Web of Data: Linked Open Data

Linked Open Data, sometimes called the “semantic web of data,” is Linked Datawhich is published as Open Data. Tim Berners-Lee coined the term Linked Dataand defined it as follows: “Linked Open Data (LOD) is Linked Data, which isreleased under an open license, which does not impede its reuse for free” (Berners-Lee 2006). Apparently the term Linked Open Data consists of the two terms LinkedData and Open Data. Before describing what Linked Open Data is and means, wewill introduce the concepts of Open Data and Linked Data.

Open Data describes all kinds of data which are published under an open license.The Open Data Handbook describes Open Data as “. . . data that can be freely used,re-used and redistributed by anyone – subject only, at most, to the requirement toattribute and share alike” (Dietrich et al. 2009). A common license applied to OpenData is CC-BY-SA, which is the Creative Commons license with the attributes BY,which defines that the author of the data has to be mentioned, and SA, which standsfor Share Alike and defines that the license also has to be attached when the data isreused.

Linked Data was defined by Berners-Lee in 2006 and is a common way to sharedata on the web in a machine-readable and machine-understandable way: “Withlinked data, when you have some of it, you can find other, related, data” (Berners-Lee 2006). Whereas on the web, documents are linked to each other with hyperlinks,in the web of data, data sets are linked to each other. On the web, those links aredescribed in hypertext, whereas in Linked Data, those links are described in RDF.To publish data as Linked Data, the publication must follow four principles:

1. Use URIs as names for things: this principle ensures that things in data sets areidentified uniquely.

2. Use HTTP URIs so that people can look up those names: this principle ensuresthat the information about the thing can be dereferenced on a web page.

14 D. Fensel et al.

3. When someone looks up a URI, provide useful information using the standards(RDF*, SPARQL): this principle ensures that the information provided is alsoread and interpretable for machines, since RDF allows/requires the semanticdescription of the things.

4. Include links to other URIs so that they can discover more things: the lastprinciple puts the “Linked” into Linked Data. Since these links, according toprinciple 3, are described in RDF, machines can follow them autonomously tofind more data on their own.

The step from Linked Data to Linked Open Data does not only imply that thedata is open and linked. It goes one step further and defines firstly that the data mustbe open and then covers the way data is accessed – preferably in nonproprietaryformats. Only the last requirement then brings in the linking aspect. The qualityof Linked Open Data is depicted in a star rating from one to five stars. The ratingcriteria build on each other, which means that having four stars implies that the dataalso matches the three-star criteria and so on:

∗ The data set gets awarded one star if the data is provided under an openlicense. This is a very nontechnical requirement, so the data could, forexample, be available as a pdf or an image.

∗∗ Two stars is awarded if the data is available as structured data. Thisrequirement is satisfied if it is machine-readable and has a certainstructure, for example, in an Excel sheet.

∗ ∗ ∗ Three stars is awarded if the data is also available in a nonproprietaryformat. Since Excel is a proprietary format, this criterion requires thedata to be available as, for example, CSV, JSON, or XML.

∗ ∗ ∗∗ Four stars is awarded if URIs are used so that the data can be referenced,as already defined in the Linked Data principles one and two.

∗ ∗ ∗ ∗ ∗ Five stars is awarded if the data set is linked to other data sets that canprovide context (according to Berners Lee 2015).

If a data set satisfies all five stars, it can become part of the “Linked Open DataCloud” (Bizer et al. 2008). The LOD-cloud (see Fig. 5) is a collection of data setswhich are all published according to the five-star criteria. As of March 2019, thereare 1,239 data sets in the cloud with 16,147 links. The LOD-cloud project started in2007, and the first version contained only 12 data sets. Back then, the central dataset was, and still is, DBpedia (Auer et al. 2007). The DBpedia project dedicateditself to making the data of the Wikipedia machine processable and publishing thecontent of Wikipedia as Linked Open Data. Other important datasets in the LOD-cloud are GeoNames (Maltese and Farazi 2013), a LOD representation of a databasecontaining more than 25 million geographical names; MusicBrainz (Swartz 2002),a comprehensive LOD representation of knowledge about music-related topics;and SNOMED clinical terms (Stearns et al. 2001), a large database of health-careterminologies. From the tourism prospective, the TourMISLOD dataset containsthe linked data encoding of European tourism statistics (Sabou et al. 2013). The


Fig. 5 The LOD-cloud, as of March 2019, with its 1239 data sets (Linked Open Data Cloudhttps://www.lod-cloud.net, accessed: 14 Jun 2019)

City Service Development Kit (CitySDK) is a system that collects open data ofgovernments to develop scalable Smart City services (Pereira et al. 2015).

Semantic Web of Services: Semantic Web Services

The web contains three main elements, namely, content, data, and services. TheSemantic Web that envisions the accommodation of intelligent agents for complet-ing tasks in an automated fashion cannot succeed without the semantic descriptionof these three elements. Therefore, the research need for semantic web serviceshas been identified by the community in the early stages of the Semantic Web(Ankolekar et al. 2002; Fensel and Bussler 2002). Semantic web services can have

https://www.lod-cloud.net

16 D. Fensel et al.

interesting application areas in e-Tourism, such as the automated composition ofdifferent services (e.g., flight, hotel, tour) in order to book travel arrangements orhelping the development of dialogue systems by explicitly providing the behaviorof a service that can guide a dialogue to complete a certain task.

This section gives an overview of the approaches for semantic web services.The first part of the section describes heavyweight approaches that mainly targetservices using the Simple Object Access Protocol (SOAP)15 as a messaging protocoland that are mostly used in the internal and external B2B communication of largeorganizations. The second part reviews the lightweight approaches that mainly targetRESTful services,16 i.e., service offers on the web mostly targeting B2C. Finally, weprovide a short illustration for semantic web services and their functionalities.

Heavyweight Semantic Web Services

The early efforts on semantic web services targeted services that use the WebService Description Language WSDL17 as a description language and SOAP asa protocol. These approaches mostly offered advanced mechanisms to describe webservices in order to enable automated agents to complete web service tasks suchas discovery, composition, and invocation automatically through logical reasoning.We consider them heavyweight in terms of supported web service protocol andadvanced mechanisms for describing web services semantically. In this section, weintroduce these heavyweight approaches.

OWL-S uses Web Ontology Language (OWL) to describe web services (Martinet al. 2007). It consists of three main components, namely, the Service Profile,Process Model, and Service Grounding. The Service Profile ontology enablesservice providers to create a description of some functional (e.g., capabilities) andnonfunctional (e.g., service category, QoS) properties. The Process Model describeshow the service will be provided (i.e., behavioral properties) and the conditionsand invocation steps to obtain certain outcomes. The third component, ServiceGrounding, describes the details of how services can be invoked concretely, forinstance, in terms of connecting with WSDL components.

Semantic Web Services Framework (SWSF)18 is built on top of the experiencegained from OWL-S. It aims to create a more expressive framework by usingfirst-order logic (FOL) instead of description logic (DL) for process modeling,which is, however, undecidable. The framework consists of two main elements,namely, The Semantic Web Services Ontology (SWSO) for modeling web servicesconceptually and The Semantic Web Services Language (SWSL) to express SWSO.In addition, SWSF uses an extended version of the Process Specification Language(PSL) (Gruninger and Menzel 2003) to define the behavioral aspects of the webservice. In general, it appears to be a valuable academic exercise.


METEOR-S (Patil et al. 2004) extends the existing web and web service technolo-gies with semantics. The core technology they adopt is the SAWSDL (SemanticAnnotations of Web Services) (Kopecky et al. 2007). The forthcoming feature ofMETEOR-S is that it supports the full life cycle of the web services and providestools for the design, discovery, composition, and execution of web services.

The Web Service Modeling Framework (WSMF) (Fensel and Bussler 2002)offers a comprehensive decoupled way of automating the whole life cycle of webservice consumption. It comprises a conceptual model, a modeling ontology WSMO(Fensel et al. 2006), a structured set of languages WSML (Fensel et al. 2006),and an execution environment WSMX (Fensel et al. 2008). WSMF has four maincomponents for describing web services: (1) ontologies that provide the means fordescribing the domain of discourse; (2) goals that represent the user’s perspectivefor the consumption of web services; (3) web service descriptions for describingfunctional, behavioral, and nonfunctional aspects of web services; and (4) mediatorsthat solve various interoperability problems. The most prominent feature of theWSMF approach is to provide an explicit mechanism for mediation and goals thatare distinguished from web service capabilities.

The Internet Reasoning Service (IRS-II) (Motta et al. 2003) is a framework andinfrastructure for the publication, storage, composition, and execution of heteroge-neous web services and their semantic descriptions. The IRS-II follows an approachthat applies UPML (Fensel et al. 1999), a framework for increasing knowledge-based system reusability through modularization, to semantic web services. TheIRS-II framework abstracts services as problem-solving methods and matches themwith tasks. In that sense, they are along the same lines as the WSMF approach.The most distinct feature of IRS-II is that it has advanced publication and registrymechanisms.

Lightweight Semantic Web Services

In this section, we describe the approaches that mainly target RESTful web serviceswith simpler mechanisms to semantically describe web services. A comprehensivesurvey of such approaches can be already found in Verborgh et al. (2014). We brieflysummarize some prominent approaches in the literature while focusing on the mostrecent ones.

WSMO-Lite (Roman et al. 2015) is a conceptual model for describing thefunctionality of RESTful web services in a lightweight, bottom-up manner. Unlikethe approaches for SOAP services (e.g., OWL-S, WSMO), it does not follow asemantics-first policy, but is connected directly to the syntax of a web servicedocumentation (i.e., HTML file) through a microformat called MicroWSMO.Although it has limitations in terms of expressiveness and description of thebehavioral aspects of web services, it provides a minimal model for web service

18 D. Fensel et al.

descriptions to foster interoperability. In fact, WSMO-Lite is also the conceptualmodel for SAWSDL, a bottom-up approach for annotating WSDL files.

RESTDesc (Verborgh et al. 2013) offers a mechanism for describing functionalaspects of RESTful web services based on N3Logic (Berners-Lee et al. 2008). Thefunctionality over a resource can be described in terms of pre- and post-conditions.An OPTIONS call to a resource would return the possible actions that can be takenon that resource along with their expected outcomes. A client can then completecertain tasks with a follow-your-nose approach.

Hydra (Lanthaler and Gütl 2013) is a RESTful web service documentationapproach mainly focused on hypermedia. The main principle of Hydra is that everyRESTful web service should be its own machine-readable documentation, meaningthat a client can consume a RESTful web service with minimal a priori knowledge(i.e., only an entry point and media types). A Hydra annotated web service operateswith JSON-LD format only and benefits from hypermedia over JSON-LD for thedescription of behavioral aspects, whereas the functional aspects are only limited toannotation of input and output of an operation on a resource.

SmartAPI (Zaveri et al. 2017) is a lightweight semantic extension to the OpenAPISpecification.19 It provides a mechanism for the annotation of input and output ofan operation on a resource of RESTful web services. Although the semantics arevery limited, SmartAPI benefits from the popularity and vast tooling support of theOpenAPI specification.

Schema.org actions20 allow RESTful web service publishers to annotate theirAPIs with semantic annotations. A high-level operation over a resource can berepresented with an action (e.g., SearchAction, BuyAction). These operations arethen mapped to an HTTP method, and the resource itself is described as theobject of the action. The relationship between input and output can be provided byconnecting the values of object and result properties. An example action is shownin Fig. 6. The SearchAction represents the description of a search operation on aLodgingReservation. The operation is expected to return an offer. The behavioralaspects are represented through potential actions on responses. Given the factthat Bing, Google, Yahoo!, and Yandex are supporting this approach for servicedescription, it may become the de facto standard for describing web servicessemantically.

Illustration: A Hotel Booking Chatbot Based on Schema.org Actions

The semantic descriptions of RESTful web services provide an opportunity forintelligent agents to be decoupled from the web services they interact with. Thiseven allows for the semiautomated generation of a dialogue system based on thesemantic descriptions. The dialogue in Fig. 7 is an interaction between a user and


Fig. 6 A partial annotationof a booking engine citedfrom Simsek et al. (2018a)

a semiautomatically generated hotel booking bot based on the action annotation inFig. 6. A generated “search” intent is triggered when the user reveals her desire tosearch for a specific hotel. After the required parameters are elicited from the user,the dialogue is dynamically guided by the potential action attached to the responseof the first request sent by the bot to the booking service. The action annotationsare also used to help conversational agent developers to create natural languagesentences for the training of the natural language understanding unit.

Knowledge Graph Technology

A “graph is a structure amounting to a set of objects in which some pairs of theobjects are in some sense related.” – discrete mathematics definition.21 Strictlyspeaking, we need to slightly extend this definition to multi-sets since the sameobject can syntactically and/or semantically appear several times in our graph. Thissimple definition can be extended in various directions, and we end up with anentire zoo of graph types: simple graphs, undirected versus directed graphs, orientedgraphs, mixed graphs, multigraphs, quiver, weighted graphs, half-edges and loose-edges graphs, finite versus infinite graphs, etc. In the semantic web community, theconsensus is to use RDF as representation formalism for representing a KnowledgeGraph.

The concept of Knowledge is somewhat hazier. If we return to what (Newellet al. 1982) called the knowledge level then, based on the assumption that an

20 D. Fensel et al.

Fig. 7 An example dialogue between a user and a generated hotel booking dialogue system citedfrom Simsek et al. (2018a)

agent follows the principle of rationality (later refined as the concept of boundedrationality (Simon 1957), including the costs for “optimal” decision making), wesubscribe knowledge to the agent, perceiving the actions it takes to achieve certaingoals. In this sense, knowledge is externally assigned to this agent by an observer.Internally, the “knowledge” is coded at the symbol level. “Beneath the knowledgelevel resides the symbol level. Whereas the knowledge level is world oriented,namely, that it concerns the environment in which the agent operates, the symbollevel is system oriented, in that it includes the mechanisms the agent has availableto operate. The knowledge level rationalizes the agent’s behavior, while the symbollevel mechanizes the agent’s behavior” (Newell et al. 1982). Obviously, we couldinterpret the Knowledge Graph in a similar way. An agent has/generates knowledgeby interpreting a graph, i.e., relates its elements to so-called real-world objects andactions. A graph is a specific encoding formalism. If we refine this further, we maywant to place the graph on the logical or epistemological level rather than on theimplementational level (Brachman 1979). At the implementation level, we havemeans such as graph-based databases, etc.

In fact, Google coined the term Knowledge Graph in 2012 (Singhal 2012) asa means to build a model of the world. Meanwhile, it has become a hype termin product and service industry. In tourism, note that tourism is one of the mostimportant economical verticals on a worldwide scale, accounting for around 10%of global GDP and total employment in 2017 (Council 2018), not necessarilythe most innovative area in general, every major player already has a knowledgegraph, and thousands of players (such as Destination Management Organizations)need or want one. The drive for this stems from how increasingly importantsuccessful e-Marketing and e-Commerce providers have become in terms of thevalue distribution in tourism and other areas.


Fig. 8 A process model for Knowledge Graphs

Summarizing the discussion, we can state that Knowledge Graphs are verylarge semantic nets that integrate various and heterogeneous information sourcesto represent knowledge about certain domains of discourse. According to Gomez-Perez et al. (2017), Knowledge Graph Technologies in a nutshell consist of:

• “knowledge representation and reasoning (languages, schema, and standardvocabularies),

• knowledge storage (graph databases and repositories),• knowledge engineering (methodologies, editors, and design patterns),• (automatic) knowledge learning including schema learning and population.”

Knowledge Graph methods and techniques must additionally reflect the specificfocus on very large amounts of instances beyond any tradition knowledge base; seeSchultz et al. (2012). We identify the following major steps for a process model (seeFig. 8):

• A traditional knowledge acquisition (perhaps better referred to as knowledgeengineering). Phase that establishes the core data for a Knowledge Graph (seesection “Knowledge Creation”).

• The process to implement this knowledge in a proper storage system, such asdocument or graph-based repository (see section “Knowledge Hosting”).

22 D. Fensel et al.

• The knowledge curation process (cf. Paulheim 2017) to establish large Knowl-edge Graphs of significant coverage and quality. We identify the followingactivities as substeps of this curation process: Knowledge Evaluation, Cleaning,and Enrichment (see section “Knowledge Curation”).

• Finally, we need to deploy and apply such a Knowledge Graph (see sec-tion “Knowledge Deployment”).

Similar process models can be found in Gawriljuk et al. (2016) and Villazon-Terrazas et al. (2017b). Each of the mentioned steps is discussed in detail in thefollowing subsections.

Knowledge Creation

The knowledge creation (also referred to as knowledge acquisition) process repre-sents the process of extracting knowledge from domain experts and available datasources, and structuring it and managing established knowledge (Schreiber et al.2000). The knowledge acquisition process was viewed as one of the most criticalaspects in the knowledge engineering process (Studer et al. 1998). Nowadays, theacquisition process is influenced by the web and has become an intense area ofresearch (Gil 2011; Schreiber 2013).

There exist different approaches for knowledge acquisition from the web.In Sánchez and Moreno (2006), the authors described the domain-independentlearning methodology modeled over multi-agent systems that crawl the web tosemiautomatically build an ontology for a given domain according to the user’sinterests. In Tandon et al. (2014), the method for automatically constructing a largecommon sense knowledge base from web contents is described. An overview ofexisting methods, tools, and techniques for knowledge elicitation, as a sub-processof knowledge acquisition, is given in Shadbolt et al. (2015). The paper describes theproblem of knowledge elicitation for knowledge-intensive systems from conven-tional expert systems through to intelligent tutoring systems, adaptive interfaces,and workflow support tools. The authors discuss the knowledge elicitation andmodeling from the perspective of knowledge engineering and in the context of theSemantic Web. Elsewhere, special attention has been paid to information extractionand natural language processing (NLP) technologies (Cambria and White 2014),as well as data mining and machine learning techniques (Silwattananusarn andTuamsuk 2012; Nickel et al. 2015). There is also interest in the use of ontologylearning techniques to create initial ontological structures and develop automaticmethods for knowledge extraction for a specific domain (Drumond and Girardi2008).

As an example, we present our methodology for semantic annotations; seeFig. 9. The methodology consists of three main parts: (i) the bottom-up part, whichdescribes the steps of the annotation process; (ii) the domain specification modeling;and (iii) the top-down part, which applies the constructed models.


Fig. 9 Methodology for semantic annotation cited from Panasiuk et al. (2018c)

The bottom-up part of the methodology helps to define the domain area,analyze domain entities, detect the format and type of data, select the ontology torepresent collected information, map data to schema.org vocabulary, provide anddeploy semantic annotations, and evaluate the results. The domain specificationsmodeling focuses on developing domain specification patterns called domain spec-ifications (DSs). A domain specification is an extended subset of types, properties,and ranges (Simsek et al. 2018b) of schema.org. The goal of a DS is to provide amodel of how a concrete domain should be represented in a semantically structuredway. The top-down part of the methodology describes how to map new incomedata to developed DSs with no need to carry out the steps of the bottom-uppart and perform annotation development according to domain specifications. Tosupport the semantic annotation process, tools are required to support the manualand semiautomatic editing process, automatic annotation tools, and mappings ofexternal schemas.

Manual editing The annotation process of web content can be done manuallyvia the semantify.it Annotation Editor22 (Kärle et al. 2017). The interface isautomatically generated based on the domain specification. To start a new manualannotation, the user selects the domain specification of the document on which theannotation will be based and obtains an appropriate editing interface. If all thenecessary fields are filled out, the user can get the source code of the annotationin the format JSON-LD. This source code can either be copied or saved on thesemantify.it platform for further use. The annotation editor can be used by usersto annotate their web content and make the semantic annotation process easier,complete, and consistent.

Semi-automatic editing The fields in the editor will be filled in by extractinginformation from the given URL or source file. If a source file is semi-structured,then the editor will suggest the mapping to JSON-LD by using the mappings

24 D. Fensel et al.

as an approximation based on the training data. If the content is unstructured,some approaches to extracting information from a web page can be applied. Theinformation can be extracted from the source web page by tracking the appropriateHTML tags. Some ontology discovery techniques for the tourism domain arediscussed in Karoui et al. (2004), and the tree-based technique of the documentobject model (DOM)23 of a web page is described in Gupta et al. (2003). In addition,semantic types and properties can be automatically inferred by using the semantictypes and properties, which have been trained to be recognized (Gupta et al. 2012).

Automatic annotation tools retrieve data from the web using natural languageprocessing (NLP) and machine learning (ML). There are some approaches toextracting knowledge of text presentation and web pages, such as named entityrecognition (Mohit 2014), information extraction (Chang et al. 2006), conceptmining (Shehata et al. 2009), text mining (Inzalkar and Sharma 2015), etc. There aremany tools or libraries, such as GATE24 for text analysis and language processing;OpenNLP,25 which supports the most common NLP tasks; and RapidMiner26 fordata preparation, machine learning, deep learning, text mining, and predictiveanalytics; see Villazon-Terrazas et al. (2017a). The typical tasks of NLP aredescribed in Moschitti et al. (2017).

Mapping However, in order for large and fast-changing data sets to be generatedeffectively and continuously, other methods are required. The data are oftenprovided by different institutions and might be both in and using different formats.To make this data assessable in the Knowledge Graph, we need to transfer itinto the format and schema of our knowledge representation formalism (Villazon-Terrazas et al. 2017a). The XLWrap approach generates graphs triples from specificcells of a spreadsheet (Langegger and Wöß 2009). Mapping Master (M2) is amapping language for converting spreadsheets to OWL (O’connor et al. 2010).A generic XMLtoRDF tool provides a mapping document (XML document) thathas a link between an XML schema and an OWL ontology (Van Deursen et al.2008). Tripliser27 is a Java library and command-line tool for creating triple graphsfrom XML. In addition, GRDDL28 translates the XML data into RDF. VirtuosoSponger29 generates Linked Data from a variety of data sources and supports a widevariety of data representation and serialization formats. R2RML30 specifies howto translate relational data into the RDF format. RDF Mapping Language (RML)(Dimou et al. 2014) extends R2RML’s applicability to define mappings of datain other formats. With RML, rules can be expressed that map data with differentstructures and serializations (e.g., databases, XML, or JSON data sources) to thedomain-specific schema.org data model (cf. Simsek et al. 2019).

The knowledge acquisition process plays an important role in the development ofknowledge graphs. It defines the methods and techniques necessary for the semanticannotation of data from various resources and extracts important information fromthe textual description. The tools above help to create and test semantic annotationsat the level available to the user.


Knowledge Hosting

Knowledge can be represented in different ways. In the context of our survey,we focus on knowledge which is present in RDF format. This means the data isrepresented in the form of subject-predicate-object triples, using an ontology todescribe things semantically. For example, the information Fritz Phantom lives inInnsbruck and is born in 1984 contains three facts and can be expressed in a tripleform in the following way:

Listing 9 RDF example Fritz Phantom"Fritz Phantom" is a Person"Fritz Phantom" lives in Innsbruck"Fritz Phantom" is born in 1984

If we want to describe those three facts with the ontology of schema.org, it wouldlook like this:

Listing 10 Schema.org example Fritz Phantomhttp://fritz.phantom.com rdf:type schema:Personhttp://fritz.phantom.com schema:homeLocation http://innsbruck.gv.athttp://fritz.phantom.com schema:birthDate "1984"

There are two different popular ways to store this data: either in a NoSQL orspecifically document database, serialized in JSON-LD, or natively in RDF form, ina graph database.

JSON-LD stands for JSON for Linked Data and is a backward-compatibleextension of JSON, the JavaScript Object Notation (Sporny et al. 2014). NoSQLdatabases such as MongoDB31 store JSON natively; hence using a document storeto store knowledge is very effective and at the same time very convenient, due tothe good support of document stores by web frameworks. Even though storage andretrieval work seamlessly, querying over JSON-LD files is not supported natively.To circumvent this limitation, manual implementation work is required.

The second way to store knowledge is in the native RDF format, into a graphdatabase supporting RDF. A typical example for such a database is GraphDB.32 Toquery the data, Knowledge Graphs typically offer a SPARQL interface over whichall read-write operations and extensive reasoning requests can be performed. Theresults of SPARQL requests are typically in Turtle format and can be transformedinto any other RDF serialization. Compared to document stores, graph databasesare much better for querying and reasoning, but on the other hand, they are alsomuch more expensive to purchase and maintain and have, de facto, no integrationwith any popular web framework. Both methods have their preferred scenario ofusage. NoSQL is more suitable in the context of web site annotation, while graphdatabases are more suitable in the context of Linked (Open) Data publication (seesection “Semantic Web of Data: Linked Open Data”). More information on RDFstorage technology can be found in Ma et al. (2004), Angles and Gutierrez (2005),Stegmaier et al. (2009), and Faye et al. (2012).

26 D. Fensel et al.

Knowledge Curation

Knowledge is an important asset in all enterprises. For instance, Gutiérrez-Cuellarand Gómez-Pérez (2014) propose the HAVAS Knowledge Graph that aims tocollect information from start-up, innovators, and tech companies; Amato et al.(2017) present KIRA that gathers data from user, social media, and multimediadata sources; Quimbaya et al. (2014) propose EXEMED that is a knowledge basefor structuring clinical guidelines; Achichi et al. (2018) propose DOREMUS, aknowledge graph of music works and events; and an Amazon product knowledgegraph is led by Dong (2018). Knowledge is being continuously gathered andmaintained in order to serve several purposes, from providing a common unifiedview on all data resources of the enterprises to powering their applications. Forinstance, the large technology companies – including Microsoft, Facebook, Google,and many more – have knowledge graphs and have invested in their curation withthe purpose of making all their web-scale services better (Pan et al. 2017). In theprevious sections, we examined different methods and tools in which knowledgecan be modeled and how knowledge graphs can be built and hosted. Building andhosting a knowledge graph is one thing. Turning them into a useful resource forproblem-solving requires additional effort. In this context, knowledge curation playsa key role. In short, knowledge curation is about (1) assessing the quality of theknowledge graph, i.e., knowledge assessing; (2) improving the correctness of theknowledge graph, i.e., knowledge cleaning; and (3) improving the completeness ofthe knowledge graph, i.e., knowledge enrichment.

Knowledge AssessmentKnowledge Assessment describes and defines the process of assessing the quality ofa Knowledge Graph. The goal is to measure the usefulness of a Knowledge Graphconsidering two major quality dimensions, namely, its correctness and complete-ness. Knowledge Graph assessment can differ along different quality dimensions(Batini et al. 2009; Zaveri et al. 2013). For example, Paulheim and Bizer (2013)tackle the identification of missing instance assertions, Fürber and Hepp (2010a)identify wrong and missing property value assertions, and Lertvittayakumjorn et al.(2017) address the identification of wrong property value assertions. The approachpresented by Mendes et al. (2012) defines the additional quality assessmentmethods.

Knowledge CleaningKnowledge Cleaning is about identifying and correcting the wrong assertions in aKnowledge Graph by deleting or modifying them. The goal of Knowledge Cleaningis to improve the correctness of a knowledge graph. The following tasks are relevantfor Knowledge Cleaning:

– Detection and correction of wrong instance assertions– Detection and correction of wrong property value assertions– Detection and correction of wrong equality assertions


The achievement of those tasks heavily relies on the employed tools. For example,SDType (Paulheim and Bizer 2013) detects wrong instance assertions, SPIN (Fürberand Hepp 2010b) identifies functional dependency violations, LOD Laundromat(Beek et al. 2014) allows for the detection and correction of syntax errors,SDValidate (Paulheim and Bizer 2014) partially identifies wrong property valueassertions, KATARA (Chu et al. 2015) identifies and corrects wrong propertynames, and HoloClean (Rekatsinas et al. 2017) can be used for detecting andcorrecting wrong property value assertions.

Knowledge EnrichmentKnowledge Enrichment identifies and adds new assertions into a Knowledge Graph.The goal of knowledge enrichment is to improve the completeness of a knowledgegraph. The following tasks are relevant for Knowledge Enrichment:

– Identifying and resolving duplicates by adding missing instance assertions andmissing equality assertions.

– Resolving conflicting property value assertions by adding or deleting missingproperty value assertions.

Several methods and tools have been developed to address entity resolution andthe related problems of conflicting property value assertions. For instance, foridentifying duplicates, some authors use methods and techniques based on stringsimilarity measures (Winkler 2006), association rule mining (Hipp et al. 2000),topic modeling (Sleeman et al. 2015), support vector machine (Sleeman andFinin 2013), property-based (Hogan et al. 2007), crowd-sourced data (Getoor andMachanavajjhala 2013), and graph-oriented (Korula and Lattanzi 2014); and forresolving duplicates, there are various tools such as Silk (Volz et al. 2009), SERIMI(Araújo et al. 2011), Legato (Amato et al. 2017), or Duke (Garshol and Borge2013). Regarding the resolution of conflicting property value assertions, this canbe tackled using Sieve (Mendes et al. 2012), FAGI (Giannopoulos et al. 2014), orODCleanStore (Michelfeit et al. 2012).

Knowledge Deployment

One of the earliest applications of knowledge graphs was provided by Google whobegan in 2012 to develop the so-called Google knowledge graph (Singhal 2012),which should contain significant aspects of human knowledge found semanticallyannotated on the web or in other data sources. Since then, a multitude of knowledgegraphs have been developed (cf. Paulheim 2017 and others) including AirbnbKnowledge Graph,33 Bing Knowledge Graph34 (previously called Microsoft’sSatori35), Cyc/OpenCyc36 (cf. Lenat 1995; Lenat and Guha 1989), datacom-mons.org,37 DBpedia38 extracted from Wikipedia (cf. Auer et al. 2007; Lehmannet al. 2015), Facebook’s Entities Graph,39 Freebase40 (see Bollacker et al. 2008)meanwhile close, bought by Google and also incrementally included in Wikidata,

28 D. Fensel et al.

Table 1 Numerical Overview of some Knowledge Graphs, taken from Paulheim (2017)

Name Instances Facts Types Relations

DBpedia (English) 4,806,150 176,043,129 735 2,813

YAGO 4,595,906 25,946,870 488,469 77

Freebase 49,947,845 3,041,722,635 26,507 37,781

Wikidata 15,602,060 65,993,797 23,157 1,673

NELL 2,006,896 432,845 285 425

OpenCyc 118,499 2,413,894 45,153 18,526

Google’s Knowledge Graph 570,000,000 18,000,000,000 1,500 35,000

Google’s Knowledge Vault 45,000,000 271,000,000 1,100 4,469

Yahoo! Knowledge Graph 3,443,743 1,391,054,990 250 800

Google’s Knowledge Graph,41 kbpedia42 (see Bergman 2018), Knowledge Vault43

(see Dong et al. 2014), NELL44 and45 (Carlson et al. 2010), Wikidata46 (Vrandecicand Krötzsch 2014), YAGO47 (Suchanek et al. 2007), (Hoffart et al. 2013), extractedfrom Wikipedia plus wordnet,48 and Yahoo!’s Knowledge Graph49; see Blanco et al.(2013). Table 1 provides a survey of the size of some of the previously mentionedKnowledge Graphs where this information is made publicly available.

These Knowledge Graphs, especially when based on schema.org, play anincreasingly important role for web-based information search. Search is in factone core application of knowledge graphs. It has evolved over time as the webhas changed. From a system based on publishing information intended for humanuser consumption (the classical web) to a system where machines can understandand consume the content (Semantic Web), the web is nowadays changing into aweb for bots. Search and search engines have evolved along the same path, frombeing very effective indexes of the web based on syntax and statistical analysisto becoming query-answering engines (see Guha et al. 2003; Harth et al. 2007).This was only achieved by using semantic annotations of website content, data, andservices through the use of de facto standards such as schema.org.

Chatbots and Intelligent Personal Assistants have become very popular in thelast couple of years. They require structured, machine-processable data, content,and services. Alexa, Bixby, Cortana, Facebook messenger, Google Assistant, Siri,and others provide personalized and (spoken) message-based access to information.Knowledge Graphs can be used to improve such dialogue systems in two majorways: (1) to power the language understanding part of the dialogue system and(2) to react to the conversations and provide additional interactions, information,and recommendations to the user engaged in conversations with the dialoguesystem. When it comes to supporting the language understanding part of thedialogue system, the goal is to use the Knowledge Graphs to provide training datafor the Natural Language Understanding. We can automatically ingest from theKnowledge Graph training data for the entity recognition task (e.g., Vienna is aCity) and provide (semi-)automatically generated intents and example questions.Based on the Knowledge Graph structure, we can generate, on the one hand, entities


and synonyms, and on the other hand intents needed in the Natural LanguageUnderstanding service based on the entities, specifically the relations betweenthese entities in the Knowledge Graph. Furthermore, we can use ontology2textapproaches to generate example questions that can be used to train the NaturalLanguage Understanding service. The second direction in which Knowledge Graphscan be used to improve dialogue systems is to react to conversations and provideadditional interactions, information, and recommendations to the user. Using theknowledge from the Knowledge Graph, the dialogue system can elaborate on thetopic of discussion and provide additional interesting facts. A Knowledge Graphcan also be used to improve the handling of the conversation context. Finally, aKnowledge Graph can also be used to refine the search for products or services in adialogue system. In case the dialogue system cannot answer the given question, theKnowledge Graph can be used to obtain more information.

Use Cases in E-Tourism

We can distinguish between proprietary and public Knowledge Graphs. For exam-ple, the Google Knowledge Graph is an internal resource of Google that improves itsanswering quality. Alternatively, a public Knowledge Graph can be the basis of anecosystem of bots that search it for products and services. Obviously, there are manyvariations and combinations of these two principles possible. In the following, wediscuss pilots for developing both internal Knowledge Graphs and open and publicones in the tourism domain.

Touristic Chatbots and Intelligent Personal Assistants

The Gartner hype cycle50 for emerging technologies (August 2018) shows bothknowledge graphs and conversational AI in the innovation trigger. Marketsand-Markets51 forecasts the global conversational AI market size to grow from USD4.2 billion in 2019 to USD 15.7 billion by 2024, at a compound annual growthrate (CAGR) of 30.2% during the forecast period (2019–2024). The major growthdrivers for the market include the increasing demand for AI-powered customer sup-port services, omni-channel deployment, and reduced chatbot development costs.As we describe in the following use case, applications of Knowledge Graphs area complementary technology for conversational platforms to scale the automationof conversations of chatbots and voice assistant at reduced costs. The growth forconversational AI is due to the evolving usage of chatbots for content marketingactivities, such as digital marketing and advertising. The technological capabilities,individuality, and customization are the main features accelerating market growth.Chatbots are there to assist, interact, and engage with customers, and they offerpersonalized marketing capabilities.52

Voice-to-text understanding has recently achieved a very high accuracy and con-tinues to improve. Nevertheless, use cases of current chatbots and voice assistants

30 D. Fensel et al.

Fig. 10 Typical Dialogue with current chatbots and voice assistants

Fig. 11 The inner process of a Knowledge-centered chatbots and voice assistants

are still basic and focus on simple question and answer solutions. A dialogue withan Alexa Box or Google Home quite often ends in a “Sorry, I don’t know” due tothe lack of knowledge these devices have. The reason for this is that the naturallanguage solutions of such devices lack knowledge of entities, e.g., restaurant androast pork (as demonstrated in the example in Fig. 10) and cannot achieve the goalsof the questions.

To support the chatbot and voice assistant type of scenarios introduced before,we need to design, implement, and deploy a knowledge-centered solution thatwill enable conversational interfaces to engage in human-like dialogues. Figure 11depicts the inner process of such a solution for chatbots and voice assistants.

At first, the natural language input of a user, in written or spoken form, undergoesa natural language understanding step (understand 1.), in which the user intent,together with parameters, is identified. The intent then needs to be resolved toan action that typically translates in a number of queries (map 2.) that can thenbe executed (query 3.) against the integrated large volumes of heterogeneous,


Fig. 12 Using Knowledge Graphs to make chatbots and voice assistants (e.g., Alexa) smarter

distributed, dynamic, and potentially (i.e., almost certainly) inconsistent statementsin order to identify the relevant knowledge parts necessary to generate the useranswer in natural language (NLG – natural language generation 4.) as text or voice.

Let us revisit our use case and see how Knowledge Graphs can enable chatbotsand voice assistants to understand the goal the human users express in naturallanguage requests. Figure 12 illustrates the different steps of the process, fromunderstanding the user request to generating and executing the query against theKnowledge Graph to generating the answer for the user. With a Touristic KnowledgeGraph in place that includes touristic entities such as restaurants and offers fromthese restaurant (e.g., roast pork), as well as actions related to these entities thatcan be performed (e.g., booking a table), intents and parameters can be derived.For example, an intent TableReservation for entities of type Restaurant can begenerated. Restaurants, and in general organizations, can be connected in theKnowledge Graphs to other entities of the type Offer (e.g., roast pork offers).Furthermore, the Knowledge Graph can be used to improve the understanding ofthe natural language understanding (NLU) by pushing entities from the KnowledgeGraph (e.g., Hofbräu Bierhaus NYC) to the NLU or by generating examplequestions for the intents. The Knowledge Graph can be also used to generate therules that restrict the view/access to the Knowledge Graph depending on the usecases. Such rules, together with the intent and parameters extracted by the NLU, areused to generate the queries to be executed against the Knowledge Graph. Last butnot least, the Knowledge Graph can be used to generate templates for the answers,the textual answers, or follow-up questions to run the dialogues.

Chatbots and voice assistants have started to play an increasing role in customercommunication for many businesses in various verticals. Especially in tourism, theyare proving to offer an increasing number of benefits in terms of convenience,availability, and fast access to information delivery and customer support throughthe entire customer journey.53 In the dreaming and planning phase, hotels andDestination Management Organizations (DMOs) can provide information throughchatbots and voice assistants about the hotel and/or the region, the surroundings,

32 D. Fensel et al.

and weather conditions to potential guests. In the booking phase, from bookingthe hotel and transport to buying connected services, e.g., ski tickets, the entireprocess becomes much simpler and efficient by using natural language. Finally, inthe experience phase, chatbots and voice assistants can also announce special offersor events. All requested information and processes are instantly available 24/7/365.For hotel guests in particular, the stay experience can be enriched by providing themwith access to hotel services and beyond. Recently, Amazon launched a program forhotel operators54 that allows guests to request room service, ask for housekeeping,configure the temperature and lights in the hotel room, set wake-up calls, and evenconnect their accounts to listen to their own music and audio-books. Last but notleast, customer support questions regarding rooms, equipment, additional services,and more are answered in a fully automated way. On the one hand, it can be arguedthat similar functionalities are available in mobile apps, but the major drawbackof these apps is that each one of them focuses on different aspects, and time isrequired to learn how each app works. Chatbots and voice assistants, on the otherhand, provide easier means to access the same functionalities by using the mostnatural way in which humans interact, i.e., natural language (as voice or writtentext).

Touristic chatbots and voice assistants are thus expected to answer questions andsatisfy commands of a different nature, for example, “What’s the most popularattraction in the city?”, “What events are happening in the coming weekend?”,“What’s the snow height?”, “Book me a table tonight for 2 people in a Tyroleanrestaurant”, “I’m looking for a bike ride that is difficult and offers huts on theway”, etc. To properly answer all these types of questions and perform tasks such asbooking, chatbots and voice assistants need machine-processable (semantic) annota-tions of content, data, and services. They need structures that encode the knowledgeabout the tourism domain, in terms of entities and relations between them, in amachine-processable form. Knowledge Graphs are such structures providing thetechnical means to integrate various heterogeneous touristic information sourcesabout accommodations, points of interests, events, sports activity locations, etc.With the help of Knowledge Graphs, not only can simple question-answering tasksbe supported but rather complex conversations/dialogues can be too.

Applying the principles, methods, and tools introduced in the previous sections,we have built a Knowledge Graph for tourism that integrates multiple sources ofcontent, data, and services from various providers, both

– closed: Feratel,55 Outdooractive,56 intermaps,57 General Solutions,58

Verkehrsauskunft Österreich59

– open: WikiData,60 DBpedia,38.OpenStreetMap61 and GeoNames62

The resulting Touristic Knowledge Graph powers several chatbots and voiceassistants of touristic regions in Tyrol, Austria.

1. The Seefeld pilot63 focuses on integrating only closed data sources, namely,from Feratel, Intermaps, Outdooractive, and General Solutions. The use case is


for the tourist region Olympiaregion Seefeld. For this use case, we also focuson question-answering for more advanced (compound questions), for instance,“Where can I have traditional Tyrolean food when going cross country skiing?”.

2. The Serfaus-Fiss-Ladis pilot64 focuses on integrating both close datasources, namely, Feratel, Intermaps, Outdooractive, General Solutions, andVerkehrsauskunft Österreich as well as open data sources, namely, Wikidata andDBpedia. The Serfaus-Fiss-Ladis tourist region envisions that users cannot onlychat about the specific tourism data but also inquire about common knowledgeof the region. The conversational interface is able to handle questions whichcombine the closed and open datasets, for instance, “How many inhabitants doesSerfaus have?” or “Traffic information from Serfaus to Via Claudia Augusta?”

Common to these pilots and use cases is the need to integrate data from multipleheterogeneous static and dynamic sources, for which we need to track provenance(e.g., data owner, temporal validity, or the integration process) and maintain onecommon evolving schema. Using knowledge cleaning and enrichment, we alsoensure a certain level of quality of the touristic knowledge. The ultimate aim is tooptimize conversational interfaces based on Knowledge Graphs by providing a richintent and entity management (e.g., automated NLU training), question-answeringover the Knowledge Graph, and supporting advanced dialogues such as guiding auser through actions or recommendations or follow-up conversations. These pilotshave been implemented and used to test and validate the usage of Knowledge Graphsto enable the better understanding of natural language dialogues and knowledgeaccess for touristic chatbots and voice assistants.

Open Touristic Knowledge Graph

We have built the Tirol Knowledge Graph (TKG) as a five-star linked open data setpublished in a graph database providing a SPARQL endpoint, Kärle et al. (2018),for the provision of touristic data of Tyrol, Austria. The TKG currently contains dataabout touristic infrastructure, such as accommodation businesses, restaurants, pointsof interests, events, and recipes. The data of the TKG fall under three categories ofdata: static data is information which is rarely changing, such as addresses of hotels,descriptions of points of interests, and suchlike. Dynamic data is fast-changinginformation, such as availabilities and prices. Active data describes actions that canbe executed, for example, the description of a purchase or reservation Web API thatcan be accessed through the TKG GraphDB platform.65

The data is collected either through the crawling of websites or mappingsfrom proprietary data sources into the Knowledge Graph, which uses schema.orgas ontology. Therefore, only websites containing schema.org annotated data areconsidered, and data sources are always mapped to schema.org before being storedin the TKG. The crawler is implemented inside the semantify.it annotation platform(Kärle et al. 2017), called broker.semantify.it. Based on a list of URLs of touristicwebsites, the data gets collected periodically and is then stored in the graph. The

34 D. Fensel et al.

Table 2 Top 10 entities usedin the TKG

Entity Count

schema:Thing 453,841,147

schema:CreativeWork 175,787,490

schema:MediaObject 175,746,110

http://purl.org/dc/dcmitype/Image 175,735,868

schema:ImageObject 175,735,868

schema:Intangible 172,124,244

schema:StructuredValue 155,482,666

schema:Place 60,996,190

schema:ContactPoint 53,155,166

schema:PostalAddress 51,706,023

mapping is provided for different data sources such as Feratel,55.General Solutions,58.

Infomax,66 Tomas,67 etc. (Panasiuk et al. 2018a, b). The data are mostly retrievedthrough SOAP or REST APIs and are originally provided in XML or JSON format.For fetching these data, translating it to schema.org, and storing it in the KnowledgeGraph periodically, wrappers are implemented inside semantify.it that are executedperiodically. The mapping is either implemented programmatically in NodeJS ordone through the mapping language RML (Dimou et al. 2014).

As of May 4, 2019, the TKG contained around 7.5 billion statements, of which55% are explicit and 45% are inferred. Every day, the Knowledge Graph grows byaround 8 million statements. The data is held in around 2000 subgraphs, where everysubgraph represents one import process per data source (see section “AnnotationLanguages”, data provenance). TKG contains more than 200 entity types; the mostfrequently used ones are shown in Table 2.

To demonstrate the possibilities of the TKG and to evaluate its usability, we builtseveral pilots.

1. Chatbot-driven room booking: among the crawled websites are many customersof the Internet booking engine development company Easybooking.68 Thefeatures, identifying a website as a customer of Easybooking, inside the sourcecode are known. The booking API structure of Easybooking is also known.Therefore, we decided to develop an Alexa skill that enables voice-drivenbooking of Easybooking hotels through the TKG. If the showcase skill is askedfor a certain hotel, it sends a requests to a web hook. The result, a list of availablehotel offers, is sent back to the skill and read to the user. The list also containsannotated API descriptions for the booking API, so if the user decides on anoffer, a booking can be executed through a voice command.

2. Show case dialogue system: as described in Simsek and Fensel (2018), Panasiuket al. (2018c), we built two dialogue systems that fetch their data from the graph.One generically answers touristic topics such as hiking or opening hours. Theother (Simsek and Fensel 2018) goes one step further and conducts genericdialogs solely based on data taken from the Knowledge Graph.

http://purl.org/dc/dcmitype/Image


3. Time series analysis of prices in touristic regions: since all the prices of offers,if available, are stored permanently, a time series analysis can be conducted. Wecompare the price development of two touristic regions over a period of time. Thetime series analysis works perfectly with Knowledge Graphs and is a promisingapplication of them in tourism.

Despite the TKG, there are other initiatives regarding touristic Knowledge Graphsworth mentioning in this context. One of these initiatives is the DACH-KG.DACH is an acronym for the German-speaking regions of Germany (D), Austria(A), Switzerland (CH), and the region of South Tyrol, whereas KG stands forKnowledge Graph. Key representatives of the touristic domain and academics fromsaid countries and from Italy are working together. To achieve this, they are workingnot only on aligning data sources technically but also on extending the expressivityof the ontology of their choice, schema.org. The achievements and progress of thisworking group are described in a living paper69 and in two blog articles.70,71 Asimilar goal is pursued by another initiative, the French DATAtourisme.72 Here, aunion of French DMOs and regional tourism boards is working on a 5* LinkedOpen Data (see section “Semantic Web of Data: Linked Open Data”) set, to cateras the official source for structured touristic data in France. Data from more than 40systems are aggregated, described by a custom-made ontology, and provided overthe project’s website and the corresponding government website.73 An academicinitiative around touristic knowledge graphs is the TourismKG workshop series.74

Conclusions

The Semantic Web began more than 20 years ago as a means to add machine-processable semantics to the web. In the meantime, it has become a fairly maturearea of research and in the past 5 years has begun to have a significant impact onhow information is presented on the web. Obviously, search engines were fairlyslow in taking up and exploiting its potential for providing intelligent access to webresources beyond simple search result listings. However, it is increasingly becominga must-have for content, data, and service providers to enrich their online resourceswith semantic markup such as schema.org. Schema.org is a rather simple and limitedapproach, but we expect richer approaches to be adopted soon (given the usual delayin take-up by big industry). This is mission accomplished, except for the fact thatthis success is in the process of opening up an even more challenging and demandingresearch area, the so-called Knowledge Graphs.

In the context of the evolving Knowledge Graph technology, we aimed to provideanswers to three important questions: What are Knowledge Graphs, how are theybuilt, and in what sense are they important? We provided a number of approachesfor constructing, hosting, curating, and deploying Knowledge Graphs and showedtheir potential usage for dialog-based information access on the Internet, usage thatmay revolutionize humans’ information access. We described applications in theareas of e-Tourism as a cornerstone for future e-Marketing and e-Commerce.

36 D. Fensel et al.

In the future, we expect Knowledge Graphs to rapidly grow to trillions of factsand beyond. This introduces harsh requirements on the methods that are able tohandle them. Even in the optimistic case, Paulheim (2018a) estimates the relatedcosts at billions of dollars. Keeping scale without a cost explosion is an obviousrequirement for the success of the Knowledge Graph approach. This may require thereturn of more traditional AI techniques where large amounts of facts are attemptedto be captured through the elegance of simple rules and axioms (like a picturethat can express more than a thousand words). Therefore, building up meaningfulTboxes on top of existing Knowledge Graphs may be an interesting avenue toinvestigate (see Töpper et al. 2012; Socher et al. 2013; Galárraga et al. 2013, 2015;Paulheim 2018b).

Expected Future Developments

To tackle the huge size of future knowledge graphs, we expect to operate not onthe knowledge graph as a single unit but on smaller subgraphs. These subgraphswill then be used to operate efficiently and effectively (by reasoning over them andcarrying out Knowledge Curation). Not only is the size of the knowledge grapha problem, but also consumers having different points of view present challenges.Therefore, it must be possible to apply different constraints for different consumersas a single set of constraints is infeasible. For example, a knowledge graph storinginformation about people and their relations to each other contains a person withmultiple spouses. Based on the points of view, different conclusions can be drawn.A point of view that allows polygamy may not constrain the cardinality of the spouseproperty but in another point of view that information may be identified as an error.By extracting subgraphs from the huge knowledge graph and directly operating onthem, it is possible to define and apply different constraints for different consumers.

We also expect that in the future knowledge graphs will be used together/tosupport machine learning algorithms. A very prominent and important example isautonomous driving. In May 2016, Joshua Brown was killed by his car because itsautopilot mixed up a very long truck (18- wheeler) with a traffic sign. If the car hadbeen connected to a knowledge graph containing traffic data, his life could have beensaved. The autopilot could have used the information from the knowledge graph todetect that there is no traffic sign and stop the car.

Not only can knowledge graphs be used to support machine learning algorithms,but also machine learning algorithms can be used to create, correct, and enrichknowledge graphs. For the creation of knowledge graphs, heuristics are frequentlyused. Those heuristics could be implemented by using techniques based on machinelearning methods. In addition, for example, adding missing type information can beconsidered as a hierarchical multi-label classification problem. For error detection,we can use machine learning algorithms, which use relations in a knowledgegraph, as positive training examples and create negative examples to detect wrongstatements.


Cross-References

�Accessible E-Tourism�AI and the Travel Experience�Big Data�Data Mining and Predictive Analytics�Data Privacy and the Travel Sector�Data Science (Open Data), Data Interoperability�Drivers of E-Tourism�Electronic Data Interchange and Standardisation�E-Tourism: A DMO Perspective�E-Tourism Research: A Review�Events and ICT� Impact of AI on the Hospitality Industry� Interactive and Context-Aware Recommendation in Tourism� IT and Wellbeing in Travel and Tourism�Knowledge Management in Tourism Organizations (DMOs, Hotels, etc.)�Online Marketing in Tourism�Ontology Building�Open and Commercial (Big) Data in Tourism� Privacy & Ethical Issues in Online Interactions�Recommender Systems� Strategic Use of IT in Tourism�Technology and Future of Travel�The Dark Side of Tourism and the Digital Transformation�The Field of e-Tourism: An Informatics Perspective�Travel Information Search�User Modeling in E-Tourism�Web Information Retrieval & Search�Web Technologies & Web Applications

Notes

1. Schema.org https://www.schema.org, accessed: 02 Jun 2019.2. Thesaurus https://www.thesaurus.com, accessed: 05 Jun 2019.3. Microformats2 http://microformats.org/wiki/microformats2, accessed: 14 Jun 2019.4. Microdata https://www.w3.org/TR/microdata/, accessed: 14 Jun 2019.5. RDFa https://www.w3.org/TR/xhtml-rdfa-primer/, accessed: 14 Jun 2019.6. JSON-LD https://www.w3.org/TR/json-ld/, accessed: 14 Jun 2019.7. JSON https://www.json.org/, accessed: 14 Jun 2019.8. JSON-LD example https://www.w3.org/TR/json-ld/#basic-concepts, accessed: 11 Jul 2019.9. JSON-LD Syntax https://w3c.github.io/json-ld-syntax/, accessed: 14 Jun 2019.

10. Dublin Core http://dublincore.org/specifications/dublin-core/dcmi-terms/, accessed: 14 Jun2019.

11. Friend of a friend http://xmlns.com/foaf/spec/, accessed: 14 Jun 2019.12. GoodRelations http://www.heppnetz.de/ontologies/goodrelations/v1, accessed: 14 Jun 2019.

https://doi.org/Accessible E-Tourism

https://doi.org/AI and the Travel Experience

https://doi.org/Big Data

https://doi.org/Data Mining and Predictive Analytics

https://doi.org/Data Privacy and the Travel Sector

https://doi.org/Data Science (Open Data), Data Interoperability

https://doi.org/Drivers of E-Tourism

https://doi.org/Electronic Data Interchange and Standardisation

https://doi.org/E-Tourism: A DMO Perspective

https://doi.org/E-Tourism Research: A Review

https://doi.org/Events and ICT

https://doi.org/Impact of AI on the Hospitality Industry

https://doi.org/Interactive and Context-Aware Recommendation in Tourism

https://doi.org/IT and Wellbeing in Travel and Tourism

https://doi.org/Knowledge Management in Tourism Organizations (DMOs, Hotels, etc.)

https://doi.org/Online Marketing in Tourism

https://doi.org/Ontology Building

https://doi.org/Open and Commercial (Big) Data in Tourism

https://doi.org/Privacy & Ethical Issues in Online Interactions

https://doi.org/Recommender Systems

https://doi.org/Strategic Use of IT in Tourism

https://doi.org/Technology and Future of Travel

https://doi.org/The Dark Side of Tourism and the Digital Transformation

https://doi.org/The Field of e-Tourism: An Informatics Perspective

https://doi.org/Travel Information Search

https://doi.org/User Modeling in E-Tourism

https://doi.org/Web Information Retrieval & Search

https://doi.org/Web Technologies & Web Applications

https://www.schema.org

https://www.thesaurus.com

http://microformats.org/wiki/microformats2



https://www.w3.org/TR/json-ld/

https://www.json.org/

https://www.w3.org/TR/json-ld/#basic-concepts

https://w3c.github.io/json-ld-syntax/

http://dublincore.org/specifications/dublin-core/dcmi-terms/

http://xmlns.com/foaf/spec/

http://www.heppnetz.de/ontologies/goodrelations/v1

38 D. Fensel et al.

13. Tourismus der Zukunft https://www.tourismuszukunft.de, accessed: 28 Nov 2019.14. DACH-KG Schema.org extension https://github.com/STIInnsbruck/dachkg-schema,

accessed: 02 Jun 2019.15. Simple Object Access Protocol https://en.wikipedia.org/wiki/SOAP, accessed: 14 Jun 2019.16. Representational state transfer https://en.wikipedia.org/wiki/Representational_state_

transfer, accessed: 14 Jun 2019.17. Web Service Description Language https://en.wikipedia.org/wiki/Web_Services_

Description_Language, accessed: 14 Jun 2019.18. Semantic Web Services Framework https://www.w3.org/Submission/SWSF/, accessed: 14

Jun 2019.19. OpenAPI Specification https://github.com/SmartAPI/smartAPI-Specification/blob/

OpenAPI.next/versions/3.0.0.md accessed: 14 Jun 2019.20. Schema.org Actions https://schema.org/docs/actions.html, accessed: 14 Jun 2019.21. Graph (discrete mathematics) https://en.wikipedia.org/wiki/Graph_(discrete_mathematics),

accessed: 14 Jun 2019.22. Semantify.it https://semantify.it/, accessed: 14 Jun 2019.23. Document Object Model https://www.w3.org/DOM/, accessed: 14 Jun 2019.24. GATE https://gate.ac.uk/, accessed: 14 Jun 2019.25. Apache Open NLP https://opennlp.apache.org/, accessed: 14 Jun 2019.26. RapidMiner https://rapidminer.com/, accessed: 14 Jun 2019.27. Tripliser http://daverog.github.io/tripliser/, accessed: 14 Jun 2019.28. GRDDL https://www.w3.org/TR/grddl/, accessed: 14 Jun 2019.29. Virtuoso Sponger http://vos.openlinksw.com/owiki/wiki/VOS/VirtSponger, accessed: 14 Jun

2019.30. R2RML https://www.w3.org/TR/r2rml/, accessed: 14 Jun 2019.31. MongoDB https://www.mongodb.com/de, accessed: 14 Jun 2019.32. GraphDB http://graphdb.ontotext.com/, accessed: 14 Jun 2019.33. Airbnb Knowledge Graph https://medium.com/airbnb-engineering/scaling-knowledge-

access-and-retrieval-at-airbnb-665b6ba21e95, accessed: 14 Jun 2019.34. Bing Knowledge Graph https://blogs.bing.com/search-quality-insights/2017-07/bring-rich-

knowledge-of-people-places-things-and-local-businesses-to-your-apps, accessed: 14 Jun2019.

35. Microsoft’s Satori http://blogs.bing.com/search/2013/03/21/understand-your-world-with-bing/, accessed: 14 Jun 2019.

36. Cyc/OpenCyc http://www.cyc.com/, accessed: 14 Jun 2019.37. datacommons.org http://datacommons.org/, accessed: 14 Jun 2019.38. DBpedia http://dbpedia.org/, accessed: 14 Jun 2019.39. Facebook’s Entities Graph http://www.facebook.com/notes/facebook-engineering/under-

the-hood-the-entities-graph/10151490531588920, accessed: 14 Jun 2019.40. FreeBase http://www.freebase.com/, accessed: 14 Jun 2019.41. Google’s Knowledge Graph https://developers.google.com/knowledge-graph/, accessed: 14

Jun 2019.42. kbpedia http://www.kbpedia.org/, accessed: 14 Jun 2019.43. Knowledge Vault https://ai.google/research/pubs/pub45634, accessed: 14 Jun 2019.44. NELL http://rtw.ml.cmu.edu/rtw/resources, accessed: 14 Jun 2019.45. NELL http://rtw.ml.cmu.edu/rtw/kbbrowser/, accessed: 14 Jun 2019.46. Wikidata Main Page https://www.wikidata.org/wiki/Wikidata:Main_Page, accessed: 14 Jun

2019.47. YAGO https://www.mpi-inf.mpg.de/departments/databases-and-information-systems/

research/yagonaga/yago/downloads/, accessed: 14 Jun 2019.48. WordNet https://wordnet.princeton.edu/, accessed: 14 Jun 2019.49. Yahoo’s Knowledge Graph https://www.slideshare.net/NicolasTorzec/the-yahoo-

knowledge-graph, accessed: 14 Jun 2019.

https://www.tourismuszukunft.de

https://github.com/STIInnsbruck/dachkg-schema

https://en.wikipedia.org/wiki/SOAP

https://en.wikipedia.org/wiki/Representational_state_transfer

https://en.wikipedia.org/wiki/Representational_state_transfer

https://en.wikipedia.org/wiki/Web_Services_Description_Language

https://en.wikipedia.org/wiki/Web_Services_Description_Language

https://www.w3.org/Submission/SWSF/

https://github.com/SmartAPI/smartAPI-Specification/blob/OpenAPI.next/versions/3.0.0.md

https://github.com/SmartAPI/smartAPI-Specification/blob/OpenAPI.next/versions/3.0.0.md

https://schema.org/docs/actions.html

https://en.wikipedia.org/wiki/Graph_(discrete_mathematics)

https://semantify.it/

https://www.w3.org/DOM/

https://gate.ac.uk/

https://opennlp.apache.org/

https://rapidminer.com/

http://daverog.github.io/tripliser/

https://www.w3.org/TR/grddl/

http://vos.openlinksw.com/owiki/wiki/VOS/VirtSponger

https://www.w3.org/TR/r2rml/

https://www.mongodb.com/de

http://graphdb.ontotext.com/

https://medium.com/airbnb-engineering/scaling-knowledge-access-and-retrieval-at-airbnb-665b6ba21e95

https://medium.com/airbnb-engineering/scaling-knowledge-access-and-retrieval-at-airbnb-665b6ba21e95

https://blogs.bing.com/search-quality-insights/2017-07/bring-rich-knowledge-of-people-places-things-and-local-businesses-to-your-apps

https://blogs.bing.com/search-quality-insights/2017-07/bring-rich-knowledge-of-people-places-things-and-local-businesses-to-your-apps

http://blogs.bing.com/search/2013/03/21/understand-your-world-with-bing/

http://blogs.bing.com/search/2013/03/21/understand-your-world-with-bing/

http://www.cyc.com/

http://datacommons.org/

http://dbpedia.org/

http://www.facebook.com/notes/facebook-engineering/under-the-hood-the-entities-graph/10151490531588920

http://www.facebook.com/notes/facebook-engineering/under-the-hood-the-entities-graph/10151490531588920

http://www.freebase.com/

https://developers.google.com/knowledge-graph/

http://www.kbpedia.org/

https://ai.google/research/pubs/pub45634

http://rtw.ml.cmu.edu/rtw/resources

http://rtw.ml.cmu.edu/rtw/kbbrowser/

https://www.wikidata.org/wiki/Wikidata:Main_Page

https://www.mpi-inf.mpg.de/departments/databases-and-information-systems/research/yagonaga/yago/downloads/

https://www.mpi-inf.mpg.de/departments/databases-and-information-systems/research/yagonaga/yago/downloads/

https://wordnet.princeton.edu/

https://www.slideshare.net/NicolasTorzec/the-yahoo-knowledge-graph

https://www.slideshare.net/NicolasTorzec/the-yahoo-knowledge-graph


50. Gartner Hype Cycle https://gartner.com, accessed: 14 Jun 2019.51. MarketsandMarkets https://bit.ly/2IAiHqw, accessed: 14 Jun 2019.52. Personalized marketing capabilities https://www.sdcexec.com/software-technology/news/

21011880/chatbot-market-to-grow-at-31-percent-cagr-from-2018-to-2024, accessed: 14Jun 2019.

53. Customer journey https://tourismeschool.com/customer-journey-mapping-tourism-brands/,accessed: 14 Jun 2019.

54. Amazon Hotel System https://techcrunch.com/2018/06/19/amazon-launches-an-alexa-system-for-hotels/, accessed: 14 Jun 2019.

55. Feratel http://www.feratel.at/en/, accessed: 14 Jun 2019.56. Outdooractive https://www.outdooractive.com/, accessed: 14 Jun 2019.57. Intermaps https://www.intermaps.com/en/, accessed: 14 Jun 2019.58. General Solutions https://general-solutions.eu/, accessed: 14 Jun 2019.59. Verkehrsauskunft Österreich https://verkehrsauskunft.at/, accessed: 14 Jun 2019.60. Wikidata https://www.wikidata.org/, accessed: 14 Jun 2019.61. Open Streetmap https://www.openstreetmap.org, accessed: 14 Jun 2019.62. Geonames https://www.geonames.org/, accessed: 14 Jun 2019.63. Pilot Seefeld https://www.seefeld.com/en/, accessed: 14 Jun 2019.64. Serfaus-Fiss-Ladis pilot https://www.serfaus-fiss-ladis.at/en/, accessed: 14 Jun 2019.65. Tirol Knowledge Graph http://graphdb.sti2.at:8080/, accessed: 14 Jun 2019.66. Infomax https://www.infomax-online.de/, accessed: 14 Jun 2019.67. Tomas https://www.tomas.travel/, accessed: 14 Jun 2019.68. Easybooking https://www.easybooking.eu/de/, accessed: 14 Jun 2019.69. Knowledge Graph DACH V3 https://www.tourismuszukunft.de/wp-content/uploads/2019/

04/Knowledge-Graph-DACH-V3.pdf, accessed: 14 Jun 2019.70. Dach KG Touristischen Knowledge Graph https://www.tourismuszukunft.de/2018/11/dach-

kg-auf-dem-weg-zum-touristischen-knowledge-graph/, accessed: 14 Jun 2019.71. Dach KG Nächste Schritte https://www.tourismuszukunft.de/2019/05/dach-kg-neue-

ergebnisse-naechste-schritte-beim-thema-open-data/, accessed: 14 Jun 2019.72. Data Tourisme https://www.datatourisme.fr/, accessed: 14 Jun 2019.73. Data Tourisme Government Website https://www.datatourisme.gouv.fr/, accessed: 14 Jun

2019.74. Tourism KG https://tourismkg.github.io/2019/, accessed: 14 Jun 2019.

References

Achichi M, Lisena P, Todorov K, Troncy R, Delahousse J (2018) DOREMUS: a graph oflinked musical works. In: Proceedings of the 17th International Semantic Web Conference(ISWC2018), Part II, Monterey, 8–12 Oct 2018. LNCS, vol 11137. Springer

Amato F, Moscato V, Picariello A, Sperlì G (2017) Knowledge-based access to art collections:the KIRA system. In: Proceedings of the 25th Italian Symposium on Advanced DatabaseSystems (SEBD2017), Squillace Lido (Catanzaro), Italy, 25–29 June 2017, CEUR-WS.org,CEUR Workshop Proceedings, vol 2037, p 82

Angles R, Gutierrez C (2005) Querying RDF data from a graph database perspective. In:Proceedings of the 2nd European Semantic Web Conference (ESWC2015), Heraklion, Crete,29 May–1 June 2005. LNCS, vol 3532. Springer, pp 346–360

Ankolekar A, Burstein M, Hobbs JR, Lassila O, Martin D, McDermott D, McIlraith SA, NarayananS, Paolucci M, Payne T et al (2002) DAML-S: web service description for the semantic web.In: Proceedings of the 1st International Semantic Web Conference (ISWC2002), Sardinia, 9–12June 2002. LNCS, vol 2342. Springer, pp 348–363

Araújo S, Hidders J, Schwabe D, de Vries AP (2011) SERIMI – resource description similarity,RDF instance matching and interlinking. In: Proceedings of the 6th International Workshop

https://gartner.com

https://bit.ly/2IAiHqw

https://www.sdcexec.com/software-technology/news/21011880/chatbot-market-to-grow-at-31-percent-cagr-from-2018-to-2024

https://www.sdcexec.com/software-technology/news/21011880/chatbot-market-to-grow-at-31-percent-cagr-from-2018-to-2024

https://tourismeschool.com/customer-journey-mapping-tourism-brands/

https://techcrunch.com/2018/06/19/amazon-launches-an-alexa-system-for-hotels/

https://techcrunch.com/2018/06/19/amazon-launches-an-alexa-system-for-hotels/

http://www.feratel.at/en/

https://www.outdooractive.com/

https://www.intermaps.com/en/

https://general-solutions.eu/

https://verkehrsauskunft.at/

https://www.wikidata.org/

https://www.openstreetmap.org

https://www.geonames.org/

https://www.seefeld.com/en/

https://www.serfaus-fiss-ladis.at/en/

http://graphdb.sti2.at:8080/

https://www.infomax-online.de/

https://www.tomas.travel/

https://www.easybooking.eu/de/

https://www.tourismuszukunft.de/wp-content/uploads/2019/04/Knowledge-Graph-DACH-V3.pdf

https://www.tourismuszukunft.de/wp-content/uploads/2019/04/Knowledge-Graph-DACH-V3.pdf

https://www.tourismuszukunft.de/2018/11/dach-kg-auf-dem-weg-zum-touristischen-knowledge-graph/

https://www.tourismuszukunft.de/2018/11/dach-kg-auf-dem-weg-zum-touristischen-knowledge-graph/

https://www.tourismuszukunft.de/2019/05/dach-kg-neue-ergebnisse-naechste-schritte-beim-thema-open-data/

https://www.tourismuszukunft.de/2019/05/dach-kg-neue-ergebnisse-naechste-schritte-beim-thema-open-data/

https://www.datatourisme.fr/

https://www.datatourisme.gouv.fr/

https://tourismkg.github.io/2019/

40 D. Fensel et al.

on Ontology Matching (OM2011), Bonn, 24 Oct 2011, CEUR-WS.org, CEUR WorkshopProceedings, vol 814

Auer S, Bizer C, Kobilarov G, Lehmann J, Cyganiak R, Ives Z (2007) Dbpedia: a nucleus for a webof open data. In: Proceedings of the 6th International Semantic Web Conference (ISWC2007)and 2nd Asian Semantic Web Conference (ASWC 2007), Busan, 11–15 Nov 2007. LNCS, vol7031. Springer, pp 722–735

Baader F, Calvanese D, McGuinness D, Patel-Schneider P, Nardi D (eds) (2003) The descriptionlogic handbook: theory, implementation and applications. Cambridge University Press, Cam-bridge

Batini C, Cappiello C, Francalanci C, Maurino A (2009) Methodologies for data quality assessmentand improvement. ACM Comput Surv (CSUR) 41(3):16

Beckett D, Berners-Lee T, Prud’hommeaux E, Carothers G (2014) RDF 1.1 Turtle. World WideWeb Consortium. https://www.w3.org/TR/turtle/

Beek W, Rietveld L, Bazoobandi HR, Wielemaker J, Schlobach S (2014) LOD laundromat: auniform way of publishing other people’s dirty data. In: Proceedings of the 13th InternationalSemantic Web Conference (ISWC2014), Riva del Garda, 19–23 Oct 2014. LNCS, vol 8796.Springer, pp 213–228

Bergman MK (2018) A knowledge representation practionary: guidelines based on CharlesSanders Peirce. Springer, Cham

Berners-Lee T (2006) Linked data. https://www.w3.org/DesignIssues/LinkedData.htmlBerners-Lee T, Hendler J, Lassila O, et al (2001) The semantic web. Sci Am 284(5):28–37Berners-Lee T, Connolly D, Kagal L, Scharf Y, Hendler J (2008) N3logic: a logical framework for

the world wide web. Theory Pract Logic Program 8(3):249–269Berners Lee T (2015) Five star open data. http://5stardata.info/enBizer C, Heath T, Idehen K, Berners-Lee T (2008) Linked data on the web (LDOW2008). In:

Proceedings of the 17th International Conference on World Wide Web (WWW2008), Beijing,21–25 Apr 2008. ACM, pp 1265–1266

Blanco R, Cambazoglu BB, Mika P, Torzec N (2013) Entity recommendations in web search. In:Proceedings of the 12th International Semantic Web Conference, Part II, Sydney, 21–25 Oct2013. LNCS, vol 8219. Springer

Bollacker K, Evans C, Paritosh P, Sturge T, Taylor J (2008) Freebase: a collaboratively createdgraph database for structuring human knowledge. In: Proceedings of the ACM SIGMODInternational Conference on Management of Data (SIGMOD’08), Vancouver, 10–12 June 2008.ACM, pp 1247–1250

Bonatti PA, Decker S, Polleres A, Presutti V (2019) Knowledge graphs: new directions forknowledge representation on the semantic web (dagstuhl seminar 18371). Schloss Dagstuhl–Leibniz-Zentrum fuer Informatik, vol 8, pp 29–111

Brachman RJ (1979) On the epistemological status of semantic networks In: Findler NV (ed)Associative networks: representation and use of knowledge by computers. Academic Press,New York

Brickley D, Guha RV, McBride B (2014) RDF schema 1.1. W3C Recommendation 25:2004–2014Cambria E, White B (2014) Jumping NLP curves: a review of natural language processing research.

IEEE Comput Intell Mag 9(2):48–57Carlson A, Betteridge J, Kisiel B, Settles B, Hruschka ER, Mitchell TM (2010) Toward an

architecture for never-ending language learning. In: Twenty-Fourth AAAI Conference onArtificial Intelligence (AAAI’10), Atlanta. AAAI Press

Chang CH, Kayed M, Girgis MR, Shaalan KF (2006) A survey of web information extractionsystems. IEEE Trans Knowl Data Eng 18(10):1411–1428

Chu X, Morcos J, Ilyas IF, Ouzzani M, Papotti P, Tang N, Ye Y (2015) KATARA: reliable datacleaning with knowledge bases and crowdsourcing. Proc VLDB Endowment 8(12):1952–1955

Council WTT (2018) Travel & tourism: economic impact 2018 China. World Travel & TourismCouncil (WTTC)

Dell’Erba M, Fodor O, Ricci F, Werthner H (2003) Harmonise: a solution for data interoperability.In: Towards the Knowledge Society. Springer, pp 433–445

https://www.w3.org/TR/turtle/

https://www.w3.org/DesignIssues/LinkedData.html

http://5stardata.info/en


Dietrich D, Gray J, McNamara T, Poikola A, Pollock P, Tait J, Zijlstra T et al (2009) Open datahandbook. Open Knowledge International. http://opendatahandbook.org

Dimou A, Vander Sande M, Colpaert P, Verborgh R, Mannens E, Van de Walle R (2014) RML:a generic language for integrated RDF mappings of heterogeneous data. In: Proceedings ofthe Workshop on Linked Data on the Web (LDOW’14), co-located with the 23rd InternationalWorld Wide Web Conference (WWW’14), Seoul, 8 Apr 2014, CEUR-WS.org

Dong XL (2018) Challenges and innovations in building a product knowledge graph. In:Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery& Data Mining (KDD’18), London, 19–23 Aug 2018. ACM, pp 2869–2869

Dong X, Gabrilovich E, Heitz G, Horn W, Lao N, Murphy K, Strohmann T, Sun S, Zhang W (2014)Knowledge vault: A web-scale approach to probabilistic knowledge fusion. In: Proceedings ofthe 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.ACM, pp 601–610

Drumond L, Girardi R (2008) A survey of ontology learning procedures. In: Proceedings of theCEUR Workshop, WONTO: 3rd Workshop on Ontologies and their Applications, Salvador,Bahia, 26 Oct 2008, 427:1–13

Faye DC, Cure O, Blin G (2012) A survey of RDF storage approaches. Revue Africaine de laRecherche en Informatique et Mathématiques Appliquées 15:11–35

Fensel D, Bussler C (2002) The web service modeling framework WSMF. Electron Commer ResAppl 1(2):113–137

Fensel D, Musen MA (2001) The semantic web: a brain for humankind. IEEE Intell Syst 16(2):24–25

Fensel D, Erdmann M, Studer R (1997) Ontology groups: Semantically enriched subnets of theWWW. In: In Proceedings of the International Workshop Intelligent Information Integrationduring the 21st German Annual Conference on Artificial Intelligence, Freiburg, Sept 1997,Citeseer

Fensel D, Richard V, Enrico B, Bobwielinga M (1999) UPML: a framework for knowledge systemreuse. In: Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence(IJCAI 99), Stockholm, July 31–Aug 6 1999, pp 16–23

Fensel D, Angele J, Decker S, Erdmann M, Schnurr HP, Studer R, Witt A (2000) Lessons learnedfrom applying AI to the web. Int J Cooperat Inf Syst 9(4):361–382

Fensel D, Lausen H, Polleres A, De Bruijn J, Stollberg M, Roman D, Domingue J (2006) Enablingsemantic web services: the web service modeling ontology. Springer Science & Business Media

Fensel D, Kerrigan M, Zaremba M (eds) (2008) Implementing semantic web services. SpringerFürber C, Hepp M (2010a) Using semantic web resources for data quality management. In:

Proceedings of the 17th International Conference on Knowledge Engineering and Managementby the Masses (EKAW2010), Lisbon, 11–15 Oct 2010. LNCS, vol 6317. Springer, pp 211–225

Fürber C, Hepp M (2010b) Using SPARQL and SPIN for data quality management on the semanticweb. In: Proceedings of the 13th International Conference on Business Information Systems(BIS2010), Berlin, 3–5 May 2010. LNBIP, vol 47. Springer, pp 35–46

Galárraga LA, Teflioudi C, Hose K, Suchanek F (2013) AMIE: association rule mining underincomplete evidence in ontological knowledge bases. In: Proceedings of the 22nd InternationalConference on the World Wide Web (WWW’13), Rio de Janeiro, 13–17 May 2013. ACM, pp413–422

Galárraga L, Teflioudi C, Hose K, Suchanek FM (2015) Fast rule mining in ontological knowledgebases with AMIE $$+ $$+. The VLDB J – Int J Very Large Data Bases (VLDB) 24(6):707–730

Garshol LM, Borge A (2013) Hafslund sesam–an archive on semantics. In: Proceedings of the10th International Conference on The Semantic Web: Semantics and Big data (ESWC2013),Montpellier, France, 26–30 May 2013. LNCS. vol 7882. Springer, pp 578–592

Gawriljuk G, Harth A, Knoblock CA, Szekely P (2016) A scalable approach to incrementallybuilding knowledge graphs. In: Proceedings of the 20th International Conference on Theoryand Practice of Digital Libraries (TPDL2016), Hannover, 5–9 Sept 2016. LNCS, vol 9819.Springer, pp 188–199

http://opendatahandbook.org

42 D. Fensel et al.

Getoor L, Machanavajjhala A (2013) Entity resolution for big data. In: Proceedings of the 19thInternational Conference on Knowledge Discovery and Data Mining: Tutorial (KDD2013),Chicago, 11–14 Aug 2013. ACM, p 1527

Giannopoulos G, Skoutas D, Maroulis T, Karagiannakis N, Athanasiou S (2014) FAGI: Aframework for fusing geospatial RDF data. In: Proceedings of the Confederated InternationalConferences “On the Move to Meaningful Internet Systems” (OTM2014), Amantea, 27–31 Oct2014. LNCS, vol 8841. Springer, pp 553–561

Gil Y (2011) Interactive knowledge capture in the new millennium: how the semantic web changedeverything. Knowl Eng Rev 26(1):45–51

Gomez-Perez JM, Pan JZ, Vetere G, Wu H (2017) Enterprise knowledge graph: An introduction.In: Pan JZ, Vetere G, Gomez-Perez JM, Wu H (eds) Exploiting linked data and knowledgegraphs in large organisations. Springer, pp 1–14

Gruninger M, Menzel C (2003) The process specification language (PSL) theory and applications.AI Mag 24(3):63–74

Guha R, McCool R, Miller E (2003) Semantic search. In: Proceedings of the 12th InternationalConference on the World Wide Web (WWW ’03), Budapest, 20–24 May 2003. ACM, pp 700–709

Gupta S, Kaiser G, Neistadt D, Grimm P (2003) Dom-based content extraction of html documents.In: Proceedings of the 12th international conference on World Wide Web (WWW’03),Budapest, Hungary, 20–24 May 2003. ACM, pp 207–214

Gupta S, Szekely P, Knoblock CA, Goel A, Taheriyan M, Muslea M (2012) Karma: A systemfor mapping structured sources into the semantic web. In: Proceedings of the 9th ExtendedSemantic Web Conference (ESWC2012), Heraklion, Crete, 27–31 May 2012. LNCS, vol 7540.Springer, pp 430–434

Gutiérrez-Cuellar J, Gómez-Pérez JM (2014) HAVAS 18 labs: a knowledge graph for innovationin the media industry. In: Proceedings of the Industry Track at the International Semantic WebConference (ISWC-IT 2014), Riva del Garda, 19–23 Oct 2014, CEUR-WS.org, vol 1383

Harth A, Hogan A, Delbru R, Umbrich J, O’Riain S, Decker S (2007) SWSE: answers before links!In: Proceedings of the Semantic Web Challenge 2007, Busan, Korea, November 13th, 2007. Co-located with the 6th International Semantic Web Conference and the 2nd Asian Semantic WebConference, Busan, 11–15 Nov 2007, CEUR-WS.org, vol 295, pp 137–144

Hipp J, Güntzer U, Nakhaeizadeh G (2000) Algorithms for association rule mining- a generalsurvey and comparison. ACM SIGKDD Exploration Newsl 2(1):58–64

Hoffart J, Suchanek FM, Berberich K, Weikum G (2013) YAGO2: A spatially and temporallyenhanced knowledge base from wikipedia. Artif Intell 194:28–61

Hogan A, Decker S, Harth A (2007) Performing object consolidation on the semantic web datagraph. In: Proceedings of the Workshop on Entity-Centric Approaches to Information andKnowledge Management on the Web (I3: Identity, Identifiers, Identification) co-located withthe 16th International World Wide Web Conference (WWW2007), CEUR Workshop, Banff, 8May 2007, vol 249

Inzalkar S, Sharma J (2015) A survey on text mining-techniques and application. Int J Res Sci Eng24:1–14

Kärle E, Simsek U, Fensel D (2017) semantify.it, a platform for creation, publication and distribu-tion of semantic annotations. In: SEMAPRO 2017: The Eleventh International Conference onAdvances in Semantic Processing. Curran Associates, Inc., New York, pp 22–30. http://arxiv.org/abs/1706.10067

Karoui L, Aufaure MA, Bennacer N (2004) Ontology discovery from web pages: application totourism. In: Proceedings of the Workshop of Knowledge Discovery and Ontologies, Pisa, 20–24Sept 2004, Citeseer

Khare R, Çelik T (2006) Microformats: a pragmatic path to the semantic web. In: Proceedingsof the 15th international conference on World Wide Web (WWW’06), Edinburgh, 23–26 May2006. ACM, pp 865–866

Kopecky J, Vitvar T, Bournez C, Farrell J (2007) SAWSDL: semantic annotations for WSDL andXML schema. IEEE Internet Comput 11(6):60–67

http://arxiv.org/abs/1706.10067



Korula N, Lattanzi S (2014) An efficient reconciliation algorithm for social networks. Proc VeryLarge Data Bases Endowment 7(5):377–388

Kärle E, Simsek U, Panasiuk O, Fensel D (2018) Building an ecosystem for the tyroleantourism knowledge graph. In: Proceedings of the International Conference on Trends in WebEngineering (ICWE2018), International Workshops, MATWEP, EnWot, KD-Web, WEOD,TourismKG: Revised Selected Papers, Caceres, 5 June 2018. Lecture Notes in ComputerScience, vol 11153. Springer, pp 260–267. https://doi.org/10.1007/978-3-030-03056-8_25

Langegger A, Wöß W (2009) XLWrap–querying and integrating arbitrary spreadsheets withSPARQL. In: Proceedings of the 8th International Semantic Web Conference (ISWC 2009),Chantilly, 25–29 Oct 2009. LNCS, vol 5823. Springer, pp 359–374

Lanthaler M, Gütl C (2013) Hydra: a vocabulary for hypermedia-driven web APIs. In: Bizer C,Heath T, Berners-Lee T, Hausenblas M, Auer S (eds) Proceedings of the Workshop on LinkedData on the Web Workshop (LDOW), Rio de Janeiro, 14 May 2013, vol 996, CEUR-WS.org

Lehmann J, Isele R, Jakob M, Jentzsch A, Kontokostas D, Mendes PN, Hellmann S, Morsey M,Van Kleef P, Auer S, et al (2015) DBpedia – a large-scale, multilingual knowledge base extractedfrom wikipedia. Semantic Web J 6(2):167–195

Lenat DB (1995) Cyc: a large-scale investment in knowledge infrastructure. Commun ACM38(11):33–38

Lenat DB, Guha RV (1989) Building large knowledge-based systems; Representation and Infer-ence in the Cyc Project, 1st edn. Addison-Wesley Longman Publishing Co., Inc.

Lertvittayakumjorn P, Kertkeidkachorn N, Ichise R (2017) Resolving range violations in DBpedia.In: Proceedings of the Joint International Semantic Technology Conference (JIST 2017), GoldCoast, 10–12 Nov 2017. Springer, pp 121–137

Ma L, Su Z, Pan Y, Zhang L, Liu T (2004) RStar: an RDF storage and query system forenterprise resource management. In: Proceedings of the 13th ACM international confer-ence on Information and knowledge management, Washington, DC, 8–13 Nov 2004. ACM,pp 484–491

Maltese V, Farazi F (2013) A semantic schema for geonames. Tech. Rep. DISI-13-004, Departmentof Information Engineering and Comouter Science, University of Trento, Trento

Manola F, Miller E, McBride B, et al (2004) RDF primer. W3C Recommendation 10(1–107):6Martin D, Burstein M, Mcdermott D, Mcilraith S, Paolucci M, Sycara K, Mcguinness DL, Sirin

E, Srinivasan N (2007) Bringing semantics to web services with OWL-S. World Wide Web10(3):243–277

Mendes PN, Mühleisen H, Bizer C (2012) Sieve: linked data quality assessment and fusion. In:Proceedings of the 2nd International Workshop on Linked Web Data Management (LWDM2012), in conjunction EDBT2012, Berlin, 30 March 2012. Citeseer, pp 116–123

Michelfeit J, Necasky M et al (2012) Linked open data aggregation: Conflict resolution and aggre-gate quality. In: Proceedings of the 36th Annual IEEE Computer Software and ApplicationsConference Workshops (COMPSAC2012), Izmir, 16–20 July 2012. IEEE, pp 106–111

Miles A, Bechhofer S (2009) SKOS simple knowledge organization system reference. World WideWeb Consortium. https://www.w3.org/TR/skos-reference/

Mohit B (2014) Named entity recognition. In: Zitouni I (ed) Natural language processing of semiticlanguages. Springer, pp 221–245

Moschitti A, Tymoshenko K, Alexopoulos P, Walker A, Nicosia M, Vetere G, Faraotti A, MontiM, Pan JZ, Wu H, et al (2017) Question answering and knowledge graphs. In: Pan JZ, VetereG, Gomez-Perez JM, Wu H (eds) Exploiting linked data and knowledge graphs in largeorganisations. Springer, pp 181–212

Motta E, Domingue J, Cabral L, Gaspari M, II I (2003) IRS-II: A framework and infrastructurefor semantic web services. In: The Semantic Web – ISWC 2003, Second InternationalSemantic Web Conference, Sanibel Island, 20–23 Oct 2003. LNCS, vol 2870. Springer,pp 306–318

Newell A, et al (1982) The knowledge level. Artif Intell 18(1):87–127Nickel M, Murphy K, Tresp V, Gabrilovich E (2015) A review of relational machine learning for

knowledge graphs. Proc IEEE 104(1):11–33

https://doi.org/10.1007/978-3-030-03056-8_25

https://www.w3.org/TR/skos-reference/

44 D. Fensel et al.

O’connor MJ, Halaschek-Wiener C, Musen MA (2010) Mapping master: a flexible approach formapping spreadsheets to OWL. In: 9th International Semantic Web Conference (ISWC 2010),Shanghai, 7–11 Nov 2010. LNCS, vol 6497. Springer, pp 194–208

Pan JZ, Vetere G, Gomez-Perez JM, Wu H (eds) (2017) Exploiting linked data and knowledgegraphs in large organisations. Springer

Panasiuk O, Akbar Z, Gerrier T, Fensel D (2018a) Representing geodata for tourism withschema.org. In: Grueau C, Laurini R, Ragia L (eds) Proceedings of the 4th InternationalConference on Geographical Information Systems Theory, Applications and Management,GISTAM 2018, Funchal, Madeira, 17–19 March 2018. SciTePress, pp 239–246. https://doi.org/10.5220/0006755102390246

Panasiuk O, Akbar Z, Simsek U, Fensel D (2018b) Enabling conversational tourism assistantsthrough schema.org mapping. In: Gangemi A, Gentile AL, Nuzzolese AG, Rudolph S,Maleshkova M, Paulheim H, Pan JZ, Alam M (eds) The Semantic Web: ESWC 2018 SatelliteEvents – ESWC 2018 Satellite Events, Heraklion, Crete, 3–7 June 2018. Revised SelectedPapers, Lecture Notes in Computer Science, vol 11155. Springer, pp 137–141. https://doi.org/10.1007/978-3-319-98192-5_26

Panasiuk O, Kärle E, Simsek U, Fensel D (2018c) Defining tourism domains for semanticannotation of web content. e-Review of Tourism Research, Research notes from the ENTER2018 Conference on ICT in Tourism, Jönköping, Sweden, January 24–26, 2018 9, URL https://journals.tdl.org/ertr/index.php/ertr/article/view/127

Patil AA, Oundhakar SA, Sheth AP, Verma K (2004) METEOR-S web service annotation frame-work. In: Proceedings of the 13th international conference on World Wide Web (WWW’04),New York, 17–22 May 2004. ACM, pp 553–562

Paulheim H (2017) Knowledge graph refinement: a survey of approaches and evaluation methods.Semantic Web 8(3):489–508

Paulheim H, Bizer C (2013) Type inference on noisy RDF data. In: Proceedings of the 12thInternational Semantic Web Conference (ISWC 2013), Sydney, 21–25 Oct 2013. LNCS, vol8218, Springer, pp 510–525

Paulheim H, Bizer C (2014) Improving the quality of linked data using statistical distributions. IntJ Semantic Web Inf Syst (IJSWIS) 10(2):63–86

Paulheim H (2018a) How much is a triple? estimating the cost of knowledge graph creation.In: Proceedings of the 17th International Semantic Web Conference (ISWC2018): Posters& Demonstrations, Industry and Blue Sky Ideas Tracks, Monterey, 8–12 Oct 2018, CEUR-WS.org, CEUR Workshop Proceedings, vol 2180, http://ceur-ws.org/Vol-2180/ISWC_2018_Outrageous_Ideas_paper_10.pdf

Paulheim H (2018b) Machine learning with and for semantic web knowledge graphs. In: d’AmatoC, Theobald M (eds) Reasoning Web. Learning, Uncertainty, Streaming, and Scalability – 14thInternational Summer School 2018, Esch-sur-Alzette, Luxembourg, 22–26 Sept 2018. TutorialLectures, Springer, pp 110–141

Pereira RL, Sousa PC, Barata R, Oliveira A, Monsieur G (2015) Citysdk tourism api-building valuearound open data. J Internet Serv Appl 6(1):24

Quimbaya AP, Muñoz O, Londoño D, Bohórquez R, García OM, González RA, Amortegui MP,Rodriguez S, Bustamante A (2014) An executable knowledge base for clinical practice guidelinerules. Proc Technol 16:1446–1455

Rekatsinas T, Chu X, Ilyas IF, Ré C (2017) Holoclean: Holistic data repairs with probabilisticinference. Proc the Very Large Data Bases Endowment (PVLDB) 10(11):1190–1201

Roman D, Kopecky J, Vitvar T, Domingue J, Fensel D (2015) WSMO-Lite and hRESTS:lightweight semantic annotations for web services and RESTful APIs. J Web Semantics 31:39–58

Sabou M, Arsal I, Brasoveanu AM (2013) Tourmislod: a tourism linked data set. Semantic Web4(3):271–276

Schreiber G (2013) Knowledge acquisition and the web. Int J Hum-Comput Stud 71(2):206–210Schreiber G, Akkermans H, Anjewierden A, Shadbolt N, de Hoog R, Van de Velde W, Shadbolt

NR, Wielinga B (2000) Knowledge engineering and management: the CommonKADS method-ology. The MIT Press, Cambridge

https://doi.org/10.5220/0006755102390246

https://doi.org/10.5220/0006755102390246

https://doi.org/10.1007/978-3-319-98192-5_26

https://doi.org/10.1007/978-3-319-98192-5_26

https://journals.tdl.org/ertr/index.php/ertr/article/view/127


http://ceur-ws.org/Vol-2180/ISWC_2018_Outrageous_Ideas_paper_10.pdf

http://ceur-ws.org/Vol-2180/ISWC_2018_Outrageous_Ideas_paper_10.pdf


Schultz A, Matteini A, Isele R, Mendes PN, Bizer C, Becker C (2012) LDIF – a framework forlarge-scale linked data integration. In: 21st International World Wide Web Conference (WWW2012), Developers Track, Lyon, 18–20 Apr 2012

Shadbolt N, Smart PR, Wilson J, Sharples S (2015) Knowledge elicitation. In: Wilson J, SharplesS (eds) Evaluation of human work, 4th edn. CRC Press, Boca Raton, pp 163–200

Shehata S, Karray F, Kamel M (2009) An efficient concept-based mining model for enhancing textclustering. IEEE Trans Knowl Data Eng 22(10):1360–1371

Silwattananusarn T, Tuamsuk K (2012) Data mining and its applications for knowledge manage-ment: a literature review from 2007 to 2012. Int J Data Min Knowl Manag Process 2(5). https://arxiv.org/abs/1210.2872

Simon HA (1957) Models of man, nueva yorkSimsek U, Fensel D (2018) Now we are talking! Flexible and open goal-oriented dialogue systems

for accessing touristic services. In: Research Notes from the ENTER 2018 Conference onICT in Tourism, Jönköping, Sweden, 24–26 Jan 2018. https://journals.tdl.org/ertr/index.php/ertr/article/view/126

Simsek U, Kärle E, Fensel D (2018a) Machine readable web APIs with schema.org actionannotations. Proc Comput Sci 137:255–261

Simsek U, Kärle E, Holzknecht O, Fensel D (2018b) Domain specific semantic validation ofschema.org annotations. In: Petrenko AK, Voronkov A (eds) Perspectives of system informatics.Springer International Publishing, Cham, pp 417–429

Simsek U, Kärle E, Fensel D (2019) RocketRML – a NodeJS implementation of a use-case specificRML mapper. http://arxiv.org/abs/1903.04969

Singhal A (2012) Introducing the knowledge graph: things, not strings. Official Google Blog 5.https://googleblog.blogspot.com/2012/05/introducing-knowledge-graph-things-not.html

Sleeman J, Finin T (2013) Type prediction for efficient coreference resolution in heterogeneoussemantic graphs. In: Proceedings of the 7th International Conference on Semantic Computing(ICSC2013). IEEE Computer Society, Irvine, 16–18 Sept 2013

Sleeman J, Finin T, Joshi A (2015) Topic modeling for RDF graphs. In: Proceedings of the3rd International Workshop on Linked Data for Information Extraction (LD4IE2015) co-located with the 14th International Semantic Web Conference (ISWC 2015), CEUR Workshop,Bethlehem, 12 Oct 2015, CEUR-WS.org, CEUR Workshop Proceedings, vol 1467, pp 48–62

Socher R, Chen D, Manning CD, Ng A (2013) Reasoning with neural tensor networks forknowledge base completion. In: Proceedings of the 26th International Conference on NeuralInformation Processing Systems (NIPS’13), Lake Tahoe, 5–10 Dec 2013, vol 1, pp 926–934

Sporny M, Longley D, Kellogg G, Lanthaler M, Lindström N (2014) JSON-LD 1.0. W3CRecommendation 16:41. https://www.w3.org/TR/json-ld/

Stearns MQ, Price C, Spackman KA, Wang AY (2001) SNOMED clinical terms: overview of thedevelopment process and project status. In: Proceedings of the AMIA Symposium, WashingtonDC, USA, 11–15 Nov 2001, American Medical Informatics Association, p 662

Stegmaier F, Gröbner U, Döller M, Kosch H, Baese G (2009) Evaluation of current RDFdatabase solutions. In: Proceedings of the 10th International Workshop on Semantic MultimediaDatabase Technologies (SeMuDaTe) at the 4th International Conference on Semantics AndDigital Media Technologies (SAMT), Graz, 2 Dec 2009, pp. 39–55

Studer R, Benjamins VR, Fensel D (1998) Knowledge engineering: principles and methods. DataKnowl Eng 25(1–2):161–197

Suchanek FM, Kasneci G, Weikum G (2007) Yago: a core of semantic knowledge. In: Proceedingsof the 16th International World Wide Web Conference (WWW2007), Banff, Canada, 8–12 May2007. ACM, pp 697–706. https://doi.org/10.1145/1242572.1242667

Swartz A (2002) Musicbrainz: a semantic web service. IEEE Intell Syst 17(1):76–77Sánchez D, Moreno A (2006) A methodology for knowledge acquisition from the web. Int J

Knowl-Based Intell Eng Syst 10(6):453–475Tandon N, De Melo G, Suchanek F, Weikum G (2014) WebChild: harvesting and organizing

commonsense knowledge from the web. In: Proceedings of the 7th ACM internationalconference on Web search and data mining (WSDM 2014), New York, 24–28 Feb 2014. ACM,pp 523–532

https://arxiv.org/abs/1210.2872

https://arxiv.org/abs/1210.2872




https://googleblog.blogspot.com/2012/05/introducing-knowledge-graph-things-not.html

https://www.w3.org/TR/json-ld/

https://doi.org/10.1145/1242572.1242667

46 D. Fensel et al.

Töpper G, Knuth M, Sack H (2012) Dbpedia ontology enrichment for inconsistency detection. In:Proceedings of the 8th International Conference on Semantic Systems (I-SEMANTICS ’12),Graz, 5–7 Sept 2012. ACM, pp 33–40

Van Deursen D, Poppe C, Martens G, Mannens E, Van de Walle R (2008) XML to RDF conversion:a generic approach. In: Proceedings of the 4rd International Conference on Automated solutionsfor Cross Media Content and Multi-Channel Distribution (AXMEDIS’08), Florence, 17–19 Nov2008. IEEE, pp 138–144

Verborgh R, Steiner T, Van Deursen D, De Roo J, Van de Walle R, Vallés JG (2013) Capturing thefunctionality of web services with functional descriptions. Multimed Tools Appl 64(2):365–387

Verborgh R, Harth A, Maleshkova M, Stadtmüller S, Steiner T, Taheriyan M, Van de Walle R(2014) Survey of semantic description of REST APIs. In: Pautasso C, Wilde E, Alarcon R (eds)REST: Advanced Research Topics and Practical Applications. Springer, pp 69–89

Villazon-Terrazas B, Garcia-Santa N, Ren Y, Faraotti A, Wu H, Zhao Y, Vetere G, Pan JZ (2017a)Knowledge graph foundations. In: Pan JZ, Vetere G, Gomez-Perez JM, Wu H (eds) ExploitingLinked Data and Knowledge Graphs in Large Organisations. Springer, pp 17–55

Villazon-Terrazas B, Garcia-Santa N, Ren Y, Srinivas K, Rodriguez-Muro M, Alexopoulos P, PanJZ (2017b) Construction of enterprise knowledge graphs (I). In: Pan JZ, Vetere G, Gomez-Perez JM, Wu H (eds) Exploiting Linked Data and Knowledge Graphs in Large Organisations.Springer, pp 87–116

Volz J, Bizer C, Gaedke M, Kobilarov G (2009) Silk – A link discovery framework for the webof data. In: Proceedings of the WWW2009 Workshop on Linked Data on the Web, LDOW,Madrid, vol 538, Citeseer

Vrandecic D, Krötzsch M (2014) Wikidata: a free collaborative knowledge base. Commun ACM57(10):78–85

Winkler WE (2006) Overview of record linkage and current research directions. Research reportseries: statistics #2006-2, Bureau of the Census, USA. https://www.census.gov/srd/papers/pdf/rrs2006-02.pdf

Zaveri A, Kontokostas D, Sherif MA, Bühmann L, Morsey M, Auer S, Lehmann J (2013) User-driven quality evaluation of DBpedia. In: Proceedings of the 9th International Conference onSemantic Systems (I-SEMANTICS ’13), Graz, 4–6 Sept 2013. ACM, pp 97–104

Zaveri A, Dastgheib S, Wu C, Whetzel T, Verborgh R, Avillach P, Korodi G, Terryn R, JagodnikK, Assis P, et al (2017) SmartAPI: Towards a more intelligent network of web APIs. In:Proceedings of the 14th European Semantic Web Conference (ESWC 2017), Portoroz, Sovenia,May 28–June 1. LNCS, vol 10250. Springer, pp 154–169

https://www.census.gov/srd/papers/pdf/rrs2006-02.pdf

https://www.census.gov/srd/papers/pdf/rrs2006-02.pdf

semantic web empowered e-tourism · 2020-04-24 · semantic web empowered e-tourism kevin angele,...

Documents