1 IOS Press Ontologies & Reasoning in Enterprise ApplicationsSAP Research Karlsruhe, Vincenz-Priessnitz-Str. 1, 76131 Karlsruhe, Germany, E-mail: [email protected] Abstract. This survey

Undefined 1 (2009) 1–5 1IOS Press

Ontologies & Reasoning in EnterpriseApplicationsEditor(s): Name Surname, University, CountrySolicited review(s): Name Surname, University, CountryOpen review(s): Name Surname, University, Country

Daniel OberleSAP Research Karlsruhe, Vincenz-Priessnitz-Str. 1, 76131 Karlsruhe, Germany, E-mail: [email protected]

Abstract. This survey paper identifies the value-adding features of ontologies & reasoning and positions them to the plethoraof established technologies. This identification is the prerequisite for explaining how and why such technological benefits leadto benefits for enterprise applications as a next step. In addition, this paper reports on several challenges that frequently occurwhen trying to adopt ontologies & reasoning in existing enterprise settings. Together with reports for several SAP Researchcase studies, this paper channels back experiences in applying ontologies & reasoning to the Semantic Web community. Severalrecommendations for future research directions are given based on the gathered experiences.

Keywords: Semantics, Semantic Technologies, Ontologies, Reasoning, Enterprise, Industry, Commercial, Application

1. Introduction

The last years have seen great academic interestin the Semantic Web and in semantic technologiesand even first industrial products appeared. [4] On thebrink to the industrial breakthrough, semantic tech-nologies have been criticized for being elusive, espe-cially for use in enterprise applications, however [98].Typical questions asked in an enterprise setting are:“What business problem is semantics solving? Whatrelevance is there between semantics and the realworld? What value proposition for the conduct of busi-ness is there that is being addressed by semantics?”[47]

Since one reason for such sentiments is on the levelof terminology, the first step in answering such ques-tions is to clarify what we are talking about. As pointedout in [50], the murkiness of the words ‘semantics’ or‘semantic technologies’ adds to the confusion. There-fore, this paper sets its focus on ontologies & reason-ing [40], i.e., a specific semantic technology, in the fol-lowing. While there is still a certain bandwidth of in-terpretation in ontologies & reasoning, it is still much

more concrete than ‘semantics’ or ‘semantic technolo-gies.’

In the second step, this paper identifies the value-adding features inherent in the technology in additionto the often quoted formality and reasoning (Section3). On the basis of the value-adding features, this papersystematically positions ontologies & reasoning to theplethora of related technologies. Thus, comparisonsbecome possible to point out which (combination of)feature(s) make ontologies & reasoning unique com-pared to established technologies.

Identifying the value-adding features serves both asa survey and, more importantly, as the prerequisite forexplaining how and why such technological benefitslead to benefits for enterprise applications in the thirdstep (Section 4). This argumentation line representsa contribution since related papers either discuss thebenefits of ontologies & reasoning on the level of ap-plication scenarios, such as interoperability or search,or technological benefits only (cf. [22,63,46,97,88]).

After the benefits have been clarified and explained,this paper draws its attention to the challenges whenadopting ontologies & reasoning drawn from more

0000-0000/09/$00.00 c© 2009 – IOS Press and the authors. All rights reserved

2 Daniel Oberle / Ontologies & Reasoning in Enterprise Applications

than five years of industrial research (Section 5). Thechallenges are mainly due to the often complex bound-ary conditions in an enterprise setting, e.g., technicalintegration into existing technology stacks becomesnecessary.

In addition to the challenges, the paper contributesexperience reports for several SAP Research case stud-ies (Section 6) where ontologies & reasoning havebeen applied in products, pilots, and prototypes. Theexperience reports highlight which value adding fea-tures, benefits, and challenges did occur in the cor-responding case study and which compromises havebeen taken.

Besides channeling back the experiences to the Se-mantic Web community, this paper also elicits poten-tial trade-offs, promising research results and providesrecommendations for future research directions (Sec-tion 7).

2. Scenario

The declared goal of the Norwegian Oil and Gas in-dustry is the integration of different field data, the val-idation of information delivered by different sources,and the exchange of information between off-shoreand on-shore facilities [95] (cf. Figure 1).

Fig. 1. Ontologies play a major role in the Norwegian Oil & Gasindustry’s strategy. [1]

This requires that Oil and Gas IT systems, e.g., assetmanagement or facility monitoring, share informationaccording to a reference model, and that they can inter-pret messages using that model [20]. Such a referencemodel is given by the ISO 15926 Oil and Gas ontol-ogy [53] which is formalized in OWL. The ontologyshall serve as a common vocabulary to exchange infor-mation between field data and and the onshore partici-pants (operators and vendors). In particular, as pointed

out in [1], the offshore industry has huge economic in-vestments in data acquisitions, analysis, visualization,documentation and archiving. Some of the data are inuse for decades and it is one of the main assets. Orga-nizational units and IT systems last rarely more thanfew years. The most stable element in this environ-ment is the terminologies used in the business domainsalong the value chain. Therefore, it is the purpose ofISO 15926 to capture the terminology in a formal andreusable way. According to [1], using ISO 15926 it isfurther possible to:

– integrate data within and across business domains– create an architecture for Web services– include reasoning as a part of the ontology for

creation of autonomous solutions– be able to store data over time– to use the ontology as a reference ontology to be

used in-house and between companies– to use the ontology in engineering

3. Value-adding Features of Ontologies &Reasoning

Value-adding features are advantages of a technol-ogy compared to related technologies. The followingsection carefully identifies the value-adding features ofontologies & reasoning.1 This serves both as a surveyand as the prerequisite for explaining how and whysuch technological benefits lead to business benefits inSection 4.

Ontologies & reasoning inherit from several relatedand precursory technologies, e.g., conceptual mod-eling languages, logics, theorem provers, deductivedatabases, AI, or semantic nets. Therefore, many of theidentified value-adding features are not unique but arealso offered by related technologies. Because of this,a comparison with related and precursory technologiesis rather tricky.

3.1. Conceptual Modeling

Explanation Ontologies are a means for modeling aspecific universe of discourse according to the well-known principle of conceptualization [27]. Conceptualmodeling languages make use of the following intu-itive modeling primitives:

1The starting point for the following analysis is [39,43] who in-troduce communication (here: Conceptual Modeling combined withFormality), computational inference (here: Reasoning) and Reuse asvalue-adding features.

Daniel Oberle / Ontologies & Reasoning in Enterprise Applications 3

Classes are collections of instances that have similarproperties.

Properties represent relations between classes and in-stances, respectively. Classes and properties canbe hierarchically organized in taxonomies viaspecialization, respectively.

Instances (also called objects or individuals) are con-crete members of a class.

Rules & Axioms Rules take the form of premise →conclusion and allow to infer implicit informa-tion combined with Reasoning (cf. Section 3.8).Axioms allow to capture additional informationabout classes and relations, e.g., that a relation istransitive or symmetric.

According to [66], conceptual models are more ef-fective because human performance in man machineinteractions depends on the gap between the human’sconceptualization of the task and the presented inter-face. When the gap is large, performance is poor. Whenthe gap is small, performance is good. Thus, concep-tual models better correspond to an end-user’s concep-tualization than machine-oriented models. [45,18] Fur-ther, [39] argues that Conceptual Modeling paired withFormality also facilitates the communication betweenimplemented computational systems and/or humans.

Scenario As pointed out in the running example, themost stable element in the Oil and Gas industry is theterminologies used in the business domains along thevalue chain. Therefore, capturing this domain in theISO 15926 ontology provides benefits since it corre-sponds better to an Oil and Gas expert’s conceptual-ization of the domain.

Related Work The use of classes and relations asmain modeling primitives is well established in otherdisciplines such as database design and software en-gineering. Therefore, conceptual modeling is not id-iosyncratic for ontologies and has been introduced asearly as the Entity-relationship Model (ERM) [19],UML [12], or XML Schema (XSD). Note, however,that ERM does not feature the instances modelingprimitive, UML devotes a separate diagrams for in-stances and XSD lacks properties.

3.2. Flexibility

Explanation Databases and (model-driven) softwareengineering typically start with a conceptual model atdesign time which is then manually or automaticallytransformed to an executable model. That means the

initial conceptual model is hardly adaptable after thetransformation during run time. In contrast, ontologylanguages and stores are typically flexible with regardsto the conceptual model. Flexibility means that classes,properties, rules & axioms as well as instances of anontology can be managed at run time of an applica-tion (frequently dubbed ‘agile schema development’ or‘schema last’ feature). This is achieved by:

API The basic means for managing classes, proper-ties, rules & axioms as well as instances at runtime is an Application Programming Interface(API) provided by an ontology store. The API al-lows to add, change or delete classes, properties,rules & axioms, and instances during run time ofan application.

Evolution Some APIs additionally support automatedevolution strategies to unambiguously define theway how changes will be resolved [90]. Ontologychanges and their effects can be managed by cre-ating and maintaining different variants of the on-tology [68]. This includes reclassification of in-stances when a class is deleted, for instance.

Dynamic Direct Programming APIs might feature adirect programming model [79], i.e., each object-oriented class or field is an element of the on-tology. In dynamically typed languages, the APImight also be generated on-the-fly, i.e., changesin the ontology are reflected in the API at run-time.

Scenario Flexibility also benefits the Oil and Gassetting, e.g., when new types of equipment have to becaptured in the ISO 15926 ontology after it is initiallydeployed in applications. When a specific class in thetaxonomy is deleted, evolution strategies care for itssubclasses and for reclassifying all its instances.

Related Work The currently prevailing relationaldatabases require that a database schema exists priorto data entry. However, flexibility is typically given forgraph-based data and conceptual modeling languages(cf. [5] for an overview). According to [68], ontologyevolution is not the same as database schema evolutionbecause ontologies comprise data (i.e., instances) aswell, are more often reused, and are de-centralized bynature. Object-oriented database management systems[52] also support APIs and evolution [14] to a certainextent.


3.3. Direct Interaction

Explanation Through the combination of Concep-tual Modeling and Flexibility, realizing the user-friendly direct interaction with the ontology and itsinstances is straightforward. An ontology and its in-stances can be created, changed, and populated bygeneric tools fostered by the standardized ontologylanguages (cf. Web Compliance). Direct interactiontypically happens via the following UI paradigms:

Graphical Tree or graph-based visualizations of theontology with editing possibilities. A survey isgiven in [48].

Wiki Enhancements of existing Wikis where eachpage represents an instance with relations to otherpages. A good example is the Semantic MediaWiki [55].

Natural language There are several user interfacesthat apply controlled natural language. One ex-ample for a guided input natural language ontol-ogy editor is [10].

Scenario ISO 15926 is engineered collaborativelyby special interest groups each of which focus on aspecific sub-domain of Oil and Gas. Examples forsuch sub-domains are drilling and completion, opera-tion and maintenance, or production and reservoir. Itis the task of each special interest group to capturethe corresponding sub-domain ontologically and inte-grate it in the overall ISO 15926. Using the SemanticMedia Wiki, it becomes possible for the domain ex-perts to collaboratively model and document their sub-domains.

Related Work Direct interaction with relational databasesrequire in-depth knowledge of the relational algebra,e.g., about primary and foreign keys. That means ad-ditional abstraction layers have to be established ontop to allow a user-friendly interaction. There are othertechnologies that provide generic tools for direct in-teraction. A good example is topic maps [89] whichalso are a means for Conceptual Modeling includingthe feature of Flexibility.

3.4. Reuse

Explanation The declarative nature of ontologiesmakes them particularly eligible for reuse. That meansthe effort of conceptualizing a universe of discourseis undertaken once and the resulting ontology can bereused in multiple applications. In addition, ontologies

are often attributed as being a ‘shared’ conceptualiza-tion, i.e., users agree on the conceptualization, whatpotentially increases the level of reuse. According to[84] there are the following levels of acceptance:

Individual The ontology is created and used by onlyone person, e.g., within an application for reason-ing and transparent to the user.

Community A group, company or sector agrees on aspecific ontology and uses it.

World The ontology is published on the Web (cf. Sec-tion 3.6) and can potentially be used by anybody.

Transitions from acceptance levels individual tocommunity or from community to world require meth-ods for reaching consensus, e.g., [91].

Scenario In the Oil and Gas setting, the ISO 15926ontology shall be used for several purposes and in dif-ferent applications. Thus, reuse is one of the inher-ent value-adding features sought by using an ontology.The level of acceptance would be community, since theontology shall be used by the Oil & Gas industry.

Related Work Although it is not their initial motiva-tion, technologies such as conceptual database design(with, e.g., ERM [19]), UML [12], Model-Driven En-gineering with its platform-independent models [49],or XML Schemas, can achieve reuse as well. In prac-tice, shared conceptualizations are often representedby such languages. Examples are the bulk of SOA ref-erence models such as SOA-RM [59] or SOA-RAF [2]given in UML.

3.5. Best Practices

Explanation High quality reference ontologies, e.g.,[71,8], are often built by ontological analysis. Thatmeans classes and properties are related to some pro-posed invariant categories of human cognition. What istypically gained is an increased understanding of one’sown ontology as well as a cleaner design. There ex-ist the following best practices to support ontologicalanalysis:

Foundational Ontologies are a means to prevent mod-eling from scratch by giving a well-engineeredstarting point for one’s own ontology. The use ofsuch an ontology prompts an ontology engineerto sharpen his/her notion of the concepts to bemodeled. A prominent example is the DOLCEfoundational ontology [37].


Ontology Design Patterns save modeling efforts byproviding templates for ever re-occurring mod-eling needs. Thus, they prevent re-inventing thewheel and make modeling decisions better trace-able. [77]

Quality Criteria The OntoClean [41] approach pro-vides explicit quality criteria for building an on-tologically correct taxonomy.

Scenario Using a foundational ontology as startingpoint for ISO 15926 has indeed been discussed andwould lead to a philosophically sound basis. It wouldprovide the means for disambiguating terms such as‘drilling’ which can be both an activity and a topic ofexpertise. Fundamental ontology design patterns suchas ‘participation in an event’ would make the overallontology better to maintain and comprehensible. Onto-Clean would provide for an ontologically sound taxo-nomic backbone of ISO 15926.

Related Work To the best of our knowledge, the ideaof having foundational ontologies and quality criteriahas not occurred in other disciplines. Note, however,that both can be applied to other conceptual modelinglanguages as well, e.g., UML class models. There is abulk of related approaches to ontology design patterns— especially the well-known design patterns in soft-ware engineering [36].

3.6. Web Compliance

Explanation Up until the late 90s there were severalproprietary and incompatible ontology languages andcorresponding tool suites. The recommendations of theW3C Semantic Web Activity provide Web Compli-ance for ontologies and allow:

Publication Standardized ontology languages such asRDF(S) [15] and OWL [61] allow to publish on-tologies on the Web. That means every element ofan ontology is identified by a URI, XML serial-izations are available, and ontologies can be dis-tributed, modularized and referenced across theWeb.Information systems can publish instances adher-ing to published ontologies as linked data [11],i.e., RDF representations of the data containinglinks to data contained in the same or other sys-tems. The information systems provide endpointswhich deliver RDF data at dereferencable URIs,i.e., URIs that are used as identifiers for resourcesand, at the same time, as pointers to additional in-formation on those resources.

Querying SPARQL [78] may be used as a unifiedquerying language across the different data sets.SPARQL resembles SQL in syntax but pays trib-ute to the fact that data is distributed in the Web.

Annotation There are W3C recommendations for an-notating existing Web resources with ontologies.Those are SA-WSDL [30] and SA-REST [38] forannotating Web service descriptions and RDFa[3] for annotating HTML pages.

Scenario One reason for formalizing ISO 15926 inOWL, besides its original language EXPRESS, is cer-tainly because OWL is a W3C recommendation andmany tools for editing, storage, and reasoning exist.Web compliance benefits the envisioned ‘smarter so-lutions’ such as Web portals based on ISO 15926 (cf.Figure 1).

Related Work Web compliance is also given for thewhole set of recommendations that revolve aroundXML. XML Schema [28], especially, is often consid-ered a full-fledged substitute for RDF(S) in the contextof Linked Data. In fact, the differences between XMLSchema and RDF(S) are elusive [23].

3.7. Formality

Explanation Typically, ontology languages such asOWL are based on formal logics with model-theoreticsemantics. On the one hand, this provides for unam-biguous meaning of modeling primitives. As an ex-ample, consider the formal semantics of the ‘subclass-of’ modeling primitive. The model theory underlyingOWL basically maps the modeling primitive to thesubset operator in set theory. Contrast this with theincomplete, ambiguous, and informal specification ofUML [81]. On the other hand, the model theory is aprerequisite for having a proof theory, and, thus, Rea-soning becomes possible as another value-adding fea-ture. There are three major families of ontology lan-guages [72]:

Logic Programming Languages in this family aregeared to deal efficiently with larger sets of in-stances and the reasoning tasks of rules and in-stance retrieval. A prominent example is F-Logic[51].

Description Logics Languages in this family are strictsubsets of first-order logics geared at the rea-soning tasks of subsumption checks, consistencychecks, and instance classification (cf. Section3.8). The most prominent example is OWL-DL[61].


First-Order Logics and higher-order logics allow forgreater expressiveness but are typically not decid-able, i.e., a sound and complete reasoner cannotbe built.

Most languages existed prior to being used as an on-tology language and stem from related disciplines suchas mathematical logics, deductive databases or artifi-cial intelligence.

Scenario Since one of the goals of ISO 15926 is tocapture the most stable element of terminologies inthe Oil & Gas domain, formality benefits the endeavorby countering ambiguity. Being a prerequisite for rea-soning, formality also helps the goal postulated in [1],which is to enable autonomous solutions by applyingreasoning on the basis of ISO 15926.

Related Work As depicted in Figure 2, the formal-ity of a (conceptual modeling) language is actually acontinuum [94]. At the one extreme, there are purelyinformal, i.e., natural, languages to specify an “ontol-ogy.” However, in our definition, speaking of an ontol-ogy is only valid in the rightmost category of ‘logicallanguages.’ In between are languages such as ERM,that do provide a mathematical definition, yet lack amodel-theoretic formal semantics as a prerequisite forreasoning.

informal formal

Terms

Data

Dictionaries

Ad-hoc

Hierarchies(Yahoo!)

‘Ordinary’

Glossaries

Thesauri

Structured

Glossaries

XML

DTDs

Informal

Hierarchies(Folksonomies)

DB

Schema

XML

Schema

Data Models

(UML, STEP)

Formal

Taxonomies

Logic

Programming

Description

Logics

First-order,

Higher-order,

Modal Logic

Glossaries &

Data Dictionaries

Thesauri,

Taxonomies

MetaData,

XML Schemas,

Data Models

Logical

Languages

Fig. 2. The continuum of formality. Adapted version from [94].

3.8. Reasoning

Explanation As already mentioned in Section 1, anylogical deduction on the basis of a formal ontology isdefined as reasoning in this paper. Different reasoningservices are offered by inference engines (also calledreasoners) depending on the ontology language.

Subsumption Checking infers implicit super- andsubclass relations between classes. Being appliedmutually to every possible pair of classes, sub-sumption checking makes it possible to build thewhole class hierarchy (called automatic classifi-cation).

Consistency Checking examines whether an ontol-ogy contains contradictions.

Instance Classification checks whether a specific in-stance is a member of a given class.

Instance Retrieval means querying for instances wherebyrules and axioms are considered.

The reasoning services are a benefit because the of-fered functionality does not have to be developed fromscratch for each application.

Scenario Section 2 discussed that reasoning shall beused for the creation of autonomous solutions in theOil and Gas domain. For example, rules can be appliedto declaratively capture and infer knowledge on the ba-sis of ISO 15926 in production optimization.

Related Work The reasoning services of logic-pro-gramming-based inference engines overlap with Rete-based business rules management systems [34]. How-ever, the latter are less formal with respect to the con-tinuum depicted in Figure 2 and are not paired with aconceptual modeling language.

4. Benefits for Enterprise Applications

The previous section identified several value-addingfeatures of ontologies & reasoning. Such technologybenefits are tied to benefits for enterprise applicationsin the following. Figure 3 and the following subsec-tions explain which value-adding features are typicallyapplied to achieve each benefit for enterprise applica-tions and give several real-world examples.

4.1. New Business Scenarios

Explanation A business scenario is a description offuture business circumstances, imagined on the basisof past and present trends, uncertainties, and assump-tions. This word-picture depicts the anticipated inter-actions of external agents (such as competitors, cus-tomers, suppliers, and economic and regulatory envi-ronment) and aims to highlight the associated opportu-nities and threats.2

2http://www.businessdictionary.com/definition/business-scenario.html


Fig. 3. Technological value-adding features of ontologies & reasoning lead to benefits for enterprise applications.

Value-adding Features The crucial value-adding fea-tures in order to reach this benefit are Reuse and WebCompliance. The latter allows information to be pub-lished and queried on the Web. As soon as the levelof Reuse of an ontology is community or even world,the ontology and instances act as a global databasewith a common schema according to the principlesof Linked Data. This, paired with the value-addingfeature of Conceptual Modeling, i.e., classes, prop-erties, instances and rules in one coherent, easy-to-understand model, opens up hitherto impossible busi-ness scenarios as the examples in the following para-graph show.

Examples The first example concerns the BBC, i.e.,the largest broadcasting corporation in the world. TheBBC publishes large amounts of content online, suchas text, audio, and video about programmes or mu-sic. Web Compliance (Linked Data) is used to providecross-domain navigation and machine-readable repre-sentations. Giving access to machine-readable repre-sentations that hold links to further such representa-tions, crossing domain boundaries, means that muchricher applications can be built on top of BBC’s data,including new BBC products. The system gives BBCthe flexibility and a maintainability benefit: the website becomes BBC’s API. The RDF representations ofthese web identifiers allow third party developers to

use BBC’s data to build and monetize new applica-tions. Reuse is required for this new business scenarioand provided by the BBC Programme ontology. [54]

The second example is the Yahoo SearchMonkey[64] which enables a business model for improved re-sult presentation by specialized search engines. In thiscase Web Compliance, especially the feature of Anno-tation, allows Yahoo to harvest RDFa-enriched Webpages and to store the extracted RDF triples in its in-dex. In turn, third party developers can build their ownsearch engine with a development kit allowing accessto Yahoo’s search index. Reuse is required to enablethis scenario and given by an ontology prescribed bythe third party developer.

4.2. Increased Productivity of Software Engineering

Explanation A prominent target area of ontologies& reasoning has been software engineering, cf. [83],where an often sought benefit is the increase of pro-ductivity. The research community tried to increase theproductivity mainly in three ways. The first way is FullAutomation of Web service discovery and composi-tion. Activities in this realm are frequently dubbed Se-mantic Web Services [62]. Second, ontologies & rea-soning are also applied to more established softwareengineering in order to achieve Cost and Time Re-


duction. Third, Quality Improvements are also oftensought in software engineering.

Value-adding Features Full Automation or SemanticWeb Services first leverage Conceptual Modeling tocapture additional information for services, e.g., themeaning of inputs and outputs. This information isthen annotated to a Web service description (Web Com-pliance), e.g., by SA-WSDL. Semantic Web Servicesapproaches typically rely on the Formality of Descrip-tion Logics or Logic Programming languages in orderto exploit Reasoning capabilities for automating taskssuch as service discovery or composition.

Approaches that target Cost and Time Reductionor Quality Improvements typically capture relevant,software-related information by Conceptual Modeling.Here, Rules are often exploited in order to infer ad-ditional knowledge. Note that the level of Reuse isoften low since the ontology is mainly used withinan application for exploiting Reasoning functionality.Therefore, as in the case of Semantic Web Services, ap-proaches typically rely on the Formality of either LogicProgramming or Description Logic languages.

Examples An example for Full Automation in therealm of Semantic Web Services is [58]. They use aDescription Logic ontology as well as SubsumptionReasoning for Web service discovery. Both Web ser-vice descriptions and search requests are representedas classes with properties and axioms. SubsumptionReasoning then checks whether the search request issubsumable under, i.e., a subclass of, an existing webservice description. If this is the case, both match andthe corresponding Web service can fulfill the request.

An example for Cost and Time Reduction is [69].Here, the developer or administrator is supported intypical middleware management tasks, such as en-suring compatibility of library licenses. In order toachieve this, scattered information about softwarecomponents and accompanying information is inte-grated by Conceptual Modeling. The correspondingontology is formalized in a Logic Programming basedlanguage including simple Rules and Axioms. The rea-soning task of Instance Retrieval is used to obtain in-formation, e.g., a pair of libraries with incompatiblelicenses.

In our running example, building Oil and Gas ap-plications promises to take less time, require less cost,or feature improved quality — independent of whetherthe application actually uses ISO 15926. The basicidea of Semantic Web Services is also foreseen in the

Oil and Gas setting since an architecture for Web ser-vices and the use of ISO 15926 is explicitly mentioned.

4.3. Increased Productivity of Information Workers

Explanation Information workers (also known asknowledge workers) in today’s workforce are indi-viduals able to act and communicate with knowledgewithin a specific subject area. They use research skillsto define problems and to identify alternatives. Fueledby their expertise and insight, they work to solve thoseproblems, in an effort to influence company decisions,priorities and strategies. [24] Information workers mayalso be found in the Oil and Gas setting, e.g., an op-erator in an onshore control room. Efficient access touseful information should promote information workerproductivity by facilitating faster, higher quality deci-sion making. [6] According to [75], ontologies & rea-soning can be used for increasing productivity of in-formation workers by improving:

Visualization Capabilities Ontologies are used forimproving the appearance of the user interface,i.e., the way information if presented to the user.More concretely, [75] list Information Clustering,Text Generation, and Adaptation of UI Appear-ance as sub-categories.

Interaction Possibilities The user has to interact withan IT system by entering data, selecting items, is-suing commands, etc. According to [75], ontolo-gies can help to improve Browsing, Input Assis-tance, Providing Help, and UI Integration.

Value-adding Features The starting point for obtain-ing this benefit is the integration of all relevant infor-mation in a way that is easily consumable by the infor-mation worker (Conceptual Modeling). This typicallyincludes classes and taxonomies, properties, and in-stances (rules and axioms are of less importance in thiscase). The second crucial value-adding feature is Di-rect Interaction, i.e., the straightforward, user-friendlyway of viewing and editing the ontology and instancesis possible for the information worker. It is typical formost approaches that they require Reuse on the level ofcommunity, since a group of information workers co-operate on specific tasks. Some approaches in this cat-egory also exploit the Reasoning capabilities of LogicProgramming languages.


Examples A good example is ontology-enhancedbusiness intelligence such as presented in [86]. Here,an ontology conceptualizes the end user’s domain inan intuitive way. This could be ISO 15926 for captur-ing the Oil and Gas terminology. The operator of anonshore control room interacts with the business in-telligence solution via the ontology which can also bepersonalized and changed for his purposes. The on-tology shields the end user from IT systems’ intrica-cies by means of mapping legacy data structures to theontology for intuitive interaction. This is achieved byInstance Retrieval applying Rules.

4.4. Improved Enterprise Information Management

Explanation Enterprise Information Management (EIM)is a particular field of interest within the informa-tion technology area. It specializes in finding solu-tions for optimal use of information within organi-zations, for instance to support decision-making pro-cesses or day-to-day operations that require the avail-ability of knowledge. It tries to overcome traditionalIT-related barriers to managing information on an en-terprise level. A specific field of EIM is enterpriseknowledge management, i.e., the formal managementof knowledge resources in order to facilitate accessand reuse of knowledge, typically by using advancedinformation technology. [73]

Value-adding Features Conceptual Modeling is of-ten used for representing an integrated and easy-to-understand view of enterprise information. Surveysthat discuss this benefit can be found in [96,67]. Inmany cases, clarity in representing and storing en-terprise information is a must — semantic ambigui-ties, let alone conceptual inconsistencies, are likely tocause misunderstandings. To ensure sustainable mod-eling, building the ontology with Best Practices, suchas Foundational Ontologies (often formalized in First-Order logic), Ontology Design Patterns and QualityCriteria, can help to capture enterprise informationwith particular high quality. Also, modeling a domainonce and its enterprise-wide Reuse is one of the goalssought by EIM.

Flexibility is of importance because it allows to ac-cess and evolve the ontology programmatically overtime according to the enterprise’s representation needs.Ontology APIs with Evolution possibilities and Dy-namic Direct Programming allow to flexibly cope withinevitable schema changes in enterprise settings.

Web Compliance allows for standardized Publica-tion, Annotation and Querying of enterprise informa-

tion with corresponding tooling. This avoids informa-tion silos and enables seamless integration with infor-mation on the Web (in particular exploiting availablelinked data). Finally, Formality counters ambiguities ofenterprise information as discussed in Section 2.

Examples Disaster management is an example whereConceptual Modeling with Best Practices can help toimprove enterprise information management. In [7,8],we explain that with the many organizations involvedin the disaster, crossing regional or even national bor-ders, information exchange is crucial but cumbersomedue to differing vocabularies and representations bothat human language and IT level. Carefully designedontologies can be instrumental in addressing this prob-lem, by providing a reference model for humans, andwith that the basis for enterprise information manage-ment at the IT level. We devise an ontology stack cov-ering the description of damages (caused by the disas-ter), resources (available to organizations fighting thedisaster), and their connection (e.g., which resourcesare relevant for which damage). To ensure sustain-able modeling, we follow the guiding principles of thefoundational ontology DOLCE.

In our running example, it is one purpose of ISO15926 to facilitate and enable information integrationacross IT systems. The declared goal of the NorwegianOil and Gas industry is to enable information integra-tion based on ISO 15926 to support, e.g., the integra-tion of different data sources and sensors, the valida-tion of information delivered by different sources, andthe exchange of information within and between com-panies via Web services [95].

5. Challenges of Adopting Ontologies &Reasoning

After the benefits have been clarified and explained,this section draws attention to the challenges when ap-plying ontologies & reasoning. The challenges wereidentified during more than five years of industrial re-search and are mainly due to the often complex bound-ary conditions in an enterprise setting.

Some of the challenges also apply to the introduc-tion of other technologies. However, the severity of thechallenges increases for ontologies & reasoning whenmoving to the right on the axis depicted in Figure 4 .


Technical Integration

Technical Integration x n ?

Cost-benefit ratio

Modeling

How to measure the benefits?

Training

Fig. 4. Overview of challenges.

5.1. Cost-Benefit Ratio

The first challenge is to arrive at a positive overallcost-benefit ratio for each use case of ontologies & rea-soning in an enterprise setting. On the one hand, on-tologies & reasoning promise benefits for enterpriseapplications as discussed in Section 4. On the otherhand, the costs, e.g., caused by technical integration ormodeling, have to be summed up and contrasted to theenvisioned benefits. Here, the term Total Cost of Own-ership (TCO) [33] comes into play which is a finan-cial estimate. The purpose of TCO is to help enterprisemanagers determine direct and indirect costs of a prod-uct or system. TCO is a management accounting con-cept that can be used in full cost accounting or evenecological economics where it includes social costs.The total cost of ownership of a software applicationhas a correlation to the number of systems/componentsand different technologies. These are the main multi-pliers of TCO drivers in a software product. One wayof formalizing the TCO is shown below

TCO ∼ TCO drivers × ]of stacks in landscape ×]of technologies

where a TCO driver might be any administrative,operational or user specific action. More specifically,such a driver could be the required acquisition of ontol-ogy experts, training of existing employees, technicalintegration costs, modeling costs, maintenance, tech-nology buy-in or redevelopment costs. Each technol-ogy (but also inside the same technology) introducesdifferent tools, UIs, and system behaviors which allend up in a higher effort to install, learn, run, and main-tain software. The stacks in the landscape is best ex-plained by an example: with complete ABAP and Javasupport, the SAP landscape would feature two technol-ogy stacks. If ontology editors, stores, or reasoners areto be integrated in both stacks, the TCO will approx-imately rise by the factor of 2. The number of tech-nologies equals 1 if ontology language, reasoner, edi-

tor, and store are to be integrated once. If different rea-soners, editors, or stores are required, the number willrise accordingly.

Coming back to our running example in Section2, the Oil company might be forced to integrate ISO15926 in its existing IT landscape in order to be in-teroperable and compliant with partners. This use caseaffects the information worker and might require theintegration of a suitable ontology store and API inits existing IT landscape. However, the Oil companymight also feature a large IT department where cus-tom software is developed in house. In order to in-crease the productivity of software engineering, thecompany fancies to establish an ontology-supportedenvironment for its developers. This might require theintegration of additional — and potentially different— ontology store, reasoner and API. Therefore, the]of technologies would equal 2 because the Oil com-pany has to integrate different ontology stores andAPIs for two different use cases. If the technologystacks of the IT landscape and the development envi-ronment differ, the ]of stacks in landscape would alsoequal 2.

5.2. Training

Two of the TCO drivers mentioned in Section 5.1are the training of existing employees and the acquisi-tion of ontology experts. Both concern the challenge oftraining and are further discussed below.

Concerning the training of existing employees, theirexpertise typically lies in one specific technology.Even if they are willing to adopt and familiarize witha new technology or paradigm, educating them re-quires training costs. Usually, the training costs arevery high and managers are not willing to expend themunless there is a compelling business case. Here, theintricacies of ontology languages and reasoning tech-niques, especially description-logic-based, are note-worthy since they are particularly hard to understandfor a non logician. [32]


In the case of existing information workers, they,too, have to be trained if they are confronted with on-tology interpretation or editing. As an example, evenif the Oil company applies a potentially intuitive toolsuch as the Semantic Media Wiki, basic knowledgeabout the structure of an ontology is required to effi-ciently use the tool.

Acquiring ontology experts is also problematic sincethe technology is still rather new and not established inthe industry. Contrast this with the wide-spread natureof relational databases. If there is no convincing busi-ness case, an enterprise might decide to realize the usecase with conventional technologies, i.e., technologieswhere there is expertise readily available in the com-pany.

5.3. How to Measure the Benefits?

Calculating the TCO of an additional technologyrepresents an estimate and depends on probabilisticvariables. The calculation and determination of thebenefits is even less concrete. Therefore, the third chal-lenge lies in measuring the benefits what boils downto finding measures for proving that an ontology-basedapplication is beneficial compared to another solution.

The ontology and Semantic Web community hasbeen struggling to evaluate their contributions accord-ingly. Indeed, one hardly finds scientific methods ormeasures to prove the benefits. Other communitiesshare similar struggles, however. It is equally hard tomeasure the benefits for knowledge management ap-plications, conceptual modeling techniques in general,or object-oriented programming languages comparedto procedural languages.

Probably even harder than a sound scientific eval-uation is coming up with a business case. A businesscase captures the rationale for initiating a project ortask and is the basis of an enterprise manager’s deci-sion for investment. A business case is often presentedin a well-structured document, but may also sometimescome in the form of a short verbal argument or pre-sentation. The logic of the business case is that, when-ever resources or effort are consumed, they should bein support of a specific business need. As an exam-ple, consider the exemplary business need of an Oilcompany to increase the productivity of its in-house ITdepartment with custom software development by 5%thus saving several thousand Euros per year. A properbusiness case must answer to how this can be achieved

– under consideration of the cost-benefit ratio

– including a deployment plan of available (human)resources

– defining quantifiable success criteria– proposing an exit strategy– concerning the business capabilities and impact– specifying the required investment– including a project plan– in an adaptable way, meaning the proposal can be

tailored to size and risk.

5.4. Technical Integration

The fourth challenge concerns the technical integra-tion of the use case, and, in particular, its required on-tology language, editor, store, and reasoner. Technicalintegration means the required technology needs to beembedded in the existing landscape of the adopting en-terprise. As an example, consider an Oil company thatwants to incorporate ISO 15926 along a specific ontol-ogy store and API in its existing IT landscape. Adap-tors have to be written to legacy code and databases,versions of programming languages might have to beswitched, specific software components might have tobe replaced because of license incompatibilities, etc.The challenge typically increases with the size of thelegacy code and size of the enterprise’s portfolio. Thefollowing issues have to be addressed:

Build or buy? If the required technology is not read-ily available, the enterprise is first prompted todecide to buy the technology or even to considerredeveloping the technology on its own. Reasonsfor redevelopment might be immaturity of the ex-isting technology or license incompatibilities orstrategic investments. In the case of adopting ex-isting software, the challenges continue below. Aredevelopment is only possible if the required ex-pertise is at hand (cf. Section 5.2).

Maturity level of tools Most use cases of ontologies& reasoning so far emerged in the realm of sci-entific contributions. Scientific work usually endswith a simple prototype and is lacking the re-quired maturity. Most of the time the software isundocumented and not maintained. It requires aconsiderable effort to bring academic prototypesto the required maturity level. Often such projectsfail when the original developer or community isnot available. This might stop the whole project orlead to redevelopment on the enterprise’s behalf.

Enterprise scale performance Large enterprises of-ten feature heaps of data in the area of tera


bytes. The scientific contributions are often onlytested against toy examples and not thoroughlyrun through a set of real-world test cases withenterprise-scale amounts of data. Depending onthe ontology language, enterprise scale perfor-mance might not even be theoretically possibledue to the complexity of the language.

How to handle ontologies in the transport system?Business applications for large enterprises typi-cally feature a transport system for dealing withupdates [44]. When mission critical software is tobe updated, a test system is set up which can betransported automatically to a productive system.The transport happens at a specific point in time,e.g., when a test period has been run successfully.Often the test systems are even cascaded to a pre-productive system for detailed tests. So far it re-mains unclear how to deal with the flexible natureof ontologies in such software transport systems.

5.5. Technical Integration ×n?

Suppose several use cases of ontologies in softwareengineering are adopted by an enterprise. Each usecase might require a different ontology language possi-bly based on a different logic. That means different usecases might require different editors, stores, and rea-soners. In this case, the challenge of technical integra-tion might have to be faced×n. The situation is furthercomplicated when the different use cases are geared atdifferent beneficiaries. Some might benefit the devel-oper and some might benefit the end user, eventually.That means editor, reasoner, or store might have to beintegrated both in the developer run time or the appli-cation run time environment. Both environments mightbe completely different.

Continuing the example of Section 5.1, the Oil com-pany is prompted to apply ontologies & reasoning fortwo different use cases. Therefore, the hope is to min-imize n ideally to 1. That means, the same ontologylanguage, editor, store and reasoner can be used forboth use cases.

5.6. Modeling

5.6.1. Target AudienceThe lackluster adoption of ontologies & reasoning

can be attributed to the lack of tools that offer higherlevels of abstraction. The audience for most ontologyediting tools even today is experts in Conceptual Mod-eling with the required knowledge in the Formality

of ontology languages. However, this is not the audi-ence who is expected to create ontologies. Often, sub-ject matter experts of domains are the most suitablepeople for authoring ontologies. They usually have notechnical expertise in complex and deep computer sci-ence subjects. So, the entire tooling in ontology do-main catered to the wrong audience to start with asconcluded by [25,9].

This phenomenon is also apparent in the Oil & Gasscenario where special interest groups have to modeltheir corresponding domain of expertise (drilling andcompletion, operations and maintenance, or produc-tion and reservoir). However, the group members aretypically no ontology experts.

5.6.2. Upfront Modeling CostsAnother reason for lack of adoption is the need

to build ontologies upfront. This in combination withtools that do not cater to the right audience, makes itvery hard to get started with ontologies & reasoningin general. [85] present a specific cost model for en-gineering ontologies and identify the following costdrivers:

Product-related cost drivers account for the impactof the characteristics of the ontology on the over-all costs. Examples are the domain analysis com-plexity, the conceptualization complexity, or thedocumentation needs.

Personnel-related cost drivers emphasize the role ofteam experience, ability and continuity with re-spect to the effort required by an ontology devel-opment project (expert capability, level of exper-tise with respect to languages or tools, and thepersonnel turnover).

Project-related cost drivers related to the character-istics of the overall ontology development projectand their impact on the total costs. Relevant costdrivers include the level of support and automa-tion provided by ontology engineering software,and the additional efforts related to operating theproject in a decentralized environment.

Ontologies are particularly often attributed to beafflicted with large upfront modeling cost. Althoughtrue, this also holds for all related technologies thatprovide the Conceptual Modeling feature. Capturing arelevant domain in a relational database schema, UMLclass model or in knowledge-based systems requiresmodeling effort as well [31,87].


5.6.3. Use Case DependencyIndeed, if several use cases are to be applied, the

hope is that modeling a domain once and capturing itin an ontology will suffice for all use cases. However,different use cases typically have different represen-tation needs. Below, we show two typical phenomenathat cause this challenge. Note that the phenomena arenot idiosyncratic to ontology languages but are alsoknown from other conceptual modeling languages ordatabase schema design (normalization theory).

Regarding the first phenomenon, consider the do-main of Oil and Gas as a simple example. When mod-eling this domain, there might be the need to rep-resent information about equipment (such as pumpsand valves) and at which production plant it is de-ployed. For one use case it might be enough to rep-resent equipment and plant as classes and to estab-lish a simple relation between them to capture the de-ployment (cf. Figure 5 top). However, another use casemight have to capture information about the deploy-ment itself, e.g., its duration. Therefore, a simple re-lation between equipment and plant is not sufficientto meet this requirement since decidable ontology lan-guages typically do not allow attaching attributes andrelations to other relations. The deployment relationhas to be reified to a separate class (cf. Figure 5 bot-tom).

Equipment Plantdeployment

Equipment Deployment Plant

Fig. 5. Different use cases might require different representations ofthe same domain of discourse.

There is a second phenomenon that leads to differ-ent representations of the same domain. For example,the modeling applied by [60] to the domain of Web ser-vices is geared towards reasoning. That means model-ing decisions are taken in a way such that subsumptionreasoning can be applied to eventually discover webservices. In contrast, [71] models the same domain butfor the purpose of establishing a reference ontology.Here, modeling decisions are influenced by Best Prac-tices including ontology quality criteria such as Onto-Clean [41] leading to a different ontology.

According to our experience, ontologies are oftennot reused and built without consensus geared to a spe-

cific use case, e.g., for purely exploiting Reasoningfeatures and transparent to the user.

6. Case Studies

The following section highlights experience reportsfor two SAP Research case studies where ontologies& reasoning have been applied and also comes backto one particular realization of the Oil & Gas scenario.The experience reports highlight which value-addingfeatures, benefits, and challenges did occur in the cor-responding case study and which compromises havebeen taken.

The first case study, viz., FindGrid, is a success storyand indeed the first product of SAP that applies ontolo-gies & reasoning. The second case study, code-namedxCarrier, has been successfully prototyped and demon-strated to board members with positive feedback butproved to be too disruptive. The third case study isin fact a negative example where the challenges out-weighed the sought after benefits.

6.1. FindGrid

Problem Knowledge in enterprises is not easily ac-cessible since information is hidden in disparate loca-tions or in the heads of experts. There is typically noway to capture and reuse best practices and know-how.Established solutions, such as search engines, enter-prise content management systems, case managementsystems or collaboration spaces only solve parts of theissue.

Instead, enterprise memory should be built via acollaborative, self-learning environment that enablesknowledge capture across people, teams and the wholeenterprise. Information workers should be enabled toactivate, consolidate and summarize artifacts such asfolders, bookmarks, tags, notes, or pictures, thus cre-ating, harmonizing and sustaining the enterprise mem-ory.

Relevant Value-adding Features The first outcomeof semantic projects at SAP, code-named “FindGrid,”provides automated tool support for all of the above. Inessence, the goal is to increase the productivity of in-formation workers of SAP’s customers by acceleratedspeed to knowledge and new ideas. FindGrid is a solu-tion that supports business research activities, such asmarket or product research. A single interface allowsa product manager to search for information, organizefindings, and collaborate with others.


As can be seen in Figure 6, the solution relies ona Semantic Integration Layer. This layer manages theEnterprise Memory upon which Semantic Functions,such as search or auto-completion, are realized.

The following value-adding features are exploited:FindGrid relies on Flexibility since it requires the‘schema last’ feature to cope with building the enter-prise memory via a collaborative, self-learning envi-ronment. The schema is captured via Conceptual Mod-eling, in this particular case via an informal ontology(i.e., a semantic net [56]). The semantic net capturesinformation about cases, folders, associated terms, etc.The level of Reuse is on Community since all users im-plicitly share its conceptualization via FindGrid.

Fig. 6. FindGrid layered architecture.

Relevant Challenges When FindGrid was initiated,SAP did not own the required ontology store, so a de-cision had to be made with respect to Build or buy?(cf. challenge Technical Integration). For the first ver-sion of FindGrid, the decision was taken to license therequired ontology store with complementary servicesfor realizing the Semantic Functions.

Another major challenge concerned the Cost-BenefitRatio, especially the TCO. In the first version of Find-Grid, the ontology store has been integrated as ‘black

box,’ i.e., unavailable to other applications that requiresuch a store as well. The rationale behind this deci-sion was that developing enabling technology shouldnot hinder the development of the actual solution. Be-sides, this decision avoided the cost of integrating thestore in the comprehensive SAP technology platformin the first place and allowed faster prototyping.

Trade off However, SAP intends to develop furtherapplications that leverage the semantic integrationlayer. The ‘black box’ integration of the ontology storeand functionality in the first version of FindGrid wouldhave to be duplicated in each of the further applica-tions. Instead, the semantic integration layer has to be-come an integral part of SAP’s comprehensive tech-nology platform. The semantic integration layer willthus be incrementally factored out of FindGrid and re-developed on the basis of SAP’s in-memory databasesystem (HANA [29]).

6.2. xCarrier

Problem This case study concerns the need of SAP’scustomers to integrate multiple carrier services for dif-ferent destinations and goods to decrease transporta-tion costs. As an example, consider a manufacturingcompany that needs to ship its goods via the services ofcarriers such as UPS, DHL, or FedEx depending on di-mension, weight, or destination. An internal SAP pro-totype, called xCarrier [82], addressed B2B integra-tion on the technical level by introducing XCE (xCar-rier Enterprise) and XCC (xCarrier Connect) on top ofthe shipper and carrier backends, respectively (cf. Fig-ure 7). XCC provides normative Web service opera-tions for rate-lookup or tracking which are invoked byits counterpart XCE. Although a significant benefit tothe one-off manual integration, this internal SAP pro-totype still did not address the following problems:

1. How to describe carrier service products?2. How to find suitable carrier services?3. How to automatically select the appropriate car-

rier service products?

Relevant Value-adding features In order to addressthe first problem, Conceptual Modeling has been ap-plied to describe carrier service products via an OWL-DL ontology. Each carrier service product is repre-sented as a class with ship from and to properties tolocations. Further classes capture information aboutshipping items such as parcels, containers or docu-ments and many more. The second problem was ad-


Fig. 7. xCarrier setup. Circles in XCC (xCarrier Connect) representWeb service operation interfaces.

dressed by introducing another software component,viz., XCD (xCarrier Discovery), which allows carriersto publish their service descriptions via XCC and ship-pers to find suitable carrier service products. A Car-rier Manager was developed which applies the bene-fit of Direct Interaction allowing the definition of car-rier service products. The solution to the third prob-lem is based on [35], where we reduce selection anddiscovery to the Formality of Description Logics andsubsequent Subsumption Reasoning. [35] also allowsto define preferences in the ontology to specify whichcarrier service product is to be used depending on aspecific shipment request. This is done graphically inthe Shipper Manager. All this increased the productiv-ity of software engineering since selection, discoveryand preferences could be reduced to existing reasoningservices.

Web Compliance improves enterprise informationmanagement since SAP could prescribe a CommonOntology about carrier service products and publishit on www.sap.com, for instance. Carriers couldthen specialize the Common Ontology for their spe-cific representation needs and publish on their domain.Specific carrier service description can still be main-tained in XCD but reference their corresponding on-tologies. Our selection and discovery approach stillworks because of the open-world assumption of OWL-DL. WSDL descriptions of XCC can apply Annotationby applying SA-WSDL in order to link to their ontol-ogy. Similar holds for the XML messages that are pay-load of actual SOAP messages during runtime. Ship-pers may also specialize the Commmon Ontology torepresent additional information required for specify-ing their preferences.

The outlined setting could even lead to a New Busi-ness Scenario, e.g., when XCD is maintained by athird-party broker which still allows carriers to main-tain their own specializations of the common ontology.

Relevant Challenges One major challenge of thiscase study has been Training. The case study relies onthe approach in [35] which makes use of intricate de-scription logic capabilities. Since this could be under-stood only by the authors but not by the remaining soft-ware engineering team, the hurdle was too high. An-other challenge proved to be Enterprise scale perfor-mance (the reasoner has to process up to 20.000 auto-matic classifications per hour) and the Maturity levelof tools (no commercially proven description logic rea-soner has been available at that time). Another chal-lenge was to convince decision makers of a positiveCost-benefit Ratio compared to a solution with estab-lished technology. Also here, the Build or buy? chal-lenge was evident and led to many problems on theside of decision makers.

Trade off The dilemma between value-adding fea-tures and challenges was solved by not productizingxCarrier. Apart from the challenges there were otherreasons, e.g., lack of adoption of the carriers and orga-nizational obstacles. In other words, the proposed so-lution was too disruptive at that time.

6.3. Oil & Gas

Problem The Oil & Gas scenario in Section 2 re-quires a middleware solution for composing existingapplications and information repositories (such as ro-tating equipment monitoring, facility monitoring, en-gineering systems, asset management) to new appli-cations (such as drilling program planning, produc-tion optimization, equipment fault detection, or planturnaround). Such composite applications depend oninformation stemming from several existing IT sys-tems, and, thus, benefit from semantically unambigu-ous information exchange provided by ISO 15926 (cf.Figure 8).

One commercially availabble middleware solutionis given by [21] which however existed prior to the in-troduction of the ISO 15926 ontology. The middlewareprovides an enterprise service bus exchanging mes-sages via a set of existing standards in Oil & Gas. Thisset of standards is proprietarily represented via a con-ceptual model, i.e., [76], which is given in UML, how-ever. The problem was to incorporate support for ISO15926 in this existing environment.


Fig. 8. Middleware solution allowing the composition of existingapplications and information repositories (bottom) to new mash-ups(top). [65]

Relevant Value-adding features The main value-adding feature sought in this particular case is Reuse— the middleware has to pay tribute to the fact thatISO 15926 is standardized, and, thus, needs to supportthe standard.

Relevant Challenges The main challenges as indi-cated above has been the Technical Integration andCost-Benefit Ratio. The middleware solution, includ-ing its UML-based conceptual model, existed prior tothe task of integrating ISO 15926, so immense integra-tion costs would have to be justified in order to supportthe ontology.

Trade off The solution chosen was to remodel andadapt parts of ISO 15926 in a self-owned modeling en-vironment, viz., [57], which required manual effort andeventually led to the deviation of the standard. The de-viation was required to allow for scalability of storage[65].

7. Recommendations

This paper delineates the value-adding features ofontologies & reasoning and explains why these fea-tures are relevant to achieve specific business bene-fits. In addition, this paper identifies a set of challengeswhen trying to adopt ontologies & reasoning in a givenenterprise setting. The following discussion elicits po-tential trade-offs and recommends future research di-rections for each challenge.

Cost-Benefit Ratio A recent trend to lower TCO is touse semantic technologies as a service (cf., [26]). Ac-cording to this idea, users and developers can exploitontologies & reasoning, e.g., an ontology store, with-out the need for an own infrastructure, benefitting fromthe elastic scalability, flexibility, ease of deploymentand maintenance in the cloud. In this context, cloud

services present a viable alternative to in-house solu-tions promising significant reduction of cost and easeof use.

Training More research should be conducted in bet-ter integration with the software engineering commu-nity and their wide-spread and mature tools. A goodexample is OWL2Ecore [80] which enables softwareengineers to develop applications on the basis of an on-tology in their familiar environment, viz., Eclipse. Thisavoids that software engineers have to acquaint withontologies & reasoning.3

Another proposed remedy is to include the topic ofontologies & reasoning in standard curricula in com-puter science courses. Also, the Semantic Web com-munity should target publishing their research resultsin the ‘benefitting community.’ A good example is[70], where ontologies & reasoning are used to facil-itate management of application servers. Correspond-ingly, the contribution has been positioned in the mid-dleware community.

How to measure the benefits? A cost model for on-tology engineering published in [85] has been pro-posed in the Semantic Web community. Together withfirst efforts of measuring and comparing cost and bene-fits, e.g., in [17], this is an early but promising researchdirection which should be strengthened. The latter isbased on user satisfaction analysis which indicates if aparticular application meets its objectives.

With respect to defining a business case, the com-muniqué in [93] is a promising endeavor in that it as-sists in making the case for the use of ontology by pro-viding concrete application examples, value metrics,and advocacy strategies. The communiqué providestips, guidelines and strategies for making the case forontology to a variety of potential beneficiaries and fu-ture stakeholders. This communiqué specifically tar-gets ontology technology evangelists who already getthe value but want to overcome the blank stares theyget when trying to explain it to people who do not.

Technical Integration When ontologies & reasoningare applied in enterprises, they almost always haveto be integrated in existing IT landscapes with sev-eral boundary conditions. One stream of research, viz.,[74], addresses this challenge of technical integra-tion by introducing a reference architecture for us-

3This is not to be confused with exploiting the plethora of exist-ing UML editors for ontology engineering, e.g., [16], what servesprimarily the Semantic Web community.


ing ontologies in existing enterprise settings in a non-intrusive way.

Technical Integration ×n? The hope is that techni-cal integration is not required on a ‘per use case’ ba-sis but that software components such as ontology ed-itor, stores, and APIs can be factored out much likewhat happened with database management systems.Endeavors, such as the NeOn API and plug-ins, [92]have been a step in this direction.

Modeling One established area of countering thechallenge of Upfront Modeling Costs is ontologylearning. Although initial learning results require man-ual adaptations, the approach helps to bootstrap on-tologies as shown in [42], for instance. Another ideais to only model what is initially required and then ex-ploit the value-adding feature of Flexibility to evolvethe schema at run time.

[13] is a way to address Use Case Dependency bycontextualizing ontologies. Specific parts of the ontol-ogy can be kept local, when they do not represent ashared view but only a party’s view of a domain.

References

[1] Integrated operations and the oil & gas ontology. Technicalreport, The Norwegian Oil Industry Association (OLF), POSCCaesar Association (PCA), 2008.

[2] Reference architecture foundation for service oriented ar-chitecture 1.0. Committee Draft 2, OASIS, Oct 2009.http://docs.oasis-open.org/soa-rm/soa-ra/v1.0/soa-ra-cd-02.pdf.

[3] B. Adida, M. Birbeck, S. McCarron, and S. Pemberton. RDFain XHTML: Syntax and processing. Recommendation, W3C,Oct 2008. http://www.w3.org/TR/rdfa-syntax/.

[4] J. Angele. Ontoprise: Semantic web technologies at business.In C. Pautasso, C. Bussler, M. Walliser, S. Brantschen, M. Cal-isti, and T. Hempfling, editors, Emerging Web Services Tech-nology, Whitestein Series in Software Agent Technologies andAutonomic Computing, pages 1–2. Birkhäuser Basel, 2007.

[5] R. Angles and C. Gutierrez. Survey of graph database models.ACM Comput. Surv., 40:1:1–1:39, February 2008.

[6] S. Aral, E. Brynjolfsson, and M. V. Alstyne. Information, tech-nology and information worker productivity: Task level evi-dence. NBER Working Paper 13172, The National Bureau ofEconomic Research, Cambridge, MA, Jun 2007.

[7] G. Babitski, S. Bergweiler, O. Grebner, D. Oberle, H. Paul-heim, and F. Probst. Soknos - using semantic technologies indisaster management software. In G. Antoniou, M. Grobelnik,E. P. B. Simperl, B. Parsia, D. Plexousakis, P. D. Leenheer, andJ. Z. Pan, editors, The Semanic Web: Research and Applica-tions - 8th Extended Semantic Web Conference, ESWC 2011,Heraklion, Crete, Greece, May 29 - June 2, 2011, Proceedings,Part II, volume 6644 of Lecture Notes in Computer Science,pages 183–197. Springer, 2011.

[8] G. Babitski, F. Probst, J. Hoffmann, and D. Oberle. Ontol-ogy design for information integration in disaster management.In S. Fischer, E. Maehle, and R. Reischuk, editors, Informatik2009: Im Focus das Leben, Beiträge der 39. Jahrestagung derGesellschaft für Informatik e.V. (GI), 28.9.-2.10.2009, Lübeck,Proceedings, volume 154 of LNI, pages 3120–3134. GI, 2009.

[9] E. G. Barriocanal, M.-Á. Sicilia, and S. S. Alonso. Usabil-ity evaluation of ontology editors. Knowledge Organization,32(1):1–9, 2005.

[10] A. Bernstein and E. Kaufmann. Gino - a guided input nat-ural language ontology editor. In I. F. Cruz, S. Decker,D. Allemang, C. Preist, D. Schwabe, P. Mika, M. Uschold,and L. Aroyo, editors, The Semantic Web - ISWC 2006, 5thInternational Semantic Web Conference, ISWC 2006, Athens,GA, USA, November 5-9, 2006, Proceedings, volume 4273 ofLecture Notes in Computer Science, pages 144–157. Springer,2006.

[11] C. Bizer, T. Heath, and T. Berners-Lee. Linked data - the storyso far. Int. J. Semantic Web Inf. Syst., 5(3):1–22, 2009.

[12] G. Booch, J. E. Rumbaugh, and I. Jacobson. The unified mod-eling language user guide. J. Database Manag., 10(4):51–52,1999.

[13] P. Bouquet, F. Giunchiglia, F. van Harmelen, L. Serafini, andH. Stuckenschmidt. Contextualizing ontologies. J. Web Sem.,1(4):325–343, 2004.

[14] P. Brèche. Advanced principles for changing schemas of objectdatabases. In P. Constantopoulos, J. Mylopoulos, and Y. Vassil-iou, editors, Advances Information System Engineering, 8th In-ternational Conference, CAiSE’96, Heraklion, Crete, Greece,May 20-24, 1996, Proceedings, volume 1080 of Lecture Notesin Computer Science, pages 476–495. Springer, 1996.

[15] D. Brickley and R. Guha. RDF vocabulary description lan-guage 1.0: RDF Schema. Recommendation, W3C, Feb 2004.http://www.w3.org/TR/rdf-schema/.

[16] S. Brockmans, R. Volz, A. Eberhart, and P. Löffler. Visualmodeling of owl dl ontologies using uml. In S. A. McIlraith,D. Plexousakis, and F. van Harmelen, editors, The SemanticWeb - ISWC 2004: Third International Semantic Web Confer-ence,Hiroshima, Japan, November 7-11, 2004. Proceedings,volume 3298 of Lecture Notes in Computer Science, pages198–213. Springer, 2004.

[17] T. Bürger and E. P. B. Simperl. Measuring the benefits ofontologies. In R. Meersman, Z. Tari, and P. Herrero, edi-tors, On the Move to Meaningful Internet Systems: OTM 2008Workshops, OTM Confederated International Workshops andPosters, ADI, AWeSoMe, COMBEK, EI2N, IWSSA, MONET,OnToContent + QSI, ORM, PerSys, RDDS, SEMELS, andSWWS 2008, Monterrey, Mexico, November 9-14, 2008. Pro-ceedings, volume 5333 of Lecture Notes in Computer Science,pages 584–594. Springer, 2008.

[18] H. C. Chan, K. K. Wei, and K. Siau. Conceptual level ver-sus logical level user-database interaction. In J. I. DeGross,I. Benbasat, G. DeSanctis, and C. M. Beath, editors, Proceed-ings of the International Conference on Information Systems,ICIS 1991, December 16-18, 1991, New York, NY, USA, pages29–40. Association for Information Systems, 1991.

[19] P. P. Chen. The entity-relationship model - toward a unifiedview of data. ACM Trans. Database Syst., 1(1):9–36, 1976.

[20] F. Chum. Semantic technologies at the ecosystem level. InP. Horowitz, editor, Semantic Web in the Enterprise, volumeSpring of Technology Forecast. PwC, 2009.


[21] R. Credle, V. Akibola, V. Karna, D. Panneerselvam, R. Pil-lai, and S. Prasad. Discovering the business value patternsof chemical and petroleum integrated information framework.Red Book SG24-7735-00, IBM, August 2009.

[22] F. Dau. Semantic technologies for enterprises. In S. Andrews,S. Polovina, R. Hill, and B. Akhgar, editors, Conceptual Struc-tures for Discovering Knowledge - 19th International Confer-ence on Conceptual Structures, ICCS 2011, Derby, UK, July25-29, 2011. Proceedings, volume 6828 of Lecture Notes inComputer Science, pages 1–18. Springer, 2011.

[23] S. Decker, S. Melnik, F. van Harmelen, D. Fensel, M. C. A.Klein, J. Broekstra, M. Erdmann, and I. Horrocks. The seman-tic web: The roles of XML and RDF. IEEE Internet Comput-ing, 4(5):63–74, 2000.

[24] P. F. Drucker. Landmarks of tomorrow: a report on the new“post-modern” world. Harper Colophon, 1965.

[25] A. J. Duineveld, R. Stoter, M. R. Kenepa, and V. R. Ben-jamins. Wondertools? a comparative study of ontological engi-neering tools. International Journal of Human-Computer Stud-ies, 52(6):1111–1133, 2000.

[26] A. Eberhart, P. Haase, D. Oberle, and V. Zacharias. Semantictechnologies and cloud computing. In D. Fensel, editor, Foun-dations for the Web of Information and Services, pages 239–251. Springer, 2011.

[27] E. Falkenberg, W. Hesse, P. Lindgreen, B. Nilsson, J. Oei,C. Rolland, R. Stamper, F. V. Assche, A. Verrijn-Stuart, andK. Voss. FRISCO — a framework of information system con-cepts — the FRISCO report. Technical report, IFIP WG 8.1Task Group FRISCO, 1998.

[28] D. C. Fallside and P. Walmsley. XML Schema part 0: Primersecond edition. Recommendation, W3C, Oct 2004. http://www.w3.org/TR/xmlschema-0/.

[29] F. Färber, S. K. Cha, J. Primsch, C. Bornhövd, S. Sigg, andW. Lehner. SAP HANA database: data management for mod-ern business applications. SIGMOD Rec., 40(4):45–51, Jan.2012.

[30] J. Farrell and H. Lausen. Semantic annotations for WSDL andXML Schema. Recommendation, W3C, Aug 2007. http://www.w3.org/TR/sawsdl/.

[31] A. Felfernig. Effort estimation for knowledge-based configu-ration systems. In F. Maurer and G. Ruhe, editors, Proceed-ings of the Sixteenth International Conference on Software En-gineering & Knowledge Engineering (SEKE’2004), Banff, Al-berta, Canada, June 20-24, 2004, pages 148–154, 2004.

[32] D. Fensel, M.-C. Rousset, and S. Decker. Workshop oncomparing description and frame logics. Data Knowl. Eng.,25(3):347–352, 1998.

[33] B. Ferrin and R. E. Plank. Total cost of ownership models:An exploratory study. Journal of Supply Chain Management,38(3), 2002.

[34] C. Forgy. Rete: A fast algorithm for the many patterns/manyobjects match problem. Artif. Intell., 19(1):17–37, 1982.

[35] A. Friesen and K. Namiri. Towards semantic service selectionfor B2B integration. In N. Koch and L. Olsina, editors, Work-shop Proceedings of the 6th International Conference on WebEngineering, ICWE 2006, Palo Alto, California, USA, July 11-14, 2006, volume 155 of ACM International Conference Pro-ceeding Series, page 17. ACM, 2006.

[36] E. Gamma, R. Helm, R. E. Johnson, and J. M. Vlissides.Design patterns: Abstraction and reuse of object-oriented de-sign. In O. Nierstrasz, editor, ECOOP’93 - Object-Oriented

Programming, 7th European Conference, Kaiserslautern, Ger-many, July 26-30, 1993, Proceedings, volume 707 of LectureNotes in Computer Science, pages 406–431. Springer, 1993.

[37] A. Gangemi, N. Guarino, C. Masolo, A. Oltramari, andL. Schneider. Sweetening ontologies with DOLCE. InA. Gómez-Pérez and V. R. Benjamins, editors, Knowledge En-gineering and Knowledge Management. Ontologies and theSemantic Web, 13th International Conference, EKAW 2002,Siguenza, Spain, October 1-4, 2002, Proceedings, volume2473 of Lecture Notes in Computer Science, pages 166–181.Springer, 2002.

[38] K. Gomadam, A. Ranabahu, and A. Sheth. SA-REST: Se-mantic annotation of web resources. Member submission,W3C, Apr 2010. http://www.w3.org/Submission/SA-REST/.

[39] M. Grüninger and J. Lee. Ontology applications and design -introduction. Commun. ACM, 45(2):39–41, 2002.

[40] N. Guarino, D. Oberle, and S. Staab. What is an ontology?In S. Staab and R. Studer, editors, Handbook on Ontologies,Handbooks on Information Systems. Springer, 2nd edition,2009.

[41] N. Guarino and C. A. Welty. Evaluating ontological decisionswith OntoClean. Commun. ACM, 45(2):61–65, 2002.

[42] H. Guo, A. Ivan, R. Akkiraju, and R. Goodwin. Learningontologies to improve the quality of automatic web servicematching. In 2007 IEEE International Conference on Web Ser-vices (ICWS 2007), July 9-13, 2007, Salt Lake City, Utah, USA,pages 118–125. IEEE Computer Society, 2007.

[43] M. Hepp. Ontologies: State of the art, business potential, andgrand challenges. In M. Hepp, P. D. Leenheer, A. de Moor,and Y. Sure, editors, Ontology Management, Semantic Web, Se-mantic Web Services, and Business Applications, volume 7 ofSemantic Web And Beyond Computing for Human Experience,pages 3–22. Springer, 2008.

[44] J. A. Hernandez, J. Keogh, and F. Martinez. SAP R/3 Hand-book, chapter 6, pages 215–270. McGraw-Hill/Osborne, thirdedition, 2006.

[45] A. R. Hevner, S. T. March, J. Park, and S. Ram. Design sciencein information systems research. MIS Quarterly, 28(1):75–105, 2004.

[46] P. Horowitz, editor. Semantic Web in the Enterprise, volumeSpring of Technology Forecast. PwC, 2009.

[47] B. Inmon. The semantics mystery. BeyeNETWORK,Article no. 4605, http://www.b-eye-network.com/print/4605, Jun 2007.

[48] A. Katifori, C. Halatsis, G. Lepouras, C. Vassilakis, and E. Gi-annopoulou. Ontology visualization methods — a survey.ACM Comput. Surv., 39(4), Nov. 2007.

[49] S. Kent. Model driven engineering. In M. Butler, L. Petre, andK. Sere, editors, Integrated Formal Methods, volume 2335 ofLecture Notes in Computer Science, pages 286–298. SpringerBerlin / Heidelberg, 2002.

[50] N. D. Kho. Semantics in practice. EContent — Digital ContentStrategies & Resources, http://www.econtentmag.com/?ArticleID=70970, Nov 2010.

[51] M. Kifer, G. Lausen, and J. Wu. Logical foundations of object-oriented and frame-based languages. J. ACM, 42(4):741–843,1995.

[52] W. Kim. Object-oriented databases: Definition and researchdirections. IEEE Transactions on Knowledge and Data Engi-neering, 2:327–341, 1990.


[53] J. W. Kluewer, M. G. Skjæveland, and M. Valen-Sendstad. ISO15926 templates and the semantic web. In W3C Workshop onSemantic Web in Oil & Gas Industry, Houston, TX, USA, 9–10December, 2008, 2008.

[54] G. Kobilarov, T. Scott, Y. Raimond, S. Oliver, C. Sizemore,M. Smethurst, C. Bizer, and R. Lee. Media meets semanticweb - how the BBC uses DBpedia and linked data to makeconnections. In L. Aroyo, P. Traverso, F. Ciravegna, P. Cimi-ano, T. Heath, E. Hyvönen, R. Mizoguchi, E. Oren, M. Sabou,and E. P. B. Simperl, editors, The Semantic Web: Research andApplications, 6th European Semantic Web Conference, ESWC2009, Heraklion, Crete, Greece, May 31-June 4, 2009, Pro-ceedings, volume 5554 of Lecture Notes in Computer Science,pages 723–737. Springer, 2009.

[55] M. Krötzsch, D. Vrandecic, and M. Völkel. Semantic Me-diaWiki. In I. F. Cruz, S. Decker, D. Allemang, C. Preist,D. Schwabe, P. Mika, M. Uschold, and L. Aroyo, editors, TheSemantic Web - ISWC 2006, 5th International Semantic WebConference, ISWC 2006, Athens, GA, USA, November 5-9,2006, Proceedings, volume 4273 of Lecture Notes in ComputerScience, pages 935–942. Springer, 2006.

[56] F. Lehmann. Semantic networks. Computers & Mathematicswith Applications, 23(2):1–50, Mar 1992.

[57] D. Leroux, M. Nally, and K. Hussey. Rational software archi-tect: A tool for domain-specific modeling. IBM Systems Jour-nal, 45(3):555–568, 2006.

[58] L. Li and I. Horrocks. A Software Framework for Match-making Based on Semantic Web Technology. In Proceed-ings of the Twelfth International World Wide Web Conference,WWW2003, Budapest, Hungary, 20-24 May 2003, pages 331–339. ACM, 2003.

[59] C. M. MacKenzie, K. Laskey, F. McCabe, P. F. Brown, andR. Metz. Reference model for service oriented architecture 1.0.Oasis standard, OASIS, Oct 2006.

[60] D. L. Martin, M. H. Burstein, D. V. McDermott, S. A. McIl-raith, M. Paolucci, K. P. Sycara, D. L. McGuinness, E. Sirin,and N. Srinivasan. Bringing semantics to web services withOWL-S. In C. L. Williamson, M. E. Zurko, P. F. Patel-Schneider, and P. J. Shenoy, editors, Proceedings of the 16thInternational Conference on World Wide Web, WWW 2007,Banff, Alberta, Canada, May 8-12, 2007, pages 243–277.ACM, 2007.

[61] D. L. McGuinness and F. van Harmelen. OWL Web Ontol-ogy Language Overview. Recommendation, W3C, Feb 2004.http://www.w3.org/TR/owl-features/.

[62] S. A. McIlraith, T. C. Son, and H. Zeng. Semantic web ser-vices. IEEE Intelligent Systems, 16(2):46–53, 2001.

[63] T. Menzies. Cost benefits of ontologies. Intelligence, 10(3):26–32, 1999.

[64] P. Mika. Year of the monkey: Lessons from the first year ofSearchMonkey. In S. Fischer, E. Maehle, and R. Reischuk,editors, Informatik 2009: Im Focus das Leben, Beiträge der 39.Jahrestagung der Gesellschaft für Informatik e.V. (GI), 28.9.-2.10.2009, Lübeck, Proceedings, volume 154 of LNI, page 387.GI, 2009.

[65] F. Myren, U. Pletat, and J. W. Kluewer. Model driven integra-tion architecture for IO G2 information – reference semanticmodel alignment to ISO 15926. Semantic Days 2009, 18 - 20May, Clarion Hotel Stavanger, Session 6: Semantic technologyfor IO Generation 2, 2009.

[66] D. Norman. Cognitive engineering. In D. Norman and

S. Daper, editors, User Centered System Design, pages 31–61.Lawrence Erlbaum Associates, Hillsdale, New Jersey, USA,1986.

[67] N. F. Noy. Semantic integration: a survey of ontology-basedapproaches. SIGMOD Rec., 33:65–70, December 2004.

[68] N. F. Noy and M. C. A. Klein. Ontology evolution: Not thesame as schema evolution. Knowl. Inf. Syst., 6(4):428–440,2004.

[69] D. Oberle. Semantic Management of Middleware, volume 1 ofSemantic Web And Beyond Computing for Human Experience.Springer, 2006.

[70] D. Oberle, A. Eberhart, S. Staab, and R. Volz. Developing andManaging Software Components in an Ontology-based Appli-cation Server. In H.-A. Jacobsen, editor, Middleware 2004,ACM/IFIP/USENIX 5th International Middleware Conference,Toronto, Ontario, Canada, volume 3231 of LNCS, pages 459–478. Springer, 2004.

[71] D. Oberle, S. Lamparter, S. Grimm, D. Vrandecic, S. Staab,and A. Gangemi. Towards ontologies for formalizing modular-ization and communication in large software systems. AppliedOntology, 1(2):163–202, 2006.

[72] D. Oberle, S. Staab, and R. Volz. Three Dimensions of Knowl-edge Representation in WonderWeb. Künstliche Intelligenz,1:31–35, 2005.

[73] D. E. O’Leary. Enterprise knowledge management. IEEEComputer, 31(3):54–61, 1998.

[74] H. Paulheim, D. Oberle, R. Plendl, and F. Probst. An architec-ture for information exchange based on reference models. InSoftware Language Engineering - Fourth International Con-ference, SLE 2011, Praga, Portugal, July 3-4, 2011, LectureNotes in Computer Science. Springer, 2011.

[75] H. Paulheim and F. Probst. Ontology-enhanced user interfaces:A survey. Int. J. Semantic Web Inf. Syst., 6(2):36–59, 2010.

[76] U. Pletat and V. Narayan. Towards an upper ontology for rep-resenting oil & gas enterprises. In W3C Workshop on SemanticWeb in Oil & Gas Industry, Houston, TX, USA, 9–10 Decem-ber, 2008, 2008.

[77] V. Presutti and A. Gangemi. Content ontology design pat-terns as practical building blocks for web ontologies. In Q. Li,S. Spaccapietra, E. S. K. Yu, and A. Olivé, editors, Concep-tual Modeling - ER 2008, 27th International Conference onConceptual Modeling, Barcelona, Spain, October 20-24, 2008.Proceedings, volume 5231 of Lecture Notes in Computer Sci-ence, pages 128–141. Springer, 2008.

[78] E. Prud’hommeaux and A. Seaborne. SPARQL query languagefor RDF. Recommendation, W3C, Jan 2008. http://www.w3.org/TR/rdf-sparql-query/.

[79] C. Puleston, B. Parsia, J. Cunningham, and A. L. Rector. Inte-grating object-oriented and ontological representations: A casestudy in Java and OWL. In A. P. Sheth, S. Staab, M. Dean,M. Paolucci, D. Maynard, T. W. Finin, and K. Thirunarayan,editors, The Semantic Web - ISWC 2008, 7th International Se-mantic Web Conference, ISWC 2008, Karlsruhe, Germany, Oc-tober 26-30, 2008. Proceedings, volume 5318 of Lecture Notesin Computer Science, pages 130–145. Springer, 2008.

[80] T. Rahmani, D. Oberle, and M. Dahms. An adjustable trans-formation from OWL to Ecore. In D. C. Petriu, N. Rouquette,and Ø. Haugen, editors, Model Driven Engineering Languagesand Systems - 13th International Conference, MODELS 2010,Oslo, Norway, October 3-8, 2010, Proceedings, Part II, volume6395 of Lecture Notes in Computer Science, pages 243–257.


Springer, 2010.[81] G. Reggio, M. Cerioli, and E. Astesiano. Towards a rigor-

ous semantics of UML supporting its multiview approach. InH. Hussmann, editor, Fundamental Approaches to SoftwareEngineering, volume 2029 of Lecture Notes in Computer Sci-ence, pages 171–186. Springer Berlin / Heidelberg, 2001.

[82] H. Roggenkemper, R. Ehret, and A. Tönne. Capture accuratesolution requirements the first time with exploratory modeling(xM): How to nearly eliminate post-go-live application designfailures. SAP Professional, 10(1), JAN 2008.

[83] F. Ruiz and J. R. Hilera. Using ontologies in software engi-neering and technology. In C. Calero, F. Ruiz, and M. Piat-tini, editors, Ontologies for Software Engineering and SoftwareTechnology, pages 49–102. Springer Berlin Heidelberg, 2006.

[84] S. Schaffert, A. Gruber, and R. Westenthaler. A semantic wikifor collaborative knowledge formation. In Proceedings of SE-MANTICS 2005, Vienna, Austria, 2006. Trauner Verlag.

[85] E. P. B. Simperl, C. Tempich, and Y. Sure. A cost estima-tion model for ontology engineering. In I. F. Cruz, S. Decker,D. Allemang, C. Preist, D. Schwabe, P. Mika, M. Uschold,and L. Aroyo, editors, The Semantic Web - ISWC 2006, 5thInternational Semantic Web Conference, ISWC 2006, Athens,GA, USA, November 5-9, 2006, Proceedings, volume 4273 ofLecture Notes in Computer Science, pages 625–639. Springer,2006.

[86] M. Spahn, J. Kleb, S. Grimm, and S. Scheidl. Supporting busi-ness intelligence by providing ontology-based end-user infor-mation self-service. In Proceedings of the First InternationalWorkshop on Ontology-supported Business Intelligence OBI’08, pages 10:1–10:12. ACM, 2008.

[87] P. Spyns, R. Meersman, and M. Jarrar. Data modelling versusontology engineering. SIGMOD Record, 31(4):12–17, 2002.

[88] A. Stark, M. Schroll, and J. Hafkesbrink. Die Zukunft des Se-mantic Web. Think!innowise Trend Report, 2009.

[89] K. Steiner, W. Eßmayr, and R. Wagner. Topic maps - anenabling technology for knowledge management. In A. M.Tjoa and R. Wagner, editors, 12th International Workshop onDatabase and Expert Systems Applications (DEXA 2001), 3-7 September 2001, Munich, Germany, pages 472–476. IEEEComputer Society, 2001.

[90] L. Stojanovic, A. Maedche, B. Motik, and N. Stojanovic. User-driven ontology evolution management. In A. Gómez-Pérezand V. R. Benjamins, editors, Knowledge Engineering and

Knowledge Management. Ontologies and the Semantic Web,13th International Conference, EKAW 2002, Siguenza, Spain,October 1-4, 2002, Proceedings, volume 2473 of Lecture Notesin Computer Science, pages 285–300. Springer, 2002.

[91] C. Tempich, H. S. Pinto, Y. Sure, and S. Staab. An argumen-tation ontology for distributed, loosely-controlled and evolv-ing engineering processes of ontologies (DILIGENT). InA. Gómez-Pérez and J. Euzenat, editors, The Semantic Web:Research and Applications, Second European Semantic WebConference, ESWC 2005, Heraklion, Crete, Greece, May 29 -June 1, 2005, Proceedings, volume 3532 of Lecture Notes inComputer Science. Springer.

[92] T. Tran, P. Haase, H. Lewen, Ó. Muñoz-García, A. Gómez-Pérez, and R. Studer. Lifecycle-support in architectures forontology-based information systems. In K. Aberer, K.-S. Choi,N. F. Noy, D. Allemang, K.-I. Lee, L. J. B. Nixon, J. Gol-beck, P. Mika, D. Maynard, R. Mizoguchi, G. Schreiber, andP. Cudré-Mauroux, editors, The Semantic Web, 6th Interna-tional Semantic Web Conference, 2nd Asian Semantic WebConference, ISWC 2007 + ASWC 2007, Busan, Korea, Novem-ber 11-15, 2007, volume 4825 of Lecture Notes in ComputerScience, pages 508–522. Springer, 2007.

[93] M. Uschold. Making the case for ontology. Appl. Ontol.,6(4):377–385, Dec. 2011.

[94] M. Uschold and M. Gruninger. Ontologies and semantics forseamless connectivity. SIGMOD Rec., 33:58–64, December2004.

[95] F. Verhelst, F. Myren, P. Rylandsholm, I. Svensson, A. Waaler,T. Skramstad, J. Ornæs, B. Tvedt, and J. Høydal. Digital Plat-form for the Next Generation IO: A Prerequisite for the HighNorth. In SPE Intelligent Energy Conference and Exhibition,2010.

[96] H. Wache, T. Vögele, U. Visser, H. Stuckenschmidt, G. Schus-ter, H. Neumann, and S. Hübner. Ontology-based integration ofinformation - a survey of existing approaches. In Proceedingsof IJCAI-01 Workshop: Ontologies and Information Sharing,Seattle, WA, 2001, pages 108–117, 2001.

[97] D. West. What semantic technology means to application de-velopment professionals. Forrester Research Report, Oct 2009.

[98] A. White. Semantic web moving ever close to the ‘semanticenterprise’? Gartner Blog Network, http://blogs.gartner.com/andrew_white/2009/04/30, Apr2009.

Documents

1 IOS Press Ontologies & Reasoning in Enterprise ApplicationsSAP Research Karlsruhe, Vincenz-Priessnitz-Str. 1, 76131 Karlsruhe, Germany, E-mail: [email protected] Abstract. This survey