Semantic Web Services, Part 2

www.computer.org/intelligent

David Martin, John Domingue, Amit Sheth, Steve Battle, Katia Sycara, and Dieter Fensel

Vol. 22, No. 6 November/December 2007

This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.

© 2007 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

For more information, please see www.ieee.org/portal/pages/about/documentation/copyright/polilink.html. Trends & Controversies

Semantic Web Services, Part 2

David Martin, SRI International John Domingue, Knowledge Media Institute, Open University

taken the first concrete steps toward building a SWS- n part 2 of this Trends & Controversies installment, we continue based solution: the W3C proposed recommendation exploring the state of the art, current practices, and future direc- SAWSDL (www.w3.org/2002/ws/sawsdl), its associated I tions for Semantic Web services. SWS aims to bring Semantic Web tools and use cases, and initial applications.1 Where do we technology—for representing, sharing, and reasoning about know- go from here? The researchers among us might be fully ledge—to bear in Web service contexts. The objective is to enable a convinced of the importance and benefit of adding seman- fuller, more flexible automation of service provision and use and the tics to Web services and impatient to see their research construction of more powerful tools and methodologies for working translated into technologies and adapted for real use. How- with services. For an overview of SWS, please see our introduction to ever, I believe we might need to do a lot more work before part 1 in the September/October 2007 issue. That introduction also we see substantial adoption of SWS technologies. includes references for major SWS initiatives, such as SAWSDL, OWL-S. First, we must be patient. Both Web services and the WSMO, SWSF, and the Internet Reasoning Service. Part 1 also includes Semantic Web are prerequisites for SWS, at least in the essays by Michael L. Brodie and Frank Leymann that discuss service tech- form in which SAWSDL or its extensions could evolve. Des- nology needs from a long-term industry perspective. In this issue, we pite a clear uptick in adoption of Web services by industry, conclude with four more essays. initial success stories, and early adoption of Semantic Web The first two essays are primarily concerned with nearer-term direc- technologies,2 broad acceptance and adoption could take tions—steps that will let us build out from the current state of the art another three years or so. For the Web services part, the rea- toward greater adoption and applicability of SWS approaches. Amit son is the complexity and confusion resulting from WS-* Sheth lays out a near-term roadmap of steps that will be essential for specifications, some of which (such as security and policy/ industry acceptance of SWS approaches, starting from SAWSDL and cur- agreement) are critical and either not yet mature or yet to be rent industry practice. (See the “Frequently Used Acronyms” sidebar for broadly adopted. For the Semantic Web, the challenges are explanations of SAWDSL and other terms.) Among other things, he coun- to convince practitioners sels that essential steps are required to make SWS approaches sufficiently accessible and economically attractive to industry. Steve Battle • that ontologies can be and have been built and are starts with an analysis of OWL-S’s strengths and limitations. He then manageable, discusses the necessary evolution of business ontologies for SWS. Along • that we can do much with limited semantics and with- with the evolution of business practices, this will allow for Web services out becoming experts in description logic, and SWS approaches to come together. • that we don’t need a single ontology for everything and The final two essays put forward longer-term agendas for the evolu- we know how to work reasonably well in a multi-ontology tion of SWS. Katia Sycara argues that SWS could benefit from decoup- environment, and ling itself from the basic stack of Web service standards rather than fol- • that technology is ready for building robust applications. lowing a more incremental trajectory tied to their evolution. She also identifies two important opportunities in which this strategy could pay We might also have to wait out the current Web 2.0 eu- off. Dieter Fensel takes a broad perspective, arguing that the character- phoria, bordering on infatuation. Clearly, Web 2.0 is se- istics of Internet-scale service usage, and problem solving in general, call ductive, easy to use and adopt, and highly valuable for for an entirely new conceptualization of some of the core challenges of some purposes, but it isn’t a solution to all problems and computer science for the 21st century. requirements. Technologies such as SWS won’t get at- —David Martin and John Domingue tention until there is a broader realization of what problems Web 2.0 can’t solve and why Web 3.0 (the new code for Semantic Web technologies-enabled Web) is needed. Beyond SAWSDL: A Game Plan for Broader Second, we must do a better job of understanding, ex- Adoption of Semantic Web Services plaining, and managing the complexity of using SWS. Take, Amit Sheth, Kno.e.sis Center, Wright State University for example, the SWS Challenge (http://sws-challenge.org), in which six teams of researchers have implemented or tried After a flurry of research led by the OWL-S, WSMO, to implement a solution to a realistic but relatively simple SWSF, and WSDL-S groups, the SWS community has application using widely different approaches, formalisms,

8 1541-1672/07/$25.00 © 2007 IEEE IEEE INTELLIGENT SYSTEMS Published by the IEEE Computer Society techniques, and technologies. Only one team input of the subsequent used SAWSDL. In SAWSDL’s defense, this chal- task or tasks). Clearly Frequently Used Acronyms lenge started well before it became a pro- demonstrating how posed recommendation. But a WSDL-S sub- SWS can ease this or BPEL: Business Process Execution Language mission to W3C (see www.w3.org/Submission/ equally vexing problems DAML-S: DARPA Agent Markup Language for Services WSDL-S), which became the starting point that process developers OWL: Web Ontology Language OWL-S: Web Ontology Language for Services for W3C’s work toward SAWSDL, has contin- face will be important. ued at the same time. More important, the Today’s enterprise ser- RDF: Resource Description Framework exercise, which has already involved four vice bus, application RDFS: RDF Schema meetings and spanned well over a year, server, and service- SAWSDL: Semantic Annotations for WSDL shows how complex the challenges are, es- based middleware ven- SWSF: Semantic Web Services Framework pecially when the problem involves legacy dors claim to solve the SOA: Service-Oriented Architecture systems or interfaces for preexisting ser- data-mediation problem UDDI: Universal Description Discovery and Integration vices. Lessons from this exercise lead to using XSLT (Extensible WSDL: Web Services Description Language two observations related to technical issues Style Sheet Language WSMO: Web Services Modeling Ontology (tools and methodologies) and business Transformation)-based issues (ROI). transformation. Al- Before practitioners (developers and real- though this addresses numerous basic • Supporting precondition and effects in word users) will embrace SWS, they must problems, as my student Karthik Goma- SAWSDL. OWL-S and others have noted have robust tools as well as methodologies dam put it, “the idea of mediating at the this need for some time, and more re- to streamline all aspects of managing the level of instance or between two fixed cently clear use cases have emphasized SWS life cycle, including annotation, publi- schemas is hackneyed.” To accelerate this a need for such support. cation, discovery, data mediation, composi- technology’s adoption, we must focus on • Semantic annotation of policy description or configuration, orchestration, and exe- issues such as data and process mediation tions. From a business perspective, this cution. In the context of SAWSDL, initial where using semantics will likely provide will be very important, especially to sup- versions of most of these have been devel- solid cases of ROI. Efforts that use semantic port nonfunctional or quality-of-service oped—some by the LSDIS (Large Scale and ontological approaches for interoperabil- requirements. Distributed Information Systems) lab (see ity and data mediation3 will need to be put www.iswc2006.semanticweb.org/projects/ side by side with current nonsemantic ap- In addition, we intend to collaborate meteor-s/SAWSDL), Kno.e.sis Center proaches to drive the point home. with others who are developing an on- (see http://knoesis.wright.edu/research/ Finally, although early examples of real- tology for SWS, which I believe will build webservices), IBM (see www.alphaworks. world SAWSDL-based services already exist on work being done in the WSMO and ibm.com/tech/wssem), and others. But these (for example, see http://glycomics.ccrc.uga. OWL-S groups. On a more strategic and tools must be robust and in a form that devel- edu/stargate/web_services.jsp), we will need longer-term path, we hope to continue to opers find attractive to learn and use. An end- many more before potential adopters feel that push toward understanding and demon- to-end open source toolset developed by an SWS technology has reached critical mass. strating semantics’ value in making pro- international collaboration of researchers As a follow-up to SAWSDL and with en- cesses more agile, adaptive, and dynamic. could facilitate this process. couragement from members of the SWS Rather than look for a clear winner among Perhaps the most important step is to community such as Dieter Fensel, Charles various SWS approaches, I believe that in the clearly answer questions any manager or Petrie and I have started a SWS testbed incu- post-SAWSDL context, significant contribu- business decision maker will ask: What is bator (www.w3.org/2005/Incubator/swsc). It tions by each of the major approaches will the ROI, what is the cost of adapting SWS will likely continue building additional expe- likely influence how we incrementally en- technology, and what are the concrete ben- rience in developing SWS-based solutions hance SAWSDL. Incrementally adding features efits? In my view, we must focus on two with the help of the SWS Challenge. Kno.e.sis (and hence complexity) when it makes sense, benefits we’ve all talked about: service Center, in some cases in collaboration with by borrowing from approaches offered by reuse and mediation/interoperability/ other researchers, is looking at three issues various researchers, will raise the chance that integration. In the late ’90s, our work led discussed in SAWSDL calls that we hope will SAWSDL can present itself as the primary op- to the development of a commercial work- be components of future solutions: tion for using semantics for real-world and flow-management product, METEOR EAppS industry-strength challenges involving Web (Enterprise Application Suite of Tools and • Development of SA-REST4 for semantic services. Services), that was licensed to numerous or- annotation of REST (Representational ganizations. The company I founded, Info- State Transfer) and other lightweight ser- References cosm, also used it to develop some real- vices. Such non-SOAP/WSDL services world applications. In the case of existing are widely used, and I see a clear value 1. K. Verma and A. Sheth, “Semantically Anno- tasks or activities, I learned that 90 percent in providing a semantics-based solution tating a Web Service,” IEEE Internet Com- puting, vol. 11, no. 2, 2007, pp. 83–85. of the total time spent developing workflows to create powerful smashups (semantic 5 is related to data-mediation issues (trans- mashups ) where it’s easier to integrate 2. A. Sheth and S. Stephens, “Semantic Web: forming the output of one task into the data and services on the client side. Technologies and Applications for the Real

NOVEMBER/DECEMBER 2007 www.computer.org/intelligent 9 World,” Proc. 16th World Wide Web Conf. as a glass-box model, providing users with a figure out how to model this business (WWW 07), 2007; http://www2007.org/ description of what the service does while data within a Semantic Web framework. tutorial-T11.php. hiding the gory implementation details. 3. M. Nagarajan et al., “Semantic Interoper- OWL-S defines a vocabulary for describing Process ability of Web Services–Challenges and the actions that a service could perform in It remains unclear whether the OWL-S Experiences,” Proc. 4th IEEE Int’l Conf. Web terms of inputs, outputs, (pre)conditions, process model describes only the necessary Services, IEEE CS Press, 2006, pp. 373–382. and effects. In this respect, it offers a tanta- (nonexecutable) process constraints on the lizing vision of the future, but it doesn’t go conversation between service and client or 4. J. Lathem, K. Gomadam, and A. Sheth, “SA- REST and Smashups—Adding Semantics to far enough. I don’t intend to criticize OWL- if it provides sufficient information to des- RESTful Services,” to be published in Proc. 1st S; it points the way, but much work remains cribe an executable workflow (or orchestra- IEEE Int’l Conf. Semantic Computing, 2007. before we have a compelling service ontol- tion). Efforts on the SWSF (www.w3.org/ ogy for business use. Here, I look at the Submission/SWSF-SWSO) sought to un- 5. A. Sheth, K. Verma, and K. Gomadam, “Seman- tics to Energize the Full Services Spectrum,” OWL-S service grounding, process, and derpin the process model with a deeper Comm. ACM, vol. 49, no. 7, 2006, pp. 55–61. profile for inspiration and see how research ontological analysis based on the Process on related ontologies has grown out of this Specification Language. This approach has original vision. an advantage over pure process calculi such Business Semantics in Service- as the pi-calculus (which provides seman- Oriented Architectures Grounding tics for BPEL) because it can capture rich Steve Battle, HP Labs Bristol, UK The OWL-S grounding in WSDL focuses descriptions of conditions and effects. on Web services’ basic plumbing, looking at In practice, service exchanges are typi- SOAs offer a powerful level of abstraction cally simple sequences that are weak on promising loosely coupled, modular design. control structure and only really under- They break away from conventional object- standable in terms of the permissions they oriented programming, which works for To make sense carry (for example, you can now deduct tightly knit developer communities in a well- money from my account) and the obliga- defined problem space but is less well ad- of the business process, tions they commit the organization to (for apted to the Web. Service interfaces cross example, you must deliver the goods). To organizational boundaries. This isn’t merely we need sophisticated models make sense of the business process, we a technical issue. Service interfaces must therefore need sophisticated models that cater to the diverse business needs and term- that capture both message capture both message exchanges and their inologies of the different parties involved. associated real-world activities. They also cross functional boundaries; no exchanges and their associated longer just the province of the IT function, Profile they must be described in a way that the real-world activities. The OWL-S service profile is surely the business analyst can understand. At the heart place where we finally get to describe the of the Information Technology Infrastruc- nature of the service being offered, as dis- ture Library/Service Management (ITIL/ tinct from its process or plumbing. However, ITSM)1 vision is the assumption that IT individual inputs and outputs. The service there’s little ontological guidance here, other powers the business. No longer just an grounding has the difficult task of overcom- than a handful of nonfunctional properties infrastructure provider, the IT function is ing the semantic mismatch between OWL- and a summary of the inputs, outputs, pre- key to business innovation. We must try S and WSDL. The SAWSDL working group conditions, and effects as seen in the process thinking about services in new ways has adopted a more incremental approach, model. before they become hopelessly mired in expanding WS* standards by extending the Experience with UDDI has proven that the baggage of traditional programming existing toolset. SAWSDL builds on an ap- we can’t describe everything a service does techniques. proach touched on in OWL-S, which is to in a simple name (even if it’s a uniform re- embed semantic annotations directly in source identifier) and retain any real seman- Service ontology is alive WSDL. This results in a more direct tics. WSMO demonstrates the use of rich and well grounding that gets straight to the gory service descriptions, enabling the matching The Semantic Web offers a new modeling details of lifting and lowering messages. of service offers and requests in service dis- paradigm that seems to fit the bill. It embod- A complementary approach could be covery. These representations of service ies the open nature of the Web in its seman- to work directly with the WSDL-to-RDF capabilities must be expressed in domain- tics and provides an approach for capturing mapping.2 Although this grounding is im- specific terms. OWL-S is an upper ontology business vocabulary and ontology. However, portant, it provides only basic hooks into and was always intended to be extended by we haven’t yet succeeded in making this a the message content, where the real bus- users into their specific domains. Business business reality. The OWL-S ontology for iness semantics lies. The challenges are users are only now becoming savvy to the services (originally DAML-S) was one of to understand the complex document- power of extensibility that ontologies offer. the first attempts to describe a service inter- exchange style interactions typical in Part of the difficulty in describing business face’s semantics. It captures what is known business-to-business interactions and to data, processes, and services is the sheer

10 www.computer.org/intelligent IEEE INTELLIGENT SYSTEMS variety of business models and processes into a single message. RDF and OWL pro- resent a healthy pull on the Semantic Web out there. No single ontology could capture vide a syntax-neutral way to model and from the real world. them all. OWL-S and WSMO are unlike manage these business objects within a busi- existing programming languages that pro- ness object repository. This approach offers References vide all the vocabulary up front. a future migration path for service providers Where are these emerging business on- that want to go native with RDF/XML, avoid- 1. The HP IT Service Management (ITSM) Ref- erence Model, Hewlett-Packard, 2003; ftp://ftp. tologies? One interesting data point is the ing the intermediate mapping into XML while hp.com/pub/services/itsm/info/itsm_rmwp.pdf. MIT Process Handbook and its associated retaining backward compatibility with the OWL mapping.3 Established in 1991, this XML world. 2. J. Kopeck and B. Parsia, eds., Web Services online repository includes over 5,000 bus- Service management software provides Description Language (WSDL) Version 2.0: RDF Mapping, World Wide Web Consortium iness activities. In media industries, the a comprehensive range of capabilities for (W3C) note, June 2007; www.w3.org/TR/ Functional Requirements for Bibliographic managing the entire service life cycle. Even wsdl20-rdf. Records4 ontology provides a useful base for before service implementation, creating de- extensions. The distinctions it makes between tailed, formal specifications could support 3. T.W. Malone et al., “Tools for Inventing Orga- works, expressions, and their manifestations powerful tools for assessing whether the new nizations: Toward a Handbook of Organiza- tional Processes,” Management Science, vol. coincide with different kinds of rights that service will actually meet real business re- 45, no. 3, 1999, pp. 425–443; www.ifi.unizh. must be tracked (copyright, performing, and quirements. Once processes go live, they can ch/ddis/fileadmin/user_upload/mklein/ mechanical rights). In financial industries, the be monitored and health-checked against malone99tools.pdf. Financial Product Markup Language (FPML; their idealized specifications. The ongoing 4. “Functional Requirements for Bibliographic see www.fpml.org) has emerged as a lingua task of managing (and eventually decom- Records,” final report, IFLA Study Group on franca for e-commerce in the currency and the Functional Requirements for Biblio- interest-rate-derivatives markets. These seem graphic Records, K.G. Saur München, 1998; a world apart from OWL-S, but really, they www.ifla.org/VII/s13/frbr/frbr.pdf. highlight the wide span of business semantics Upper service ontologies can 5. E. Crewe, “Extending the Open Journals Sys- in real-world applications, ranging from tem OAI Repository with RDF Aggregation industry-specific vocabularies to generic address only a small part of the and Querying,” Proc. 9th DELOS Network of high-level ontologies. Excellence Thematic Workshop: Digital problem; the bulk of the work Repositories, 2005; http://delos-wp5.ukoln. Crossing the RDF/XML chasm ac.uk/dissemination/pdfs/crewe.pdf. XML is the de facto standard for ex- lies in developing industrial- 6. D. Booth, “RDF and SOA,” position paper, changing business messages in SOAs. Mov- W3C Workshop: Web of Services for Enter- ing to RDF message exchange is beneficial scale ontologies that capture prise Computing, 2007; www.w3.org/2007/ where there are strong requirements to com- 01/wos-papers/booth. municate structured content unanticipated at design time. An example is the Open Ar- real business semantics. chives Initiative protocol for metadata har- Untethering vesting, where embedding RDF/XML in the Semantic Web Services message content is a natural fit.5 In other missioning) services requires us to maintain Katia Sycara, Carnegie Mellon University Web service applications where the message dynamic models of business services, sup- content is relatively locked down, XML re- porting assets, and their virtualized infra- Researchers and practitioners are in- mains the ideal choice, being both compact structures. The Configuration Management creasingly realizing that semantics would and easy to validate against an XML schema. Database lies at the heart of this ITIL con- be a beneficial addition to Web services. Presenting XML and RDF as an either/or figuration-management process. The ability Since the release of DAML-S (precursor of choice is therefore a mistake; Web service to map these models into a Semantic Web OWL-S) version 0.5 in 2001, Semantic Web developers will rightly go with the tools framework enables ease of integration, so- services have seen a lot of activity. Recent they’re familiar with. However, RDF can be phisticated inference, and complex query notable events include the W3C-proposed viewed as an augmentation of XML pro- that would provide invaluable business in- recommendation status of SAWSDL and the viding a new level of abstraction;6 individ- telligence to managers who rely on high- recent SWS testbed incubator activity. So, ual lexical items are cast into their value level views of their service operations. this is a good time to pause and think, space, and the document structure is mapped The take-home message is that upper “What next?” What’s the way forward to to an object-relational model conforming service ontologies can address only a small greater adoption of SWS in industry and to the XML schema. part of the problem, laying out the overall diffusion of SWS technologies into the Viewing RDF as an extension of XML modeling framework. The bulk of the work research and business communities? enables us to look at new ways in which the still lies in the development of good, indus- Semantic Web can add value to existing trial-scale ontologies that capture real busi- Next steps SOAs. The ontological approach makes us ness semantics. These ontological exten- One approach is to follow the current look at schema with new eyes, viewing them sions come from industrial stakeholders, as incremental trajectory. SWS ontologies as definitions of business objects assembled seen in vocabularies such as FPML, and rep- (such as OWL-S) were developed with the

NOVEMBER/DECEMBER 2007 www.computer.org/intelligent 11 realization that SWS was the child of the standardized measurement procedures to • operational issues, such as improving Semantic Web and Web services. This real- ascertain the quality of the service pro- quality and productivity; ization led to the explicit strategy of devel- vided.1 Finally, some organizational cul- • managerial issues, such as improved de- oping SWS technologies not in isolation, tures might be averse to Web services. If cision making, planning, and resource but on top of Web services and the Seman- this indeed is the case, I am not sure that management; tic Web. This strategic goal translated into introducing semantics to Web services • strategic issues, such as support for busi- design decisions to build on technologies at would go far enough to rapidly overcome ness alliances, business growth, innova- the core of the Semantic Web, such as RDF the barriers. This would indicate that the tion, and product differentiation; and and OWL, and on Web Services standards, value and power of Semantic Web services • organizational issues, such as support such as WSDL, SOAP, UDDI, and BPEL4WS lie elsewhere. for organizational change, building a (BPEL for Web Services). This philosophy common vision, and facilitation of orga- is also evident in SAWSDL, which prescribes Two big opportunities nizational learning. incrementally adding semantics to WSDL. Incrementally adding semantics to Web ser- Following this incremental view binds the vices standards might be in line with business To better understand this impact, we must SWS adoption to the adoption of Web ser- caution vis-à-vis semantics (this caution pos- understand how business goals translate into vices. Web services aim to enable interoper- sibly has its roots in the AI experiences of the computational business processes that enact able, dynamic business-to-business interac- ’80s). However, focusing solely on inserting those goals. This reflects a top-down view. tion; they must, therefore, also strongly semantics incrementally might mean missing IT is under pressure to justify the value it support enterprise application integration. out on bigger opportunities to show SWS’ adds to the enterprise, and its continued ef- Because organizations’ IT infrastruc- value. I see at least two such opportunities: forts to do so have echoed this realization tures consist of numerous disparate legacy in a bottom-up way. This has resulted in a systems, enterprises want to bridge these clamor for automated integration of higher- systems and enable them to interoperate level business functionality to the IT ser- flexibly. From a technical perspective, an A need exists to model vices that implement high-level business important advantage of Web services over goals and relations. Achieving this integra- previous distributed computing technolo- informal business requirements tion will require technologies and stan- gies is that they employ open, platform- dards for modeling and re-engineering the independent standards. Adopting open in ways that make it feasible enterprise—not just integrating business standards promises not only to facilitate applications.1 A need exists, then, to model interoperability among disparate applica- to translate them into informal business requirements in ways tion platforms, but also to lower the barriers that make it feasible to translate them into to application integration. The complexity precise business-service precise business-service specifications, in- of integration, the integration challenge,is cluding operational interfaces and rules for increasing much faster than the ability to specifications. procedures, timing, integrity, and quality. integrate. Service-oriented architectures Such modeling must be driven from the and Web service technologies have been top down, directly from business require- heralded as a solution. ments. The models express intended busi- As a matter of fact, most attention and • supporting higher-level enterprise mod- ness semantics in a vertically integrated studies on whether Web services will suc- eling to enable integration of business way. The modeling would provide a func- cessfully deliver on their promise have fo- goals with services that computationally tionality that’s entirely understandable from cused on the integration challenge and IT enact those goals, and a business perspective; it would depend on improvement issues (such as cost reduction • supporting increased autonomous and business context, goals, and operational and increased infrastructure capability and dynamic Web services behavior. standards, but shouldn’t depend on the tech- flexibility). Undoubtedly, introducing se- nology used to implement them. The mod- mantics into SOA-based implementations Besides being big opportunities, these two els would provide business value directly can aid in the integration challenge by areas are attractive because, for various relating to business purposes and could be helping clients understand the intended technological and cultural reasons, they’re understood and used without knowledge of semantics of services and the meaning of not tightly coupled (technically) or com- underlying IT artifacts. exchanged messages’ content. It can also mitted (culturally) to using the current Web Current Web services proposals don’t add flexibility in discovery of services that services standards infrastructure. enable the semantic representation of busi- can be reused across a workflow. ness relations, contracts, or business rules However, despite Web services’ promise, Business process semantics in a machine-understandable way. Current it’s unclear that adoption is widespread. The business community has been in- business-process languages aim at repre- Security is a major concern.1 Web services creasingly interested in the possible impact senting and describing business processes allow direct access to a company’s applica- of Web services on ROI and other business and achieving a joint understanding of a tions, exposing corporate networks to secu- issues, while decrying the dearth of studies process involving several parties. However, rity threats. Web services standards’ imma- on these issues.3,4 Such big-picture issues they’re at low abstraction levels and don’t turity is another concern,2 as is the lack of include provide formal business semantics. For ex-

12 www.computer.org/intelligent IEEE INTELLIGENT SYSTEMS ample, the Business Process Modeling No- portunity areas, and work by the SWS re- References tation5 is a graphical notation with XML as search community is under way.6–8 The the underlying representation, lacking for- skeptic asks, “But what about the barriers 1. D. McGoveran, Embracing SOA: The Bene- fits of Integration Independence, report mal semantics. BPEL is well suited for des- to adoption that you mentioned?” A num- 20060125, Alternative Technologies and cribing communication behavior from a ber of answers come to mind. Regarding TIBCO Software, 2006; www.tibco.com/ particular individual service’s point of view the barriers to current adoption of Web ser- resources/solutions/soa/embracing_soa_wp.p and can map to the code to execute the pro- vices, decoupling SWS from the Web ser- 2. H. Deitel et al., Web Services: A Technical cess representation. The technical literature vices stack might mitigate some challenges Introduction, Pearson Education, 2003. spends much time discussing BPEL’s short- and make the new technology nimbler. Addi- comings (for example, it can’t represent tionally, neither area of opportunity involves 3. R. Hailstone and R. Perry, “IBM and the choreographies) and the Web Services an accumulation of legacy technologies, so Strategic Potential of Web Services: Assess- Choreography Description Language’s in- embracing the SWS promise wouldn’t entail ing the Customer Experience,” ICD white paper, IBM, 2002. ability to map to BPEL or to support sce- abandoning past investment (unlike the case narios where the number of participants is of the application integration challenge). 4. A. Manes, Web Services: A Manager’s Guide, known only at runtime. For all the nomen- Moreover, the higher level of conceptualiza- Addison Wesley, 2003. clature similarities between service and tion called forth would increase understand- 5. Business Process Modeling Notation Speci- business processes, current process lang- ing of security issues and how security and fication, tech. report, Object Modeling Group, uages don’t concern themselves with try- other service-quality concerns might tie in 2006; www.bpmn.org/Documents/OMG% ing to link workflows and message choreo- with business goals and commitments. 20Final%20Adopted%20BPMN%201- graphies to high-level business notions, such Finally, if the SWS approaches can provide 0%20Spec%2006-02-01.pdf. as organizational goals, commitments, and 6. N. Desai et al., “OWL-P: A Methodology for contracts. Business Process Modeling and Enactment,” Proc. 7th Int’l Bi-Conf. Workshop Agent- Web service autonomy and dynamism The notion of 100 percent Oriented Information Systems III (AOIS 05), The cell phone’s emergence as a primary LNCS 3529, Springer, 2006, pp 79–94. access point for Web content and services completeness and correctness 7. M. Paolucci and K. Sycara, “Autonomous has fueled the need for Web services to Semantic Web Services,” IEEE Internet Com- operate more autonomously and dynami- doesn’t make sense anymore; puting, vol. 7, no. 5, 2003, pp. 34–41. cally. Ultimately, the growing Web services infrastructure (the Web services stack) 8. R. Masuoka, B. Parsia, and Y. Labrou, “Task the underlying fact base is Computing—The Semantic Web Meets Per- facilitates the emergence of agreements vasive Computing,” Proc. 2nd Int’l Semantic between programmers and the coding of changing faster than any Web Conf., LNCS 2870, Springer, 2003, pp. those agreements. The result, however, is 866–881. an inherently brittle infrastructure that’s difficult to reconfigure to accommodate reasoning process can process it. new Web services or react to failures and Computing for the World: that’s inevitably expensive to maintain. For Incomplete, Incorrect, example, service registries don’t directly breakthroughs that add value in orders of but Requested! support sufficiently flexible, dynamic, magnitude, as I anticipate, this alone will Dieter Fensel, University of Innsbruck automated service discovery; nor do they provide compelling incentives to deal with and STI International support service context, such as location- any apparent barriers. dependent or context-sensitive information. The SWS community has the oppor- Dave Brown from Worcester Polytech- The way to overcome the infrastructure’s tunity to lead the development of standards nic Institute recently asked whether the brittleness is to make Web services more and specifications for concepts such as work on problem-solving methods of the autonomous by letting them reconfigure contracts, business-level agreements, and late ’90s is dead.1 I answered, “Yes, but only their interaction patterns. Any increase in commitments and how they relate to the in a certain sense.” Heuristic problem solv- autonomy lets Web services react to changes processes that computationally enact them. ing for closed, fixed applications doesn’t while minimizing direct intervention from Current standards can’t express these pro- deserve much attention anymore. However, programmers. Crucially, the lack of explicit cesses because they require more semanti- heuristic problem solving has begun to re- semantics prevents Web services from acting cally rich representations. Additionally, by appear on a worldwide scale. Two distinct autonomously, which prevents them from supporting dynamism and autonomy, SWS approaches define its essence: dynamically discovering one another or can support service decentralization and understanding what each other’s messages adaptive business processes that support • a goal-driven approach where the solu- mean and what tasks each performs. changing business needs. This new orienta- tion process is only partially determined tion of SWS work, decoupled from the cur- ahead of time and actually decided at A possible way forward rent Web services stack, could lead to a runtime on the basis of available data The presence of semantics is a sine qua more flexible and expressive services infra- and services, and non for making progress in these two op- structure that meets business needs. • a heuristic approach that gives up on the

NOVEMBER/DECEMBER 2007 www.computer.org/intelligent 13 absolute notion of completeness and cor- discovery approaches will focus on utility approaches for heuristic problem solving rectness to gain scalability. (stopping the search when a service is such as KADS,3 CommonKADS,4 and the found that’s good enough to fulfill a re- Unified Problem-Solving Method Develop- Both aspects seem to be essential for quest). Also, it’s unrealistic to assume that ment Language (UPML)5 to conceptual mod- working in an open, heterogeneous, distrib- semantic descriptions of services are cor- els for open, distributed, and heterogeneous uted, and fast-changing environment such as rect and complete—that is, that they dupli- problem solving. WSMO is augmented by a the World Wide Web. Therefore, areas of cate a service’s functionality at the descrip- family of formalization languages called the computer science that initially aimed toward tion level. Web Service Modeling Languages: 100 percent complete and correct solutions In all areas, you must define the trade-off start to incorporate these principles as soon between the guarantees you provide in • WSML-Core captures RDFS. as they want to be usefully applied in a Web terms of service-level agreements (com- • WSML-DL (Description Logics) cap- context. pleteness and correctness are just examples tures OWL. In a large, distributed, and heterogeneous of very strong guarantees) and what this • WSML-Rule captures the Rule environment, the classical ACID (atomic, requires in terms of assumptions and com- Interchange Format (RIF). consistent, isolated, and durable) guarantees putational complexity. Various relationships • WSML-Full provides a full-fledged, lay- of the database world no longer scale in exist between these factors, and they’re ered logical framework. any sense. It’s completely out of scope to worth further analysis and more detailed assume that the Web could remain static study. Different heuristic problem-solving As a third cornerstone, WSMX (Web Ser- until some crawler has searched it in its approaches are just different combinations vice Execution Environment) and SEE (Se- entirety to produce a defined query result. of three factors: mantic Execution Environment) provide Therefore, a simple read operation in an reference architectures for scalable prob- environment such as the Web, a peer-to-peer • Service-level agreements (or goals) de- lem solving on the Web. storage network, a set of distributed reposi- fine what type of result the problem Recent work derives simple description tories, or a space can’t guarantee complete- solving must provide. Do we request an means for services from this framework. ness in the sense of assuming that if data optimal solution, a semi-optimal solu- WSMO-Lite combines wasn’t returned, it didn’t exist. It’s often the tion, or just any solution? case that the data just wasn’t found in the • Assumptions describe the generality of •SAWSDL as means to reference semantic amount of time given to search for it. Simi- the problem-solving approach. Assum- descriptions, larly, a write can’t guarantee that a consis- ing that the problem has only one solu- • a small meta-ontology in RDFS to define tent state replicates immediately to all stor- tion allows stopping the search for an the pragmatic meaning of a semantic age facilities at once. optimum once a solution has been found. description (that is, specifying whether What holds for simple data lookup holds Instead of a global optimization method, it’s a condition or an effect of an opera- in an even stronger sense for reasoning on a much simpler heuristic search method tion), and Web scale.2 The notion of 100 percent com- can be used in this case and would still • RDF syntax. pleteness and correctness as usually assumed deliver a global optimum. in logic-based reasoning doesn’t make sense • Computational complexity (scalability) MicroWSMO will drive this into a simpler, anymore because the underlying fact base is denotes the resources required to fill the useful framework for describing services changing faster than any reasoning process gap between the assumptions and the semantically. It will provide structured tags can process it. So, we must develop a notion goals. (concepts with attributes) that link to onto- of usability of inferred results and relate logies in the background as means for com- them with the resources that are requested Computer science in the 20th century was munities of service users to tag services. for those results. about perfect solutions in closed domains Returning to Brown’s question with which Even in information retrieval, the notion and applications. Computer science in the I started this essay: problem-solving meth- of completeness (recall) becomes more and 21st century will be about approximate solu- ods are alive in the same way a baby is alive more meaningless in the context of Web- tions and frameworks that capture the rela- when it leaves its closed and cozy world. A scale information infrastructures. It’s un- tionships of partial solutions and require- child’s transition to a new, open world can likely that users will request all the infor- ments in terms of computational costs—that be inconvenient and painful. Its a new world mation available about and relevant to a is, the proper balance of their ratio. In a nut- is full of risk, but also large enough to pro- certain topic because this might exceed the shell, heuristic problem solving on the Web vide for future growth and development to- amount of information processing they’re scale will likely soon become an entire re- wards its full potential. investing in achieving a certain goal. So, search area. This shift is comparable to the instead of investigating the full space of transition from classical physics to relativity precision and recall, information retrieval theory and quantum mechanics, where rela- References is starting to focus more on improving pre- tivistic notions and the principle limits of cision and ranking of results. precision replace the notion of absolute 1. D. Fensel, Problem-Solving Methods: Under- In a world of billions of services, the space and time. standing, Development, Description, and cost of finding the “optimal” service might First steps in this direction are efforts such Reuse, LNAI 1791, Springer, 2000. outweigh the benefits. Pragmatic service- as WSMO, which translates earlier modeling 2. D. Fensel and F. van Harmelen, “Unifying

14 www.computer.org/intelligent IEEE INTELLIGENT SYSTEMS Reasoning and Search to Web Scale,” IEEE David Martin is a Steve Battle is a Internet Computing, vol. 11, no. 2, 2007, pp. senior computer sci- researcher at HP 94–96. entist in the Artifi- Labs, Bristol. Con- cial Intelligence tact him at steve. 3. A.T. Schreiber, B.J. Wielinga, and J.A. Center at SRI Inter- [email protected]. Breuker, eds., KADS: A Principled Approach national. Contact to Knowledge-Based System Development, him at martin@ Academic Press, 1993. ai.sri.com.

4. G. Schreiber et al., Knowledge Engineering John Domingue Katia Sycara is a research professor at Car- and Management: The CommonKADS is deputy director of negie Mellon University, where she directs Methodology, IT Press, 1999. the Open Univer- the Intelligent Software Agents and Seman- sity’s Knowledge tic Web Technologies Lab, and holder of the 5. D. Fensel et al., “The Unified Problem-Solv- Media Institute. Sixth Century Chair in Computing Science ing Method Development Language UPML,” Contact him at j.b. at the University of Aberdeen. Contact her at Knowledge and Information Systems, vol. 5, domingue@open. [email protected]. no. 1, 2002, pp. 83–131. ac.uk.

Amit Sheth is the Dieter Fensel is a LexisNexis Ohio computer science Eminent Scholar professor at the Uni- director of Kno.e.sis versity of Innsbruck Center at Wright and president of Se- State University and mantic Technology For more information on this or any other com- an IEEE fellow. Institute Interna- puting topic, please visit our Digital Library at Contact him at tional. Contact him www.computer.org/csdl. amitpsheth@ at dieter.fensel@ ieee.org; http://knoesis.wright.edu/amit. deri.at.

New from The MIT Press The Robotics Primer Autonomous Toward Brain-Computer Maja J Matarić Bidding Agents Interfacing “As with electronics and then computers, Strategies and Lessons from the Trading edited by Guido Dornhege, the world is about to be transformed yet Agent Competition Jose del R. Millan, Thilo Hinterberger, again by the age of robotics. Every age needs its technology to be adopted by Michael P. Wellman, Amy Greenwald, Dennis J. McFarland, smart kids and dedicated hobbyists to and Peter Stone and Klaus-Robert Müller provide the tide of people who will carry This book presents algorithmic advances The latest research in the development out the transformation. The Robotics and strategy ideas developed within an of technologies that will allow humans to Primer is a bridge between academia and integrated bidding agent architecture that communicate, using brain signals only, everyperson, and its readers will become have emerged from recent research in this with computers, wheelchairs, prostheses, the crucibles of the new age.” fast-growing domain of AI. and other devices. — Rodney Brooks, Director, MIT Computer Intelligent Robotics and Autonomous Agents series Neural Information Processing series Science and Artificial Intelligence Lab 256 pp., 25 illus., $35 cloth 519 pp., 150 illus., $45 cloth Intelligent Robotics and Autonomous Agents series 288 pp., 88 illus., $30 paper Predicting Large-Scale Kernel Introduction to Statistical Structured Data Machines edited by Gökhan H. Bakir, edited by Léon Bottou, Olivier Chapelle, Relational Learning Thomas Hofmann, Bernhard Schölkopf, Dennis DeCoste and Jason Weston edited by Lise Getoor and Ben Taskar Alexander J. Smola, Ben Taskar Topics covered include fast implementa- Advanced statistical modeling and and S. V. N. Vishwanathan tions of known algorithms, approxima- knowledge representation techniques for a State-of-the-art algorithms and theory in a tions that are amenable to theoretical newly emerging area of machine learn- novel domain of machine learning, predic- guarantees, and algorithms that perform ing and probabilistic reasoning; includes tion when the output has structure. well in practice but are diffi cult to analyze introductory material, tutorials for diff erent Neural Information Processing series theoretically. proposed approaches, and applications. 8 x 10, 360 pp., 61 illus., $45 cloth Neural Information Processing series Adaptive Computation and Machine Learning series 8 x 10, 416 pp., 116 illus., $45 cloth 480 pp., 134 illus., $50 cloth The MIT Press To order call 800-405-1619 • http://mitpress.mit.edu

NOVEMBER/DECEMBER 2007 www.computer.org/intelligent 15