"A Review of Semantic Web Services"
A Review of Semantic Web Services David Brokenshire Simon Fraser University 13450 102 Ave. Surrey BC V3T 5X3 firstname.lastname@example.org ABSTRACT The motivation also derives directly from Tim Berners- This paper conducts a review of recent work in the ﬁeld of Lee’s vision of the semantic web proposed in the 2001 Sci- Semantic Web Services. We start by discussing the large entiﬁc American article. There he proposed a variety of problems in SWS, we then review the major approaches. services which would be accessed and composed by agents Next, for each approach we discuss the problem(s) it aims acting on behalf of people or groups. to addresses, main idea of the solution, current status and re- sults, and unresolved problems. Then we present a summary 1.3 Technologies of the status of the ﬁeld. Finally we discuss the remaining problems as a potential for further research. 1.3.1 Web Services Semantic Web Services rely on technologies from both ﬁelds, we brieﬂy describe some of the major ones. From web Keywords services we have Universal Description, Discovery and Inte- Ontologies, OWL, DAML+OIL, Semantics, Semantic Web, gration (UDDI) which is a registry standard for discovering Web Services web services. UDDI describes businesses based on physi- cal descriptions and can supply extra info through TMod- els which can classify services, but it doesn’t provide info 1. INTRODUCTION on what a service does. The Web Services Deﬁnition Lan- guage (WSDL) is an XML-based language for describing 1.1 Basics web services. Simple Object Access Protocol (SOAP) is a Semantic Web Services (SWS) are a combination tradi- XML protocol web service communication and invocation. tional Web Services and Semantic Web technologies. Speciﬁ- The Business Process Execution Language for Web Services cally, SWS are Web Services which have been marked/annotated (BPEL4WS) is another XML language. It is used to for- with machine-interpretable semantic markup, in the form mally specify business processes and interactions for web of ontologies. One deﬁnition is ”the augmentation of Web services. A good review of issues in web services, partic- Service descriptions through Semantic Web annotations, to ularly the wide variety of standardization eﬀorts and lan- facilitate the higher automation of service discovery, compo- guages can be found in . sition, invocation, and monitoring in an open, unregulated, and often chaotic environment (that is, the Web).” . 1.3.2 Semantic Web The complaint against all of the web service XML tech- 1.2 Motivation nologies is that they are purely syntactical (see for example, Semantic Web Services are motivated in part by the pop-  or ). The Semantic Web provides the ability to add ularity of web services and in part by their diﬃculties. Web semantics to web services. First we have OWL and its prede- services have been widely adopted by industry as a way of cessor DAML+OIL, both of which are XML languages for organizing systems online. The challenge of composing com- representing ontologies in the web. The DAML people at plex services from simple services has driven the creation of DARPA then created the DAML-S which became OWL-S a variety of process languages. Composing web services has an upper ontology for describing semantic web services, de- typically involved a programmer who links up the various ﬁned in OWL. Other languages and frameworks have been services. Creating workﬂows of services is considered chal- proposed but OWL-S is the most popular currently. These lenging and requires an expert. Finding web services has technologies are used to build upon the existing web services also been limited despite the existence of registries because technologies to create Semantic Web Services. descriptions are in text and can only be searched by key- word. Semantic description of web services is an attractive prospect because it has the potential to allow the use of AI 2. LARGE PROBLEM OVERVIEW technologies to semi-automatically or automatically search Semantic Web Service research is not a single area of en- for, compose, and invoke web services, with far less eﬀort. deavor. It draws from many aspects of computer science and    artiﬁcial intelligence research as well as other disciplines such as workﬂow analysis. There are three main aspects to re- Copyright is held by the author/owner(s). WWW2006, May 22–26, 2006, Edinburgh, UK. search in semantic web services. In chronological order they . are service discovery, composition, and invocation. These are supported by three connected areas, interoperation, ver- 3.2 Service Discovery iﬁcation, and monitoring. Finally two issues arise as a result Service discovery is essential in Semantic Web Services, of SWS, privacy and security. In the following section we before we can use any services, we must ﬁnd them. In tradi- give a deﬁnition of each area and address the major tech- tional web services, registries like UDDI are used to record niques used to approach it. simple descriptions of services in a publicly accessible reg- istry. These descriptions are then searched with simple key- word searching. Humans read through the results and then 3. REVIEW OF APPROACHES investigate the service further to determine the suitability of As there are many diﬀerent goals/aspects to semantic web the service for a particular task. With SWS we have seman- services research, there are many diﬀering approaches. We tic annotations of web services using ontologies. This richer will cover the main approaches to each area now. description allows for more speciﬁc and powerful searches, and results can be analyzed exactly and automatically to 3.1 Annotation determine if they meet semantically speciﬁed requirements. Annotation of web services is the fundamental concept Paolucci et al.  argue for semantic web service discov- underlying Semantic Web Services. The idea, as mentioned ery by matching descriptions using DAML-S. As they state previously, is to combine the ubiquity and engineering con- it service discovery is ”location of web services on the basis venience of web services with the representational power of of the capabilities that they provide.” There view of the so- the semantic web. This is to be accomplished by using on- lution is: ”The solution of this problem requires a language tological descriptions of what services do instead of just syn- to express the capabilities of services, and the speciﬁcation tactical descriptions of how  . of a matching algorithm between service advertisements and The are a variety of mechanisms which have been pro- service requests that recognizes when a request matches an posed for this annotation. OWL-S, (formerly known as advertisement.” DAML-S), is an upper level ontology for describing seman- Paolucci also deﬁnes the major diﬃculty with naive match- tic web services using OWL. It was proposed by the DAML ing: ”The major problem with capability matching is that group   and is now a member submission to the W3C it is unrealistic to expect advertisements and requests to be . The OWL-S ontology consists the Service class, with equivalent, or even that exists a service that fulﬁlls exactly three main parts, ServiceProﬁle, ServiceGrounding, and Ser- the needs of the requester.” Assuming registries of services viceModel. they then describe a matching algorithm for linking service DAML-S and OWL-S have been the most popular repre- advertisements with service requests. They deﬁne matching sentations for web service ontologies but they are not with- as ”An advertisement matches a request, when the adver- out ﬂaws, competitors, and possible improvements. tisement, describes a service that is suﬃciently similar to In  Mika et al. propose a foundational ontology called the service requested.” DOLCE to clarify the semantics of OWL-S, increase the Benatallah et al.  propose an algorithm for matching precision of its axiomatization, and broaden its deﬁnition of requests with web services. ”The proposed approach enables services to better represent real world services. Real world to select the combinations of Web services that best match services are those which have an eﬀect on the world outside a given request Q and eﬀectively computes the extra infor- the system, not just a web service process. mation with respect to Q (e.g., the information required by Motta et al  propose IRS-II (Internet Reasoning Ser- a service request but not provided by any existing service).” vice), ”...a framework and implemented infrastructure, whose They claims SWS still in infancy (2003) and most areas main goal is to support the publication, location, compo- have not yet been achieved. Also call service discovery ’dy- sition and execution of heterogeneous web services, aug- namic discovery’. Their proposed algorithm uses a rewriting mented with semantic descriptions of their functionalities.” approach to the query into the closest description expressed IRS-II was based on the Uniﬁed Problem-solving Method as a conjunction of web services which exist. ”The matching Development Language (UPML) but is now moving to an algorithm takes as input a service request (or query) Q and OWL based implementation. Similarly to OWL-S it oﬀers a DAML-S ontology T , and computes the best combination several ontologies describing key components of semantic of Web services that satisﬁes as much as possible the out- web services, and it also includes an implementation of all puts of the query Q and that requires as little as possible of the proposed components. IRS-II diﬀers from OWL-S in inputs that are not provided in Q.” that it explicitly represents tasks separately from services, To deal with the ﬂexibility requirement posed by  they and oﬀers registries for discovering tasks as well. use a diﬀerence operation which computes the distance be- Fensel et al  propose the Web Service Modeling Frame- tween descriptions and allows them to use their algorithm. work (WSMF) for describing web services and their com- Approach is extension of best covering problem from hyper- position. It follows a similar approach to OWL-S, but it graph theory. has since been incorporated in the Web Services Distributed These approaches are a signiﬁcant improvement on the Management (WSDM) standard. keyword based search typical of existing web service sys- The common practical thread between these various ap- tems. Two measures of search quality are precision and proaches is that they all attempt to present a useful frame- recall. Precision measures the ability of the system to avoid work for solving the major problems of SWS: discovery, false positives, only returning correct results. Recall mea- composition, and invocation. Theoretically they are tied sures the ability of the system to return as many correct together by their common use of ontologies and their fun- results as possible, avoiding false negatives. Bernstein and damental use of logic and knowledge representation tech- Klein  compared keyword and table based web service dis- niques, and the necessary trade oﬀ between expressiveness covery techniques against semantic methods and found that and complexity. keyword and table-based techniques had roughly compara- ble recall, but had much worse precision. out human intervention to clarify diﬀerences in representa- tion versus diﬀerences in meaning. 3.3 Composition The OWL-S proposal  describes composite services 3.5 Invocation thus: ”complex or ’composite’ services are composed of mul- Invocation involves actually calling the individual web ser- tiple more primitive services, and may require an extended vices and making use of the results. Web services focus most interaction or conversation between the requester and the directly on invocation. Existing XML description languages set of services that are being utilized.” So the task of com- and registries are almost exclusively focused on representing position involves selecting appropriate ’primitive’ services the necessary inputs, outputs, and preconditions of services which can work together when invoked to accomplish some as well as the necessary location at which to ﬁnd and invoke goal or task. The need for composing services comes from them. web services. In web services to date this is done manually by a pro- 3.6 Privacy and Security grammer or other expert. The BPEL4WS is one language Privacy and security work has mostly taken the form of which attempts to enable composition of web services into policy languages with which to annotate services. Security executable processes. Unfortunately BPEL4WS is limited information includes encryption and digital signature infor- by its lack of formal semantics . mation about how inputs/outputs will be passed and stored. In Semantic Web Services composition is performed auto- The also include information about to where information matically by the system based only on declarative descrip- may be sent and for what purpose. Denker et al.  pro- tions of the task and the services. poses security ontologies to annotate web services with secu- From our deﬁnition we can see that there are several nec- rity information and describes a prototype. Kagal et al.  essary steps to accomplishing this goal. First we must have a describes a semantic policy language and distributed policy representation for our declarative descriptions, with accom- management for the semantic web (agents, resources, ser- panying semantics and axioms. Variations on this ability vices). Tonti et al.  compares three languages for policy are provided by all of the ontologies described in the Anno- representation: KAoS, Rei and Ponder. tation section, including OWL-S,IRS-II, and WSDF. There Kagal et al.  describes policies for authorization and are also extensions proposed for OWL-S. Solanki et al.  privacy for SWS which ”Aims to provide security and pol- propose adding two ’compositional’ assertions to web service icy annotations for OWL-S service descriptions”. Web ser- annotation. More detail on Solanki here. vices will handle private information and will need to provide Once we have a representation and can discover services privacy guarantees. They will need to do reasoning about (see section 3.2) we need to reason about which services can privacy and information sharing and because SWS are au- accomplish the task, and which are the ’best’ choices. tomated the guarantees need to be that much better. Current approaches for web service composition have drawn Previous work proposes high-level security requirements. largely from the AI planning community. Chen et al.  pro- They try to match a client’s request with appropriate ser- pose a knowledge based service composition technique/framework vices based on security considerations as well as functional- and prototype. Kuter et al.  describe a Hierarchical Task ity. Kagal’s work is more ﬁne grained. They express anno- Network (HTN) planning technique for web service compo- tations in Rei - ”a logic-based language that lets you deﬁne sition with the ENQUIRER algorithm. Wu et al.  also rules and constraints over domain-speciﬁc ontologies” Kagal use HTN planning for composition. Another use of plan- claims OWL and DAML+OIL are insuﬃcient to specifying ning is describe by Traverso et al in . They do auto- policy constraints. matic web service composition using planning (model check- ing).Spencer et al.  work on integrating SWS using data transformation and a Description Logic tableau-algorithm 4. CONCLUSION reasoner. 4.1 Summary of Status of Field 3.4 Interoperability Semantic Web Services as a ﬁeld is only just beginning to Interoperability is necessary for composite services since develop. It has identiﬁed the main areas of research which output data from one service may become the input to an- deﬁne it, and techniques have been proposed and imple- other service. In traditional web services interoperability mented for all of the areas, however they approaches are was ensured by having a programmer manually compose ser- not yet to a large scale production level. Sure et al.  vices and write any necessary adapting software. Since SWS ran a one day experiment to see what experts could create proposes automatic composition, this approach is no longer from semantic web technologies, including web services, in possible. Instead we have approaches stemming from the just 24 hours. The results were impressive, demonstrating ability of ontologies to specify the descriptions in enough the possibilities of these technologies. Given the amount of detail that a machine can connect disparate services. money, hype, and eﬀort invested in Semantic Web Services Mandell et al  propose extending BPEL4WS with se- I expect it will actively develop and we will see production mantics for interoperation. HeB et al  propose three ma- systems in the near future. chine learning techniques for annotating web services for in- teroperability. Burstein et al  deal with the important 4.2 Remaining Problems issue of interoperability across unfamiliar ontologies. This As I’ve described there are several open problems in Se- is essential because ontologies on the web are rarely com- mantic Web Services. I n the fundamental area of annota- pletely agreed upon standards, and we are assuming that tion there is not yet any consensus as to what representation services are to be composed and work automatically, with- to use. OWL-S has seen the most use but problems have been identiﬁed with its representative power, and particu-  Martin, D., Burstein, M., Hobbs, J., Lassila, O., larly the granularity at which it represents services. Argu- McDermott, D., McIlraith, S., Narayanan, S., ments have also been made that OWL-DL lacks the expres- Paolucci, M., Parsia, B., Payne, T., Sirin, E., sive power to record all the necessary information. Other Srinivasan, N., and Sycara, K. Owl-s: Semantic additions continue for the diﬀerent areas in SWS. markup for web services. W3C Member Submission, Each of the areas also has its own techniques and prob- November 2002. lems. Composition has mostly used techniques from AI  McIlraith, S., and Martin, D. Bringing semantics planning but these techniques are intractable for large prob- to web services. Intelligent Systems, IEEE [see also lems. Service discovery has several promising algorithms IEEE Intelligent Systems and Their Applications] 18, proposed, but the expressiveness of techniques for annotat- 1 (2003), 90–93. ing services limits the possibilities in searching. Discovery  Mika, P., Oberle, D., Gangemi, A., and Sabou, algorithms proposed have achieved good precision but still M. Foundations for service ontologies: Aligning owl-s need to develop their recall. to dolce. In In Proceedings of the Thirteenth International Conference on The World Wide Web (2004). 5. REFERENCES  Motta, E., Domingue, J., Cabral, L., and  Ankolekar, A., Burstein, M., Hobbs, J. R., Gaspari, M. IRSII: A Framework and Infrastructure Lassila, O., Martin, D., McDermott, D., for Semantic Web Services, vol. 2870. Jan. 2003. McIlraith, S. A., Narayanan, S., Paolucci, M.,  Paolucci, M., Kawamura, T., Payne, T. R., and Payne, T., and Sycara, K. DAML-S: Web Service Sycara, K. Semantic Matching of Web Services Description for the Semantic Web, vol. 2342. Jan. Capabilities, vol. 2342. Jan. 2002. 2002.  Payne, T., and Lassila, O. Guest editors’  Ankolekar, A., Huch, F., and Sycara, K. introduction: Semantic web services. Intelligent Concurrent Execution Semantics of DAML-S with Systems, IEEE [see also IEEE Intelligent Systems and Subtypes, vol. 2342. Jan. 2002. Their Applications] 19, 4 (2004), 14–15.  Benatallah, B., Hacid, M.-S., Rey, C., and  Peer, J. Bringing Together Semantic Web and Web Toumani, F. Request Rewriting-Based Web Service Services, vol. 2342. Jan. 2002. Discovery, vol. 2870. Jan. 2003.  Preece, A., and Decker, S. Intelligent web services.  Bernstein, A., and Klein, M. Towards Intelligent Systems, IEEE [see also IEEE Intelligent High-Precision Service Retrieval, vol. 2342. Jan. 2002. Systems and Their Applications] 17, 1 (2002), 15–17.  Burstein, M. Dynamic invocation of semantic web  Solanki, M., Cau, A., and Zedan, H. Augmenting services that use unfamiliar ontologies. Intelligent semantic web service with compositional speciﬁcation. Systems, IEEE [see also IEEE Intelligent Systems and In Proceedings of The Thirteenth International World Their Applications] 19, 4 (2004), 67–73. Wide Web Conference (2004).  Chen, L., Shadbolt, N. R., Goble, C., Tao, F.,  Spencer, B., and Liu, S. Inferring Data Cox, S. J., Puleston, C., and Smart, P. Towards Transformation Rules to Integrate Semantic Web a Knowledge-Based Approach to Semantic Service Services, vol. 3298. Jan. 2004. Composition, vol. 2870. Jan. 2003.  Staab, S., van der Aalst, W., Benjamins, V.,  Denker, G., Kagal, L., Finin, T., Paolucci, M., Sheth, A., Miller, J., Bussler, C., Maedche, A., and Sycara, K. Security for DAML Web Services: Fensel, D., and Gannon, D. Web services: been Annotation and Matchmaking, vol. 2870. Jan. 2003. there, done that? Intelligent Systems, IEEE [see also  Fensel, D., Bussler, C., and Maedche, A. IEEE Intelligent Systems and Their Applications] 18, Semantic Web Enabled Web Services, vol. 2342. Jan. 1 (2003), 72–85. 2002.  Sure, Y., Hitzler, P., Eberhart, A., and  HeB, A., and Kushmerick, N. Learning to Attach Studer, R. The semantic web in one day. Intelligent Semantic Metadata to Web Services, vol. 2870. Jan. Systems, IEEE [see also IEEE Intelligent Systems and 2003. Their Applications] 20, 3 (2005), 85–87.  Kagal, L., Finin, T., and Joshi, A. A Policy Based  Tonti, G., Bradshaw, J. M., Jeffers, R., Approach to Security for the Semantic Web, vol. 2870. Montanari, R., Suri, N., and Uszok, A. Semantic Jan. 2003. Web Languages for Policy Representation and  Kagal, L., Finin, T., Paolucci, M., Srinivasan, Reasoning: A Comparison of KAoS, Rei, and Ponder, N., Sycara, K., and Denker, G. Authorization and vol. 2870. Jan. 2003. privacy for semantic web services. Intelligent Systems,  Traverso, P., and Pistore, M. Automated IEEE [see also IEEE Intelligent Systems and Their Composition of Semantic Web Services into Applications] 19, 4 (2004), 50–56. Executable Processes, vol. 3298. Jan. 2004.  Kuter, U., Sirin, E., Nau, D., Parsia, B., and  Wu, D., Parsia, B., Sirin, E., Hendler, J., and Hendler, J. Information Gathering During Planning Nau, D. Automating DAML-S Web Services for Web Service Composition, vol. 3298. Jan. 2004. Composition Using SHOP2, vol. 2870. Jan. 2003.  Mandell, D. J., and McIlraith, S. A. Adapting BPEL4WS for the Semantic Web: The Bottom-Up Approach to Web Service Interoperation, vol. 2870. Jan. 2003.