Context-aware Semantic Web Service Composition Yasser Ganji Saffar ganji@ce.sharif.edu Semantic Web Laboratory Computer Engineering Department Sharif University of Technology http://sw.ce.sharif.edu Outline • What is the Problem? – – – – Semantic Web Services Service Discovery Service Composition Context-awareness • Related Works • Contributions – Proposed Architecture for a Context-aware Service Broker – Proposed Methods for Service Matchmaking 2 What is the Problem? Web Services • Web-accessible • Self-describing • Platform independent • Definition of World Wide Web Consortium (W3C): – “a software system designed to support interoperable machine-to-machine interaction over a network. It has an interface described in a machine-processable format (specifically WSDL). Other systems interact with the Web service in a manner prescribed by its description using SOAP messages, typically conveyed using HTTP with an XML serialization in conjunction with other Web related standards” 4 Service Oriented Architecture WSDL Publish SOAP Service Broker Service Provider Only Syntax, Agents can not Find understand meanings SOAP Bind SOAP Service User UDDI SOAP UDDI –– Universal, Simple Object Description, Access Protocol Discovery, / SOA andProtocol Integration WSDL Web Services Description Language 5 Semantic Web Services • Semantic Web – Sharing Information on the Web – Computer-interpretable • Web Services – Sharing Programs on the Web • Semantic Web Services – Web Services + Semantic Web – Using ontologies to describe web services 6 Ontologies for Semantic Web Services • Use ontologies to describe Web Services – DAML-S (since 2001) – OWL-S (since 2003) • Based on OWL-DL – WSMO (since 2004) – SWSO (since 2005) • Based on FLOWS (First-order Logic Ontology for Web Services) 7 Issues in Semantic Web Services • Discovery (Matchmaking): Locate different services suitable for a given task • Selection: Choose the most appropriate services among the available ones • Composition: Combine services to achieve a goal – Automatic Service Composition might enable programmer to become specifying what to do and not anymore how to do it! • Execution: Invoke services following programmatic conventions • Monitoring: Control the execution process 8 A trivial Example 9 An Obvious Solution 10 Inputs Composition: An Example BookName Goal Available Services CardType Login UserName CardName Password CardExpiryDate HotelReservation BookLookUp ShipItem CreditCardCheck PutInCart CarRental GetInfo FlightBooking BookShipped 11 Composition: An Example UserName Password BookName Login UserType CardType CardName CardExpiryDate CreditCardCheck BookLookUp GetInfo ProfileExists BookInStock ISBN PutInCart Approved InCart ShipItem BookShipped 12 Related Works Composition Approaches • Manual – Design-time composition – BPEL4WS (Business Process Execution Language for Web Services) • Workflow-based – Only works when the web service environment doesn’t, or only rarely changes • Automatic – AI Planning & Workflow-based 14 AI Planning B C A B C Initial State A Goal State move(a,table) move(c,a) move(b,c) Actions 15 AI Planning for Composition (1) • Planning Domain Definition Language (PDDL) – – – – PDDL is a standardized input for state-of-the-art planners PDDL and OWL-S representations are very similar. OWLS2PDDL is available. Different planners have different capabilities and by using this method we can use the best suited planner for each particular composition task. 16 AI Planning for Composition (2) • Rule-based Planning – Medjahed (2003) – Composability rules are used to determine whether two services are composable. • Message composability (output of one service is compatible with input of another). • Operation semantic composability (defines the compatibility of domains and categories and purposes of two services). • Qualitative composability (defines the requester’s preferences for quality of operations). • Composition soundness (determines whether a composition of services is reasonable). Composition templates that define dependencies between services are used. 17 AI Planning for Composition (3) • Rule-based Planning – SWORD • It uses Entity-Relation model to specify web services. • A service is modeled by its preconditions and postconditions and is represented in the form of a Horn rule that denotes postconditions are achieved if the preconditions are true. • User specifies the initial and final states. • A rule-based Expert System is used for plan generation. 18 AI Planning for Composition (4) • Situation Calculus – Activities users perform on the web can be viewed as customizations of reusable, high-level generic procedures. – Runtime customization of these generic procedures. – Situation calculus is a logic language for reasoning about action and change. – GOLOG is a logic programming language built on top of the situation calculus. – McIIrith et. al. (2001,2002), adapt and extend the GOLOG language for automatic construction of Web services. – Web Service = Action • Primitive – World-altering: change the state of the world – Information-gathering: change the state of the knowledge • Complex – Compositions of individual actions – Main Problem: GOLOG programs are difficult to create 19 AI Planning for Composition (5) • Hierarchical Task Network Planners • Composite task decomposition in HTN planning is very similar to Composite process decomposition in OWL-S. 20 AI Planning for Composition (6) • Hierarchical Task Network Planners – User must give an abstract task list. – SHOP2 – More efficient than other planning languages such as GOLOG. – OWL-S can be translated to SHOP2. – JSHOP2 is open source. – Main Problems: • Lack of parallel execution, a feature frequently needed for efficient web service usage. • Processes either must have outputs or effects, but not both. – It enables information gathering during planning. • it is not possible to directly express the semantics of OWL DL using SHOP2 axioms. • A task can not be both primitive and nonprimitive. 21 AI Planning for Composition (7) • OWLS-Xplan – An open source composition tool released Dec. 2005 – Based on Xplan planner 22 Template-based Composition • Sirin et al., Nov. 2005 • A workflow template describes the outline of activities that need to be performed to solve a problem. – Some of the activities are defined as abstract activities. • Recursive decomposition of templates. • Generic templates can be customized for a specific instance of the problem based on the users’ preferences: – Use only certified services – Try to find non-fee services – Do not buy the plane ticket if we can not reserve the hotel room. 23 Semi-automatic Composition • Current automatic composition approaches can not scale with the amount of knowledge on Semantic Web. • Sirin et al. (2004) • Automatic planner and human being can work together to generate the composite service. • The user starts the composition process by selecting one of the services registered to the engine. A query is sent to the KB to retrieve the information about the inputs of the service, and for each of the inputs, a new query is run to get the list of the possible services that can supply the appropriate data for this input. 24 Context-awareness • What is context? (in our work) – context encompasses all information about the client of a web service that may be utilized by the web service for adjusting the execution and output to provide the client with a customized and personalized behavior. • For example, – Profile • General-info: – Name, email, credit-card number,... • Preferences: – Currency, Language, ... – Location – CC/PP (Composite Capabilities / Preferences Profile) – Bandwidth 25 Context-based Adaptation What is the Screen size? Is it JavaScript Enabled? 26 Contributions A Context-aware Service Broker Profile Server Service Broker Adapted Inputs request outputs Generated Ask Adapted for inputs inputs outputs Request + Context Adapted Request Context Manager Service Requester Adapter Communication Manager Composition Selected RequestService Composer Composed Services Cache Service Matchmaker Service Specification + Profile Request Inputs Matchmaking Request Service Executor Registered Services Info Service Provider Registry Manager Service Registry 28 Matchmaking Matchmaker Service Request Keyword-based Filter Service Registry OutputsFilter Matched Services Inputs Filter Concept Manager Ontologies Ontology Processor Concepts DB Ontology Cache 29 Proposed method for Matchmaking • Fuzzy Matchmaking: – Instead of using strict levels of matching, let’s use a value between 0 and 1. – We can use the concept of Semantic Distance. 1/2 CCP C1 1/4 1/4 1/8 C2 1/16 30 Evaluation 100 95 Precision 90 85 Proposed Method Conventional Methods 80 75 70 65 60 0 10 20 30 40 50 60 70 80 90 100 Recall 31 Input/Output Matching is not sufficient • Add and Multiply services both have similar signatures: Integer Integer Add Integer Integer Multiply Integer Integer • We need to find what services actually do. 32 Publications • Y. Ganji Saffar, H. Abolhassani, R. Jalili, “An Architecture for a Context-aware Service Broker for Ubiquitous Computing Environments”, to appear, The 4th ACS/IEEE International Conference on Computer Systems and Applications (AICCSA-06), UAE, March 2006 • Y. Ganji Saffar, H. Abolhassani, “Context-aware Semantic Web Service Brokering”, to appear, 11th Computer Society of Iran Computer Conference, Iran, 2006 33 Future Plan Design & Implementation of a Service Composer 3 months Context modeling and implementation of a context manager 1 month Design & Implementation of an Adapter 1 month Adding Context-awareness capability to Matchmaker and Composer components 2 months Finalizing the thesis 1 month 34 Questions ? Main References • • • • • • • • • M. Klusch, B. Fries, and M. Khalid, “OWLS-MX: Hybrid Semantic Web Service Retrieval”, In Proceedings of 1st International AAAI Fall Symposium on Agents and the Semantic Web, Arlington VA, USA, 2005. M. Klusch, A. Gerber, and M. Schmidt, “Semantic Web Service Composition Planning with OWLS-Xplan”, AAAI Fall Symposium Series, Arlington, Virginia, USA, Nov. 2005. B. Medjahed, A. Bouguettaya, and A. K. Elmagarmid, “Composing Web services on the Semantic Web”, The VLDB Journal, vol. 12, no. 4, Nov. 2003. S. McIlraith and T. C. Son, “Adapting Golog for composition of Semantic Web services”, In Proceedings of the 8th International Conference on Knowledge Representation and Reasoning (KR2002), Toulouse, France, April 2002. S. R. Ponnekanti and A. Fox, “SWORD: A developer toolkit for Web service composition”, In Proceedings of the 11th World Wide Web Conference, Honolulu, HI, USA, 2002. D. Wu et al., “Automatic Web services composition using SHOP2”, In Proceedings of the Workshop on Planning for Web Services, Trento, Italy, June 2003. E. Sisrin, B. Parsia, and J. Hendler, “Filtering and selecting semantic web services with interactive composition techniques”, IEEE Intelligent Systems, vol. 19, no. 4, pp. 42-49, 2004 M. Paolucci et al., “Semantic matching of web services capabilities”, In Proceedings of the 1st International Semantic Web Conference (ISWC), Springer Verlag, 2002, pp. 333-347. S. Ben Mokhtar et al., “Context-aware Service Composition in Pervasive Computing Environments”, In Proceedings of the 2nd International Workshop on Rapid Integration of Software Engineering techniques (RISE’05), Heraklion Crete, Greece, Sep. 2005. 36 Discovery Approaches • Based on service inputs and outputs BookName ? ISBN • Pure logic-based – – – – Exact Subsume Plug-in Fail • Logic-based methods + Information retrieval methods – Using implicit knowledge which is available in service descriptions 37 AI Planning for Composition (2) • Situation Calculus – Every situation is defined by a world history, that is a sequence of actions. – The constant s0 describes the initial situation, that is a situation where no actions have occurred yet. – A state do(putDown(A), do(walk(L), do(pickUp(A), s0))) describes the situation created by the execution of a sequence [pickUp(A),walk(L), putDown(A)]. – A composite service is a set of atomic services which connected by procedural programming language constructs (like if-thenelse, while, for and so forth). 38