1. A Bottom-Up Approach to Automating Web Service Discovery, Customization, and Semantic Translation Dan Mandell and Sheila McIlraith Knowledge Systems Lab Stanford University ESSW 2003

2. Overview  Bottom-Up approach to Web service interoperation  Motivating example  BPEL4WS and automated Web service execution  The Semantic Discovery Service (SDS) and automated Web service discovery, customization, and semantic translation  Summary: contributions, future directions

3. A Bottom-Up Approach  Web services long-term goal: seamless interoperation between programs and devices  Industry provides standards, computing infrastructure, and recently choreography models akin to work in process modeling  These include WSCI, BPML, XLANG, WSFL, WSCL, WSFL, WSCL, BPSS, now BPEL4WS  Still far from seamless interoperation

4. A Bottom-Up Approach  In parallel, Semantic Web community developed languages and computing machinery for authoring and reasoning about unambiguous, machine interpretable Web content  Efforts are based on AI technology, and include RDF, RDF(S), DAML+OIL, DAML-S, and OWL  Though powerful, these efforts remain largely disconnected from industrial standards and infrastructure

5. A Bottom-Up Approach  We argue that: Web Services must embrace representation and reasoning ideas from Semantic Web community Web Services must embrace representation and reasoning ideas from Semantic Web community Must also recognize evolutionary influence of industry standards and machinery on Semantic Web services Must also recognize evolutionary influence of industry standards and machinery on Semantic Web services  From this viewpoint, we build on BPEL4WS A leading process modeling framework A leading process modeling framework Co-authored by IBM, Microsoft, BEA, SAP, Siebel Co-authored by IBM, Microsoft, BEA, SAP, Siebel Merges ideas from XLANG and WSFL Merges ideas from XLANG and WSFL  Integrate Semantic Web technology to enable automated service discovery, customization, and semantic translation

6. A Motivating Example  Consider integrating services to provide a loan finding service:

7. A Motivating Example  Possible scenario:

8. A Motivating Example  Questions: How are the service partners How are the service partners Selected?Selected? Ordered?Ordered? Invoked?Invoked? Integrated?Integrated?

9. BPEL4WS - Automated Service Execution  BPEL4WS A BPEL4WS document A BPEL4WS document Provides notation for describing WS interactions as business processes, following in tradition of workflow modelingProvides notation for describing WS interactions as business processes, following in tradition of workflow modeling Integrates services by treating them as partners that fill roles in a process modelIntegrates services by treating them as partners that fill roles in a process model Directs workflow using traditional control constructs: if, then, else, while-loopDirects workflow using traditional control constructs: if, then, else, while-loop Communication level params (e.g. service partner bindings) are described in accompanying WSDL docs Communication level params (e.g. service partner bindings) are described in accompanying WSDL docs

10. BPEL4WS - Automated Service Execution  BPWS4J Engine released by IBM alongside BPEL4WS Engine released by IBM alongside BPEL4WS Implements subset of features defined in BPEL4WS Implements subset of features defined in BPEL4WS Consumes a BPEL4WS doc along with accompanying WSDL docs defining service partner bindings to physical ports Consumes a BPEL4WS doc along with accompanying WSDL docs defining service partner bindings to physical ports Establishes a single endpoint for accessing BPEL4WS process as a Web service Establishes a single endpoint for accessing BPEL4WS process as a Web service

11. BPEL4WS - Automated Service Execution  BPEL4WS and the loan example A service provider writes a BPEL4WS doc describing the loan finding process model -- a program that orchestrates interaction of the service partners A service provider writes a BPEL4WS doc describing the loan finding process model -- a program that orchestrates interaction of the service partners BPEL4WS allows service partners to be unbound to physical ports until runtime through dynamic assignment of Service References BPEL4WS allows service partners to be unbound to physical ports until runtime through dynamic assignment of Service References Current implementation of BPWS4J does not implement Service Reference assignment, so author selects service partners at design time Current implementation of BPWS4J does not implement Service Reference assignment, so author selects service partners at design time

12. BPEL4WS - Automated Service Execution  Critical analysis of BPEL4WS automation: Limitations in BPWS4J Limitations in BPWS4J Service provider assigns partners a prioriService provider assigns partners a priori System cannot customize partner selection for each user.System cannot customize partner selection for each user. Loan finder example: user may wish to use in-state lender to benefit from in-state tax incentivesLoan finder example: user may wish to use in-state lender to benefit from in-state tax incentives If service provider defines lending partner prior to receiving user’s request, the preference is ignoredIf service provider defines lending partner prior to receiving user’s request, the preference is ignored Manual work means more responsibility and maintenance time demands for service providerManual work means more responsibility and maintenance time demands for service provider

13. BPEL4WS - Automated Service Execution  Critical analysis of BPEL4WS automation: Limitations in BPEL4WS Limitations in BPEL4WS Relies on expressivity of XML / XML SchemaRelies on expressivity of XML / XML Schema Interface-oriented: insufficient for automating many tasks.Interface-oriented: insufficient for automating many tasks. E.g., credit assessor for an ex-UK resident provides UKCreditReports, while lending service comsumes USCreditReports. Even if differ only in representation of dates, failing to recognize their semantic compatibility leaves a potentially successful integration unrealizedE.g., credit assessor for an ex-UK resident provides UKCreditReports, while lending service comsumes USCreditReports. Even if differ only in representation of dates, failing to recognize their semantic compatibility leaves a potentially successful integration unrealized Need service-oriented descriptions of service form and function in an well-defined ontology languageNeed service-oriented descriptions of service form and function in an well-defined ontology language

14. Automated, Customized, Service Discovery with SDS  To alleviate shortcomings in BPEL4WS / BPWS4J, introduce a Semantic Discovery Service (SDS) to enable automated service discovery automated service discovery automated service customization automated service customization automated semantic translation automated semantic translation  Use Semantic Web technologies to enable description of services in computer interpretable format and discovery of services with desirable properties

15. Automated, Customized, Service Discovery with SDS  Supporting technologies DAML-S: A well-defined ontology based on DAML+OIL, used to describe services DAML-S: A well-defined ontology based on DAML+OIL, used to describe services DAML Query Language (DQL): Language and protocol used for querying repositories of DAML-S service profiles. DQL server interfaces with automated reasoner operating over knowledge base (KB) of DAML-S profiles DAML Query Language (DQL): Language and protocol used for querying repositories of DAML-S service profiles. DQL server interfaces with automated reasoner operating over knowledge base (KB) of DAML-S profiles Java Theorem Prover (JTP): Hybrid reasoning system based on FOL model elimination. Use as DQL server’s automated reasoner Java Theorem Prover (JTP): Hybrid reasoning system based on FOL model elimination. Use as DQL server’s automated reasoner

16. Automated, Customized, Service Discovery with SDS  Form and function of the SDS Sits between a BPWS4J process and potential service partners Sits between a BPWS4J process and potential service partners Locates appropriate partners, acts as dynamic proxy between them and BPWS4J Locates appropriate partners, acts as dynamic proxy between them and BPWS4J

17. Automated, Customized, Service Discovery with SDS  The SDS is portable between BPWS4J actions and processes because it is: Agnostic as to the content of the service descriptions and invocation messages it receives Agnostic as to the content of the service descriptions and invocation messages it receives Stateless, with no knowledge of prior interactions or service-specific properties Stateless, with no knowledge of prior interactions or service-specific properties  The SDS enables automated service customization and semantic translation

18. Automated, Customized, Service Discovery with SDS  Automated service customization:

19. Automated, Customized, Service Discovery with SDS Interaction flow between BPWS4J, SDS, DQL server, and discovered service partners

20. Automated, Customized, Service Discovery with SDS  Automated semantic translation In the Web services context, semantic translation means redefining well-defined data types in terms of their relationships to each other via translational axioms In the Web services context, semantic translation means redefining well-defined data types in terms of their relationships to each other via translational axioms Enables integration of service partners operating on messages that differ syntactically but are semantically translatable Enables integration of service partners operating on messages that differ syntactically but are semantically translatable

21. Automated, Customized, Service Discovery with SDS  Automated semantic translation Uses translational axioms encoded as Web services to integrate partner inputs and outputs Uses translational axioms encoded as Web services to integrate partner inputs and outputs Uses a back-chaining algorithm to find sequence of service invocations, or service chain Uses a back-chaining algorithm to find sequence of service invocations, or service chain

22. Automated, Customized, Service Discovery with SDS  SDS and the loan example Recall ex-UK resident seeking a loan from an in-state lender Recall ex-UK resident seeking a loan from an in-state lender BPWS4J could not satisfy request given the constraints BPWS4J could not satisfy request given the constraints Lender located in user’s stateLender located in user’s state UKCreditReport represents dates as MM/DD/YYYY, US version uses DD/MM/YYYYUKCreditReport represents dates as MM/DD/YYYY, US version uses DD/MM/YYYY

23. Automated, Customized, Service Discovery with SDS  SDS and the loan example With SDS, the request is satisfiable With SDS, the request is satisfiable Automated service customization: include DAML-S restriction that lender partner be physically located in the user’s state in request messageAutomated service customization: include DAML-S restriction that lender partner be physically located in the user’s state in request message Automated semantic translation: back-chaining algorithm inserts a DateTranslator translational axiom:Automated semantic translation: back-chaining algorithm inserts a DateTranslator translational axiom: DateTranslator translates between UKCreditReport and USCreditReport DateTranslator translates between UKCreditReport and USCreditReport Forms service chain (Assessor -> DateTranslator -> Lender) which can successfully complete request Forms service chain (Assessor -> DateTranslator -> Lender) which can successfully complete request

24. Summary  Seamless interoperability is critical for Web services to provide an infrastructure for ubiquitous computing  Towards this goal, the bottom-up approach brings Semantic Web technology to industrial standards and computing machinery

25. Summary  By integrating the SDS with BPEL4WS, the industrial system gained the following abilities: Automatic, runtime binding of service partners Automatic, runtime binding of service partners Selection between multiple service partners based on user-defined constraints Selection between multiple service partners based on user-defined constraints Integration of service partners with syntactically different but semantically translatable service descriptions Integration of service partners with syntactically different but semantically translatable service descriptions

26. Summary  To work towards seamless interoperation, it is critical that: Web service providers publish descriptions of Web service form and function in a well- defined ontology language like DAML-S Web service providers publish descriptions of Web service form and function in a well- defined ontology language like DAML-S Web service interoperation frameworks embed semantic technology into their systems and specificaitons that is capable of reasoning about such descriptions Web service interoperation frameworks embed semantic technology into their systems and specificaitons that is capable of reasoning about such descriptions

27. THANKS - Q/A