WSMO Implementation Workshop 2004 Woogle meets Semantic Web

WSMO Implementation Workshop 2004 Woogle meets Semantic Web Fred U. Keller, M. Stollberg, D. Fensel

Roadmap … o o o WSMO and Web Service Discovery A uniform semantic Modelling Approach Discovery Process & WSMO Woogle & a Woogle-Prototype An Application: Semantic Web Fred Conclusions & Future work

Discovery & WSMO Web Service Discovery n n n o Process of identifying web services which can (possibly) help in resolving a user‘s goal Web Service vs. Service Requires: Goal and Service specification Currently, we plan to support a wide range of techniques … n n n Require different efforts for specification Allow for different degrees of accuracy Have different computational complexity

Discovery & WSMO (II) o Three principle approaches … n n n Syntactical approaches o Keyword-based search, natural language processing techniques, Controlled vocabularies Lightweight semantic approaches o Ontologies, Action-Object-Modelling, Coarsegrained semantic description of a service Heavyweight semantic approaches o Describes the service capability in detail, takes in-out relationship into account, Fine-grained web service description » Here: we are concerned with semantic-based approaches!

WS Discovery: Modelling I o DL-Style Modelling n n The world is understood as sets of things Goal = Set of items ; WS = Set of items Checking a match means to establish a certain relationship between these sets Simple and intuitive way to see the world.

WS Discovery: Modelling Ib o Illustration: o o Intersection: G WS (weak contribution) Subsumes: G WS (strong contribution) Plugin: G WS (coverage) Equality: G = WS (exact match) Goal = WS

WS Discovery: Modelling II o Extended DL-Style Modelling n n The world is understood as sets of things Goal = Set of items ; WS = Set of items Match means presence a certain relationship between these sets o Intersection, Plugin, Subsumes, Equality Now: The language for defining these sets is very rich (unlike common DLs) goal(x) ws(x) where , are arbitrary WSML-FOL formulas. e. g. Plugin-Match: for. All x (goal(x) ws(x))

WS Discovery: Modelling III o EDL-style-Modelling with Relations n n n The world is understood as sets of things Goal = Set of items; WS has inputs i 1, …, i. N WS delivers a set of items for each tuple of input values i 1, …, i. N (WS = relation ) Matching means establishing a certain relationship between these sets wrt. (hypoth. ) executions of a web service Now: Match can refer to (hypothetical) execution of a web service goal(x) ws(x, i 1, …, i. N) pre(i 1, …, i. N) post(i 1, …, i. N, x) where , pre, post are arbitrary WSML-FOL formulas

WS Discovery: Modelling IIIb o Illustration: Goal = i 1, …, i. N WS(i 1, …, i. N) initial state poststate

WS Discovery: Modelling III o Now … Matching criteria can distinguish between single executions of a WS and an arbitrary number of such executions n Plugin-Match n n o with single execution n o goal(x) ws(x, i 1, …, i. N) pre(i 1, …, i. N) post(i 1, …, i. N, x) exists i 1, …, i. N (for. All x ( goal(x) ws(x, i 1, …, i. N)) with an arbitrary number of executions n for. All x (exists i 1, …, i. N ( goal(x) ws(x, i 1, …, i. N))

From Web Service Discovery to Service Discovery Only this o First step: WS Discovery n n o Identify possible Web Services which might resolve the user‘s goal This Ignores concrete service input / is up to applications part precond Second step: Service Discovery n o is part of WSMO Discovery Actually execute the web service and check whether a service can actually be provided (by the WS) which resolves the goal Note … By using „heavyweight“ modelling, one can in principle have an intermediate step between (1) and (2), where a web service is executed solely on a logical level (not in reality) WOOGLE!

An Example … (I) o A goal … n „I want to know about all sports events in Frankfurt in the period of Sept 29 th to Oct 14 th 2004“ goal(x) exists ? D ( x[where has. Value frankfurt, when has. Value ? D] member. Of Sports. Event and date-after-eq(? D, 29. 09. 2004) and date-before-eq(? D, 14. 10. 2004))

An Example … (II) o A web service description n „The service delivers all events for a given day in the current year in a given region in Germany“ ws(x, ? i. Day, ? i. Region) exists ? C( x[where has. Value ? C, when has. Value ? i. Day] member. Of Event and located-in(? C, ? i. Region) and year-of-date(? i. Day, current. Year) and located-in(? i. Region, germany)

An Example … (III) NO MATCH! o The proofobligation … n User selects Plugin-Match (coverage) with single invocation exists ? i. Day, i. Region (for. All x ( exists ? D ( x[where has. Value frankfurt, when has. Value ? D] member. Of Sports. Event and date-after-eq(? D, 29. 09. 2004) and date-before-eq(? D, 14. 10. 2004)) -> exists ? C( x[where has. Value ? C, when has. Value ? i. Day] member. Of Event and located-in(? C, ? i. Region) and located-in(? i. Region, germany) and year-of-date(? i. Day, current. Year) ))

An Example … (IV) o MATCH! The proofobligation … n User selects Plugin-Match (coverage) with multiple invocation for. All x (exists ? i. Day, i. Region ( exists ? D( x[where has. Value frankfurt, when has. Value ? D] member. Of Sports. Event and date-after-eq(? D, 29. 09. 2004) and date-before-eq(? D, 14. 10. 2004) ? i. Region = Hessen -> ? i. Day = ? D exists ? C( ? C = frankfurt x[where has. Value ? C, when has. Value ? i. Day] member. Of Event and located-in(? C, ? i. Region) and located-in(? i. Region, germany) and year-of-date(? i. Day, current. Year) ))

Woogle: A prototypical implementation … o o o Translation-based Approach n Supports full WSML-FOL n Translate WSML-FOL to First-order Logic Main inference engine n First-order Logic Theorem Prover (with Equality) n Currently: TPTP format for loose coupling with prover Does it scale ? n Sorry, we don‘t know … … yet : - ( ! n Interesting approach: Tuplespace-based architecture n Prepocessing techniques to extract relevant axioms n Exploit Template/Instance Scenario if possible (SWF)

A prototypical implementation … o Pros … n n n o Supports full WSML-FOL Generality gives Flexibility Prover can be reused for different tasks Cons … n n n Full First-order Logic Theorem Prover Scalability is a concern Perhaps WSML-FOL too heavy for many applications ?

A WSMO-enabled Application: Semantic Web Fred o What is Semantic Web Fred? n n Not a weird Internet junkie ; -) But … An agent-based system that supports automated goal -resolution in a cooperative environment o n A WSMO-enabled software system It combines techniques from … o o o Agent community, Semantic Web and Web Services

Semantic Web Fred (III) WOOGLE

Conclusions & Future Work o Currently n n n o General framework for Service Discovery has been outlined We are in the process of implementing a prototype implementation Implementation is used in SWF project Future work n n Evaluation of modelling approach Evaluation of the Woogle prototype