1. Automating DAML-S Web Services Composition Using SHOP2 Based on an article by Dan Wu, Bijan Parsia, Evren Sirin, James Hendler and Dana Nau in Proceedings of 2nd International Semantic Web Conference (ISWC2003), Sanibel Island, Florida, October 2003 Dumitru Roman

2. Outline Introduction Background overview DAML-S SHOP2 Translation from DAML-S to SHOP2 Conclusion

3. Introduction - Motivation: no single service capable of performing that task, but there are combinations of existing services that could - Goal: creation of compound Web services with little or no direct human intervention - How: sufficiently rich descriptions: Semantic Web languages DAML-S’ process ontology allows description of WS composition - representing services as actions -> could treat service composition as a planning problem the concept of task decomposition in HTN planning is very similar to the concept of composite process decomposition in DAML-S process ontology

4. DAML-S Process Ontology (1) - DAML-S processes: Atomic: “single step” WS Composite: compound WS -> can be decomposed into atomic or composite processes; decomposition specified by control constructs: Sequence, Unordered, Choice, If-Then-Else, Iterate, Repeat-Until, Repeat- While, Split and Split+Join Simple: abstraction of an atomic or composite process - Each process has: Preconditions: things that must be true in order for an agent to execute a service Effects: physical side-effects of execution Inputs & Outputs: knowledge preconditions & effects

5. DAML-S Process Ontology (2) SimpleProcess CompositProcess(If-Then-Else) AtomicProcess CompositProcess(Sequence) SimpleProcess CompositProcess(Unorderd) AtomicProcess CompositProcess(Sequence) - Composite process decomposition in DAML-S process ontology

6. SHOP2 - SHOP2: domain-independent HTN planning system (creates plan by task decomposition) planning domain (D): - Operator: (h(v) Pre Del Add) - describes a primitive task - Method: (h(v) Pre1 T1 … Pren Tn) - describes a compound task planning problem: triplet (S, T, D) - S: initial state, T: a task list, D: a domain description -> SHOP2 returns plan P = (p1, p2…pn): a sequence of instantiated operators method operator method operator method

7. SHOP2 example

8. Encoding DAML-S Process Models as SHOP2 Domains (1) - Two basic transformations: each atomic process is encoded as a SHOP2 operator each simple or composite process is encoded as one or more SHOP2 methods; these methods will tell how to decompose an HTN task that represents the simple or composite process

9. Encoding DAML-S Process Models as SHOP2 Domains (2) method operator method operator SimpleProcess CompositProcess(If-Then-Else) AtomicProcess CompositProcess(Sequence) SimpleProcess CompositProcess(Unorderd) AtomicProcess CompositProcess(Sequence) method operator method

10. Encoding DAML-S Process Models as SHOP2 Domains (3) Translate-Sequence-Process (Q) Input: a DAML-S definition Q of a simple process C with Sequence control construct. Output: a SHOP2 methods M. Procedure: 1. v = the list of input parameters defined for C as in Q 2. Pre = conjunct of all preconditions of C as defined in Q 3. B = Sequence control construct of C as defined in Q 4. (b1,..., bm) = the sequence of component processes of B as defined in Q 5. T = ordered task list of (b1,... bm) 6. Return M = (C(v) Pre T )

11. Conclusions no data available on complexity of WS composition problems solved simplifications: no concurrency: possible limit for efficiency all atomic processes can either have effects or outputs, but not both no open source