1. 1 Service Discovery using Diane Service Descriptions Ulrich Küster and Birgitta König-Ries University Jena Germany ukuester|koenig@informatik.uni-jena.de

2. 2 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Agenda  Introduction to DIANE Service Descriptions (DSD)  Improvements to Discovery Solution  estimation step handling  temporal reasoning goals  dynamic product listings  user preferences  basic composition

3. 3 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 What is DIANE and DSD?  Project goal: Complete efficient automation of service discovery, matchmaking and invocation  Assumption: Simple two-phase choreography  n stateless estimation (negotiation) steps  1 execution step  DSD language:  lightweight ontology language (object oriented)  special constructs to describe services  Keep efficient matching in mind

4. 4 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Service specific elements of DSD (1)  Operational elements  to capture world altering effects  states (instances of state ontology: Owned, Known, Printed, Shipped, Accessible, …)  Aggregational elements  Moon sells more than one item, Muller transports to a variety of countries  Usually describe sets of effects  Normal semantic: One out of a set of effects is requested / created

5. 5 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Service specific elements of DSD (2)  Selecting elements  to support configuration by the requestor (select one out of the offered effects)  to inform requestor about produced effect  variables (input / output)  Valuing elements  to express preferences of the requestor  fuzzy sets (the higher the membership, the higher the preference)  strategies (specify how to i.e. trade-off price versus shipping time, underspecified offers, …)  unbiased, deterministic, precise matching

6. 6 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Example: Excerpt from Muller offer description

7. 7 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Matching  Given fuzzy request r and configurable offer o solve the following problem: a)Compute fuzzy containment value subset Є [0, 1] of o in r (How well is the offer contained in the requested effects?) b)Where possible, configure o such as to maximize subset  Implementation descends through description graphs, fills variables with optimal values, recursively computes subset for each element, combines subset values

8. 8 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Agenda  Introduction to DIANE Service Descriptions (DSD)  Improvements to Discovery Solution  estimation step handling  Temporal reasoning goals  dynamic product listings  user preferences  basic composition

9. 9 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Improved Estimation Operation Handling  Estimation Operation (e.g. inquire price from Muller) only executed if necessary

10. 10 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Temporal Reasoning  Express shipping times  rule based  Express requirements on pickup time frame  certain time a day  minimum and maximum advance notice  minimal length of interval  Problematic: -length of interval -rule based shipping times

11. 11 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Rule based shipping time computation  "Ships in 2/3 days (national/international) if collected by 5pm"  Rule evaluation delegated to external service  integrated via estimation operation

12. 12 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Length of pickup time frame  Introduced conditions combining arbitrary attributes  Problematic during automatic offer configuration

13. 13 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Discovery Scenario II – Dynamic Product Listings  Dynamic offer descriptions using dynamic sets  Completed at the beginning of the matchmaking  Mapping rules to lift xml listing to semantic instances (minor problem: need to determine type of instance)

14. 14 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Discovery Scenario II – User Preferences

15. 15 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Discovery Scenario II – User Preferences (cont)  Fuzzy sets map extremely well to balance competing optimization goals (HDD, RAM, processor, price, …)  Fuzzy sets do not map well to rules (buy the white one if it is more than 100$ cheaper than the black one)  Price: ~==[0.0, 1800.0] 0.0 100$ difference  0.055555 difference in match value  Color: in {white [0.9444], black [1.0]}  0.5 * entity + 0.5 * price  min (entity, price, 0.5 * entity + 0,5 * price)  min(entity,price,(0.5 * entity + 0.5 * price) 5 ))

16. 16 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Discovery Scenario II – Basic Composition 1  Automatically composes offers  Ensures atomicity  Handles offers at different granularity

17. 17 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Discovery Scenario II – Basic Composition 2  Unable to solve Goal C2  DockingStation.compatibleTo{list of GTIN}  Notebook.GTIN  insufficient support for lists  Maybe next workshop ;-)

18. 18 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Discovery Scenario II – Basic Composition 3/4

19. 19 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Summary  Able to model all but one goal  Main improvements:  multi attribute conditions  arithmetic expressions  Main directions of future work:  implement list support in matchmaking  configuration in the presence of multi attribute conditions  Not in the near future:  direct support for rules

20. 20 Fourth International SWS-Challenge workshop Innsbruck, Austria, June 2007 Thank you for your attendance! Questions? Ulrich Küster DIANE project (services in ad hoc networks) http://hnsp.inf-bb.uni-jena.de/DIANE/