Semantic Web Services Composition Peter Bartalos
Outline Basics of service composition Why we need semantics? Different kinds of user goals
Service composition Arranging web services into complex workflows Based on user goal
Service chaining Requirement I/O compatibility Syntactic level - WSDL Ancestor Successor O I ?
Service chaining Output: meteoowl:Temperature Input: meteoowl:Elevation meteoowl:LocalPressure getTemperature() getAtmPressure() Output: Input:
Meaningful chaining Output: meteoowl:Elevation getElevation() getAtmPressure() Output: Input: Input: meteoowl:Elevation meteoowl:LocalPressure
Problem again Output: meteoowl:Elevation Input: meteoowl:Elevation meteoowl:LocalPressure getElevation() getAtmPressure() Output: Input: In meters In ft.
Problem again Output: meteoowl:Elevation Input: meteoowl:Elevation getElevation() getAtmPressure() Output: Input: Postcondition: inMeters (getElevationReturn) Precondition: inFeet (heightAboveSeaLevel)
Considering conditions Output: meteoowl:Elevation Input: meteoowl:Elevation getElevation() getAtmPressure() Output: Input: Postcondition: inMeters (getElevationReturn) Precondition: inMeters (heightAboveSeaLevel)
Semantic service composition Basic service description I/O Binding parameters to the concepts in the ontology Pre-/post-conditions Logical expressions Predicates combined with logical operators Binding to ontological elements
Semantic compatibility evaluation Compatibility evaluation I/O Checking if O subsumes I Pre-/post-conditions Checking if Post => Pre
How it starts User goal definition Required outputs Required condition Provided inputs
How it starts User goal example The user wants the air pressure in kPa in Smolenice corrected to sea level. Required output meteoowl:AtmosphericPressure Required condition inkPa(pressure) Provided inputs geoowl:Area = Smolenice meteoowl:LocalPressure = 98.5kPa
Example composition getElevation() getAtmPressure() Output: meteoowl:Elevation Input: meteoowl:Elevation meteoowl:LocalPressure Output: meteoowl:AtmosphericPressure Input: geoowl:Area Postcondition: inkPa(pressure) Smolenice 98.5kPa
Best composition Generally we have several composite services satisfying the user goal We have to choose the best one Soft and hard constraints
Simple condition The user wants to book a flight. Condition booked(flight)
Value restriction The user wants to book a flight to Los Angeles. Condition booked(flight)^hasDestination(flight, destination) Value restriction destination = LosAngeles
Local variables The user wants to book a flight to Los Angeles and a hotel for overall price at most 500$. Condition booked(flight)^hasDestination(flight, destination)^hasPrice(flight, flightprice)^booked(hotel)^hasPrice(hotel, hotelprice) Value restriction destination = LosAngeles ^ price < 500$ Local variable definition price = flightprice + hotelprice
Soft constraints - fuzzy The user wants to book a flight for about 350$. Condition booked(flight) ^ hasPrice(flight, flightprice) Value restriction flightprice = fuzzy number
Predicate evaluation I The user wants to book a flight through weekend. Condition booked(flight)^hasFlightTime(flighttime) Value restriction isWeekendTime(flighttime)
Predicate evaluation II The user wants to book the cheapest flight. Condition booked(flight) ^ hasPrice(flight, flightprice) Value restriction cheapest(flightprice)
Conditional restriction The user wants to book a flight if it includes also insurance, else he wants to use a train. Alternatives booked(flight)^includesInsurance(flight) book(traintravel) If Then Else …
Event handling The user wants to book a flight, but if that flight is delayed by more than 24 hours, he wants to cancel the flight and take a train. Basic condition booked(flight) Event booked(flight) ^ isDelayed(flight, delay) delay > 24hour Action cancelled(flight) ^ booked(traintravel)
Open problems How to get the required representation of the user goal? How to present the results? Booking problem …? Do we need additional metadata? How to develop “composition friendly” services?