Web Service Composition : Semantic Links based Approach · Web service, Semantic Web and Semantic Web Services Nowadays Web: syntax-based Web. Semantic Web is an extension of current
Post on 10-Oct-2020
12 Views
Preview:
Transcript
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition : Semantic Linksbased Approach
Freddy Lécuéfreddy.lecue@manchester.ac.uk
November 23rd -24th 2010Ecole des Mines de Saint-Etienne, France
http://tinyurl.com/33y69h6
1/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Outline
1 Introduction
2 Web Services Composability
3 Automated Web Service Composition Approaches
4 Robust Composition
5 Quality in Web Service Composition
6 Our Approach in a France Telecom Scenario
7 Conclusions and Perspectives
2/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Motivation and Aim
As Web services proliferate:
It becomes possible tocompose them at hand;
... especially when there is norelevant single service;
Web Service Composition
Selecting and combining existingservices, available on the Web, toprovide added-value servicesfeaturing higher level functionalities.
3/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Motivation and Aim
As Web services proliferate:
It becomes possible tocompose them at hand;
... especially when there is norelevant single service;
Web Service Composition
Selecting and combining existingservices, available on the Web, toprovide added-value servicesfeaturing higher level functionalities.
3/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Automated, Dynamic and Semantic Web Service Composition
Driving Idea
1 Automated2 and Dynamic Web service composition
in the Semantic Weband in Industrial settings.
Details
4/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Web Service
A Web Service is a software application identified by a URI, whose
interfaces and binding are capable of being defined, described and
discovered by XML artifacts and supports direct interactions with other
software applications using XML based messages via Internet-based
protocols (W3C definition).
A protocol communication.
5/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Web service, Semantic Web and Semantic Web Services
Nowadays Web: syntax-based Web.Semantic Web is an extension of current Web in whichinformation is given well-defined meaning.
Ontology: a key enabling technology (RDF, OWL)
Semantic web principles applied to web servicesGive a semantics to services description;Description languages with a semantics;
6/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Related Work
Reactive Advanced Restricted Non−Classical Classical
SWS Compostion Planners
Contingency
at planning time• Service execution
at planning time (interleaving)
• Pure reactive,• Any service • Only info gathering • Only info gathering • Contingency
• Conformantservices services• Deterministic• Complete Initial States
Planning under uncertainty• Replanning (changes)
• No service execution
7/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Related Work
Functional
Input/Output
Behavioural
Message
Description
DescriptionAndName
ParametersInput Output
Parameters
Pre-Conditions Post−Conditions
Functional Description
Reactive Advanced Restricted Non−Classical Classical
SWS Compostion Planners
Level
ProcessLevel
at planning time• Service execution
at planning time (interleaving)
• Pure reactive,• Any service • Only info gathering • Only info gathering • Contingency
• Conformantservices services• Deterministic• Complete Initial States
Planning under uncertainty• Replanning (changes)Contingency
• No service execution
7/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Related Work
OWLS−XPlan2(Klusch+, 06)
Reactive Advanced Restricted Non−Classical Classical
SWS Compostion Planners
GOAL (Pfalzgraf, 06)Agora−SCP (Rao+, 06)SAWSDL−SCP (Wu+, 07)OntoMat−S (Agarwal+, 04)(Medjahed+, 03)SemaPlan (Akkiraju+, 06)Onto−Comp (Arpinar+, 05)
MetaComp (Botelho+, 07)IW−RTC (Agre+,07)
FunctionalLevel
OWLS−XPlan1 (Klusch+, 06)
Advanced Semantics
LevelProcess
FFPanner (Hoffmann+, 07)(Lassila, 04)
Roman Model (Berardi+, 05)
at planning time• Service execution
at planning time (interleaving)
• Automation
• Applicability
• Expressivity
• Composability
• Optimization
• Pure reactive,• Any service • Only info gathering • Only info gathering • Contingency
• Conformantservices services• Deterministic• Complete Initial States
PLCP (Pistore+, 05)SHOP2 (Sirin+, 02)
Planning under uncertainty• Replanning (changes)Contingency
Pure Planning
WSPLan (Peer, 05)
Golog-SCP (McIlraith+, 02)
Optop (McDermott, 02)
Mealy Model (Hull+, 03)
Optop (McDermott, 02)
• No service execution
7/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Objective
Context
Web Service Composition at Functional Level.
Sequential, Conditional and Concurrent compositions.
A Proposal
Semantic links between parameters of services→ Key elements for:
1 automated composition of stateless Web Services2 the optimisation of their candidate compositions.
8/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
What Kind of Services and Compositions?
Semantic Web Services at Functional Level
Stateless Web services:→ No Behaviour-aware Web services.
Input and Output Parameters:→ concepts in an ontology T .
Preconditions and Effects:→ properties on inputs and outputs. Details
Composition Constructs
Sequential;
Non Determinism;
Concurrency.
9/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Your Turn!!!!
Ontology and Description Logics
Elaborate the taxonomy of the TBox:
Offer ≡ ∀priceOffer .Price u∀interfacedBy.Service, Commercial_offer ≡ ∀comOffer .Offer ,
NetworkConnection ≡ ∀netPro.Provider u ∀netSpeed.Speed ,
SlowNetworkConnection ≡ NetworkConnection u ∀netSpeed.Adsl1M,
FastNetworkConnection ≡ NetworkConnection u ∀netSpeed.AdslMax ,
Speed ≡ ∀ mBytes.NoNilSpeed ,
Adsl1M ≡ Speed u ∀ mBytes.1M; AdslMax ≡ Speed u ∀ mBytes.Max ,
Max v 1M v NoNilSpeed , ZipCode @ >, Email @ >, Address @ >, PhoneNum @ >,
Invoice @ >, DeliveryID @ >, Service @ >,
ZipCode v ¬Email, Invoice v ¬Service,
IPAddress ≡ Address u ∀protocol.IP, VoIPId ≡ Address u ∀network.FTLocal,
VideoDecoder ≡ Decoder u ∀decrypt.Video
10/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition
Web Services at Functional Level
Parameters (i.e., Input and Output) of Web services insemantic Web are concepts referred to in a TBox T of anontology T :
WSDL-S, SA-WSDL (W3C Proposed Recommendation);OWL-S profile level;WSMO capability level.
11/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Web Service Composition and its Semantic Links
Semantic Link: Semantic connection between services;... more particulary between Output and Input parameters;... denoted by sly ,x and valued by SimT (Out_sy , In_sx );
Academic Contributions
F. Lécué and A. LégerA formal model for semantic Web service compositionIn ISWC, pages 385–398, Athens, USA, November 2006.
F. Lécué and A. LégerSemantic Web service composition based on a closed world assumptionIn ECOWS, pages 233-242, Zurich, Switzerland, December 2006.
12/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Web Service Composition and its Semantic Links
Semantic Link: Semantic connection between services;... more particulary between Output and Input parameters;... denoted by sly ,x and valued by SimT (Out_sy , In_sx );
Academic Contributions
F. Lécué and A. LégerA formal model for semantic Web service compositionIn ISWC, pages 385–398, Athens, USA, November 2006.
F. Lécué and A. LégerSemantic Web service composition based on a closed world assumptionIn ECOWS, pages 233-242, Zurich, Switzerland, December 2006.
12/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Web Service Composition and its Semantic Links
Semantic Link: Semantic connection between services;... more particulary between Output and Input parameters;... denoted by sly ,x and valued by SimT (Out_sy , In_sx );
Academic Contributions
F. Lécué and A. LégerA formal model for semantic Web service compositionIn ISWC, pages 385–398, Athens, USA, November 2006.
F. Lécué and A. LégerSemantic Web service composition based on a closed world assumptionIn ECOWS, pages 233-242, Zurich, Switzerland, December 2006.
12/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Web Service Composition and its Semantic Links
Semantic Link: Semantic connection between services;... more particulary between Output and Input parameters;... denoted by sly ,x and valued by SimT (Out_sy , In_sx );
SimT is reduced to the five matchmaking functions[M.Paolucci et al. ISWC’02, Li and Horrocks WWW’03]:
Exact i.e., T |= Out_sy ≡ In_sx ;PlugIn i.e., T |= Out_sy v In_sx ;Subsume i.e., T |= In_sx v Out_sy ;Intersection i.e., T 6|= Out_sy u In_sx v ⊥;Disjoint i.e., T |= Out_sy u In_sx v ⊥;
12/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Web Service Composition and its Semantic Links
Semantic Link: Semantic connection between services;... more particulary between Output and Input parameters;... denoted by sly ,x and valued by SimT (Out_sy , In_sx );
SimT is reduced to the five matchmaking functions[M.Paolucci et al. ISWC’02, Li and Horrocks WWW’03]:
Exact i.e., T |= Out_sy ≡ In_sx ;PlugIn i.e., T |= Out_sy v In_sx ;Subsume i.e., T |= In_sx v Out_sy ;Intersection i.e., T 6|= Out_sy u In_sx v ⊥;Disjoint i.e., T |= Out_sy u In_sx v ⊥;
12/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Web Service Composition and its Semantic Links
Semantic Link: Semantic connection between services;... more particulary between Output and Input parameters;... denoted by sly ,x and valued by SimT (Out_sy , In_sx );
SimT is reduced to the five matchmaking functions[M.Paolucci et al. ISWC’02, Li and Horrocks WWW’03]:
Exact i.e., T |= Out_sy ≡ In_sx ;PlugIn i.e., T |= Out_sy v In_sx ;Subsume i.e., T |= In_sx v Out_sy ;Intersection i.e., T 6|= Out_sy u In_sx v ⊥;Disjoint i.e., T |= Out_sy u In_sx v ⊥;
12/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Web Service Composition and its Semantic Links
Semantic Link: Semantic connection between services;... more particulary between Output and Input parameters;... denoted by sly ,x and valued by SimT (Out_sy , In_sx );
SimT is reduced to the five matchmaking functions[M.Paolucci et al. ISWC’02, Li and Horrocks WWW’03]:
Exact i.e., T |= Out_sy ≡ In_sx ;PlugIn i.e., T |= Out_sy v In_sx ;Subsume i.e., T |= In_sx v Out_sy ;Intersection i.e., T 6|= Out_sy u In_sx v ⊥;Disjoint i.e., T |= Out_sy u In_sx v ⊥;
12/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Web Service Composition and its Semantic Links
Semantic Link: Semantic connection between services;... more particulary between Output and Input parameters;... denoted by sly ,x and valued by SimT (Out_sy , In_sx );
SimT is reduced to the five matchmaking functions[M.Paolucci et al. ISWC’02, Li and Horrocks WWW’03]:
Exact i.e., T |= Out_sy ≡ In_sx ;PlugIn i.e., T |= Out_sy v In_sx ;Subsume i.e., T |= In_sx v Out_sy ;Intersection i.e., T 6|= Out_sy u In_sx v ⊥;Disjoint i.e., T |= Out_sy u In_sx v ⊥;
12/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Your Turn!!!!
Semantic Links
Computing Semantic links of the set of Services SA, S−A ,
SA+, SB, SC and SD.What do you require to do this?
13/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Limits of Standard Matching functions
Even if some of the latter match levels are relevant for Webservices (i.e., Semantic links) composition e.g.,
the Exact match X is clearly appropriate;the PlugIn match X is also appropriate;the Disjoint match X informs about Incompatibility;
14/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Limits of Standard Matching functions
Even if some of the latter match levels are relevant for Webservices (i.e., Semantic links) composition e.g.,
the Exact match X is clearly appropriate;the PlugIn match X is also appropriate;the Disjoint match X informs about Incompatibility;
Some match levels are not robust enough for semantic linkcomposition!
... indeed the Intersection ✗ and Subsume ✗ match levelrequire some refinements i.e., an Extra Description;
e.g.,
14/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Limits of Standard Matching functions
Even if some of the latter match levels are relevant for Webservices (i.e., Semantic links) composition e.g.,
the Exact match X is clearly appropriate;the PlugIn match X is also appropriate;the Disjoint match X informs about Incompatibility;
Some match levels are not robust enough for semantic linkcomposition!
... indeed the Intersection ✗ and Subsume ✗ match levelrequire some refinements i.e., an Extra Description;
e.g.,
14/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link
Limits of Standard Matching functions
Even if some of the latter match levels are relevant for Webservices (i.e., Semantic links) composition e.g.,
the Exact match X is clearly appropriate;the PlugIn match X is also appropriate;the Disjoint match X informs about Incompatibility;
Some match levels are not robust enough for semantic linkcomposition!
... indeed the Intersection ✗ and Subsume ✗ match levelrequire some refinements i.e., an Extra Description;
Robust semantic links in Web service composition are keycomponents to obtain robust composition.
14/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Two Methods to overcome the Robustness problem in Semantic W eb Service Composition
Non Robust Semantic Links in Web Service Composition
The open issue : How could we transform a non robustsemantic link SimT (Out_sy , In_sx ) in its robust form?
The suggested approach : by retrieving informationcontained by In_sx and not by Out_sy through ConceptDifference or Concept Abduction .
S. Brandt, R. Kusters, A. Thurhan.Approximation and difference in description logics.In KR, pages 203–214, Toulouse, France, 2002.
T. Di Noia, E. Di Sciascio et al.Abductive matchmaking using description logics.In IJCAI, pages 337–342, Acapulco, Mexico, 2003. MK.
15/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Two Methods to overcome the Robustness problem in Semantic W eb Service Composition
Non Robust Semantic Links in Web Service Composition
The open issue : How could we transform a non robustsemantic link SimT (Out_sy , In_sx ) in its robust form?
The suggested approach : by retrieving informationcontained by In_sx and not by Out_sy through ConceptDifference or Concept Abduction .
S. Brandt, R. Kusters, A. Thurhan.Approximation and difference in description logics.In KR, pages 203–214, Toulouse, France, 2002.
T. Di Noia, E. Di Sciascio et al.Abductive matchmaking using description logics.In IJCAI, pages 337–342, Acapulco, Mexico, 2003. MK.
15/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Method to overcome the Robustness problem in Semantic Web S ervice Composition
Concept Difference in Web Service Composition
Definition (Concept Difference)
The difference between two concept descriptions In_sx andOut_sy is given by
In_sx\Out_sy := min�d
{H|H uOut_sy ≡ In_sx uOut_sy}
The Extra Description In_sx\Out_sy represents what isunderspecified in Out_sy in order to completely satisfy In_sx ;
⇒ Explain why Out_sy and In_sx can not be chained by arobust semantic link.
The Common Description Out_sy u In_sx refers to informationrequired by In_sx and effectively provided by Out_sy .
16/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Method to overcome the Robustness problem in Semantic Web S ervice Composition
Concept Difference with an Example (1)
Definition (Concept Difference)
The difference between two concept descriptions In_sx andOut_sy is given by
In_sx\Out_sy := min�d
{H|H uOut_sy ≡ In_sx uOut_sy}
e.g., in case of non robust semantic link valued by theSubsume match level.
17/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Method to overcome the Robustness problem in Semantic Web S ervice Composition
Concept Difference with an Example (1)
Definition (Concept Difference)
The difference between two concept descriptions In_sx andOut_sy is given by
In_sx\Out_sy := min�d
{H|H uOut_sy ≡ In_sx uOut_sy}
e.g., in case of non robust semantic link valued by theSubsume match level.
17/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Method to overcome the Robustness problem in Semantic Web S ervice Composition
Concept Difference with an Example (2)
Definition (Concept Difference)
The difference between two concept descriptions In_sx andOut_sy is given by
In_sx\Out_sy := min�d
{H|H uOut_sy ≡ In_sx uOut_sy}
e.g., in case of non robust semantic link valued by theIntersection match level.
18/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Method to overcome the Robustness problem in Semantic Web S ervice Composition
Concept Difference with an Example (2)
Definition (Concept Difference)
The difference between two concept descriptions In_sx andOut_sy is given by
In_sx\Out_sy := min�d
{H|H uOut_sy ≡ In_sx uOut_sy}
e.g., in case of non robust semantic link valued by theIntersection match level.
18/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Method to overcome the Robustness problem in Semantic Web S ervice Composition
Concept Difference in Web Service Composition
Definition (Concept Difference)
The difference between two concept descriptions In_sx andOut_sy is given by
In_sx\Out_sy := min�d
{H|H uOut_sy ≡ In_sx uOut_sy}
Explain Where, Why a semantic link is not robust...... hence a way to replace (How) a non robust semanticlink in its robust form:
Subsume match level⇒ Exact match level;Intersection match level⇒ PlugIn match level.
19/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Method to overcome the Robustness problem in Semantic Web S ervice Composition
Your Turn!!!!
Robustness
Computing the Difference of semantic links sl1, sl2, sl3, sl4.What do you require to do this?
VoiceOverIP
TVOverIP
LiveBoxAdsl
ZipCode
PhoneNumber
PhoneNumber
PhoneNumber
NetworkConnection
NetworkConnection
NetworkConnection
PhoneNumber
Input Parameter
Output ParameterService
Semantic Link sl
EligibilityDecoder
IPAddress
VideoDecoder
VoIPId
Invoice
Slow
Fast
(Subsume Match w) Semantic Link sl4
Service SdService Sa
Service Sc
Service Sb
(PlugIn Match v)
Semantic Link sl2
Semantic Link sl3
(Intersection Match u)
(Subsume Match w)Semantic Link sl1
20/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link based Composition
Semantic Link Matrix SLM (1)
Our Proposal
An appropriate and innovative formal model:used as a starting point for the automation of WSC;that improves the way to store semantic links;that eases Web service composition and selection;... under semantic composability sx ◦ sy constraints;
Key Contribution of SLMs
controlling a set of relevant services for composition;
pre-computing all possible interactions (sx ◦ sy ).
Academic Contributions
F. Lécué and Olivier Boissier and Alexandre Delteil and A. LégerWeb Service Composition as a Composition of Valid and Robust Semantic LinksIn IJCIS, Vol 17, No 4, December 2008. World Scientific.
21/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link based Composition
Semantic link matrix: A formal model for Web service composi tion (2)
SLM Definition
A SLM is defined as Mp,q(P(SWs × (0, 1])).Rows ri,i∈{1,...,p} are labelled by Input(SWs) ⊆ T ;Columns cj,j∈{1,...,q} are labelled by (Input(SWs) ∪ β) ⊆ T ;
Each entry mi ,j of a SLM is defined as a set of (sy , score)∈ SWs × (0, 1] with (sy , score) := (sy , SimT (Out_sy , cj ))
m1,1 m1,2 . . . . . . m1,q
m2,1 m2,2 . . . . . . m2,q...
... mi ,j−1 {(sy , score)}...
mp,1 mp,2 . . . . . . mp,q
22/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Semantic Link based Composition
Your Turn!!!!
SLM Construction
Computing the SLM of SA, S−A , SA+, SB, SC and SD with
goal β := Invoice.What do you require to do this?
23/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (1)
Requirements:A TBox T to infer concepts Matching;
24/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (1)
Requirements:A TBox T to infer concepts Matching;An AI planning problem Π = 〈SWs,A, β〉;
SWs i.e., a set of possible state transitions;
24/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (1)
Requirements:A TBox T to infer concepts Matching;An AI planning problem Π = 〈SWs,A, β〉;
SWs i.e., a set of possible state transitions;A is the Initial state as an ABox. Individuals e.g., instancesof concepts Email, PhoneNum and ZipCode.
24/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (1)
Requirements:A TBox T to infer concepts Matching;An AI planning problem Π = 〈SWs,A, β〉;
SWs i.e., a set of possible state transitions;A is the Initial state as an ABox.β ⊆ T is an explicit goal representation. A TBox elemente.g., the concept Invoice.
24/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (1)
Requirements:A TBox T to infer concepts Matching;An AI planning problem Π = 〈SWs,A, β〉;
SWs i.e., a set of possible state transitions;A is the Initial state as an ABox.β ⊆ T is an explicit goal representation.
A semantic link matrixM and its semantic links;
∅ ∅ {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} ∅ ∅ {(S−
a ,1),(Sa, 12 ),(S+
a , 34 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc, 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc, 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−
a ,1),(Sa, 12 ),(S+
a , 34 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} ∅ ∅ {(S−
a ,1),(Sa, 12 ),(S+
a , 34 )} ∅ ∅
24/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (1)
Requirements:A TBox T to infer concepts Matching;An AI planning problem Π = 〈SWs,A, β〉;
SWs i.e., a set of possible state transitions;A is the Initial state as an ABox.β ⊆ T is an explicit goal representation.
A semantic link matrixM and its semantic links;
Methodology:A Regression-based approach to compute consistent,correct and complete compositions of Web services.
24/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (1)
Requirements:A TBox T to infer concepts Matching;An AI planning problem Π = 〈SWs,A, β〉;
SWs i.e., a set of possible state transitions;A is the Initial state as an ABox.β ⊆ T is an explicit goal representation.
A semantic link matrixM and its semantic links;Methodology:
A Regression-based approach to compute consistent,correct and complete compositions of Web services.
Assumptions
The set of Web services SWs is closed.
Implicit goal, Fuzzy Web service together withbehaviour description are out of scope.
24/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs is referred byM
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs is referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, A is referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
in case Email and PhoneNum and ZipCode are in A.
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, β is referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
in case Invoice is in β.
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In.
In.
?
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In.
In.
(Sd , 1)
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In.
De.
Ph.IP.
In.
(Sd , 1)
(Sd , 1)(Sd , 1)
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., the new goal De.
De.
Ph.IP.
In.
(Sd , 1)
(Sd , 1)(Sd , 1)
?
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., the new goal De.
De.
Ph.IP.
In.
(Sd , 1)
(Sd , 1)(Sd , 1)
(Sc , 34 )
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., the new goal De.
De.
Ph.IP.
In.
Ph.
Fa. (Sd , 1)
(Sd , 1)(Sd , 1)
(Sc , 34 )
(Sc , 34 )
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., the new goal Fa.
De.
Ph.IP.
In.
Ph.
Fa.
?(Sd , 1)
(Sd , 1)(Sd , 1)
(Sc , 34 )
(Sc , 34 )
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., the new goal Fa.
De.
Ph.IP.
In.
Ph.
Fa.
(Sa, 12 )
(Sd , 1)
(Sd , 1)(Sd , 1)
(Sc , 34 )
(Sc , 34 )
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., the new goal Fa.
De.
Ph.IP.
In.
Ph.
Fa.
Ph.
Em.
Zi.
(Sa, 12 )
(Sa, 12 )
(Sa, 12 ) (Sd , 1)
(Sd , 1)(Sd , 1)
(Sc , 34 )
(Sc , 34 )
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., Fa., the new goal IP.
IP.
Ph.De.
In.
X
Sc ◦ Sa
(Sd , 1)
(Sd , 1)(Sd , 1)
?
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., Fa., the new goal IP.
IP.
Ph.De.
In.
X
Sc ◦ Sa
(Sd , 1)
(Sd , 1)(Sd , 1)
(Sb, 14 )
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., Fa., the new goal IP.
IP.
Ph.De.
In.
Ph.
Sl.
X
Sc ◦ Sa
(Sd , 1)
(Sd , 1)(Sd , 1)
(Sb, 14 )
(Sb, 14 )
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., Fa., the new goal IP.
IP.
Ph.De.
In.
Ph.
Sl.
X
Sc ◦ Sa
(Sd , 1)
(Sd , 1)(Sd , 1)
(Sb, 14 )
(Sb, 14 )
?
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., Fa., the new goal IP.
IP.
Ph.De.
In.
Ph.
Sl.
X
Sc ◦ Sa
(Sd , 1)
(Sd , 1)(Sd , 1)
(Sb, 14 )
(Sb, 14 )
(S+a , 3
4 )
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., Fa., the new goal IP.
IP.
Ph.De.
In.
Ph.
Sl.
X
Em.
Ph.
Zi.
Sc ◦ Sa
(Sd , 1)
(Sd , 1)(Sd , 1)
(Sb, 14 )
(Sb, 14 )
(S+a , 3
4 )
(S+a , 3
4 )
(S+a , 3
4 ) 25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., Fa. and IP., with a potential solution:
IP.
Ph.De.
In.
Ph.
Sl.
X
X
Sc ◦ Sa
(Sd , 1)
(Sd , 1)(Sd , 1)
(Sb, 14 )
(Sb, 14 )S+
a
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
AI planning and SLMs: A regression-based approach (2)
Suppose a SLMM and Π = 〈SWs,A,β〉;By the SLM definition, SWs, A and β are referred byM.
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} ∅ ∅ ∅ ∅ ∅ ∅
∅ ∅ ∅ ∅ ∅ ∅ ∅ {(Sd ,1)}
∅ {(Sc , 34 )} {(S−
a , 12 ),(Sa, 1
2 ),(S+a ,1)} {(Sb, 1
4 )} ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ {(Sd ,1)}
∅ ∅ ∅ {(Sb, 14 )} ∅ ∅ ∅ ∅
∅ ∅ {(S−a , 1
2 ),(Sa, 12 ),(S+
a ,1)} ∅ ∅ {(S−a ,1),(Sa, 1
2 ),(S+a , 3
4 )} ∅ ∅
The composition process: a recursive and regression-based approach;From the goal In., De., Fa. and IP., with other solutions e.g.,
IP.
Ph.De.
In.
Ph.
Sl.
X
X
Sc ◦ Sa
(Sd , 1)
(Sd , 1)(Sd , 1)
(Sb, 14 )
(Sb, 14 )S−
a
25/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Web Service Composition as an AI Planning Problem
Your Turn!!!!
Computation of Composition
Computing the candidate compositions that achieve goalβ := Invoice with Initial SituationA := {Email , PhoneNum, ZipCode}.
What do you require to do this?
26/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
What about Robust Web Service Composition?
Robust Service Composition
A full Automation of Web service composition?
Still not a reality... especially in case the latter compositionis consisting of non robust semantic links;However two ways to obtain the Extra Description Hrequired by non robust semantic links:
discovering new relevant services but time consuming;relaxing some constraintsinfv{In_sx\Out_sy |〈sy , SimT (Out_sy , In_sx ), sx 〉}.
27/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
What about Robust Web Service Composition?
Robust Service Composition
A full Automation of Web service composition?
Still not a reality... especially in case the latter compositionis consisting of non robust semantic links;However two ways to obtain the Extra Description Hrequired by non robust semantic links:
discovering new relevant services but time consuming;relaxing some constraintsinfv{In_sx\Out_sy |〈sy , SimT (Out_sy , In_sx ), sx 〉}.
27/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
How to Perform Robust Web Service Composition?
Robust Service Composition... by retrieving new Web servic es
A full Automation of Web service composition?
By discovering new relevant Web services.
The main constraint is related to the complexity ofcomposition.
28/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
How to Perform Robust Web Service Composition?
Robust Service Composition... by retrieving new Web servic es
A full Automation of Web service composition?
By discovering new relevant Web services.
The main constraint is related to the complexity ofcomposition.
28/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
How to Perform Robust Web Service Composition?
Robust Service Composition... by retrieving new Web servic es
A full Automation of Web service composition?
By discovering new relevant Web services.
The main constraint is related to the complexity ofcomposition.
28/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
How to Perform Robust Web Service Composition?
Robust Service Composition... by retrieving new Web servic es
A full Automation of Web service composition?
By discovering new relevant Web services.
The main constraint is related to the complexity ofcomposition.
28/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
How to Perform Robust Web Service Composition?
Robust Service Composition... by retrieving new Web servic es
A full Automation of Web service composition?
By discovering new relevant Web services.
The main constraint is related to the complexity ofcomposition.
28/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
How to Perform Robust Web Service Composition?
Robust Service Composition... by retrieving new Web servic es
A full Automation of Web service composition?
By discovering new relevant Web services.
The main constraint is related to the complexity ofcomposition.
28/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
How to Perform Robust Web Service Composition?
Robust Service Composition... by relaxing some constraint s
A full Automation of Web service composition?
By relaxing some constraints during composition:H := infv{In_sx\Out_sy |〈sy , SimT (Out_sy , In_sx), sx 〉}.
e.g., by suggesting H to the end user as requiredinformation the composition process.
For instance
H := infv{H1, H2, H3, H4, H5}
29/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
How to Perform Robust Web Service Composition?
Robust Service Composition... by relaxing some constraint s
A full Automation of Web service composition?
By relaxing some constraints during composition:H := infv{In_sx\Out_sy |〈sy , SimT (Out_sy , In_sx), sx 〉}.
e.g., by suggesting H to the end user as requiredinformation the composition process.
For instance
H := infv{H1, H2, H3, H4, H5}
29/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
How to Perform Robust Web Service Composition?
Robust Service Composition... by relaxing some constraint s
A full Automation of Web service composition?
By relaxing some constraints during composition:H := infv{In_sx\Out_sy |〈sy , SimT (Out_sy , In_sx), sx 〉}.
e.g., by suggesting H to the end user as requiredinformation the composition process.
For instance
H := infv{H1, H2, H3, H4, H5}
A Concluding Remark
In both cases the more robust semantic links in acomposition the better.
29/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Your Turn!!!!
Computation of H
Computing H of the following compositions(Sx , Sy ∈ {SA, S−
A , S+A }).
What do you require to do this?
GoalComposition
Valid Semantic LinksA service such that Sx, Sy ∈ {S−a , Sa, S
+a }
Output Parameters of Web services
Input Parameters of Web services
Invoice
Semantic Link sl1
VideoDecoder
XNetworkConnection
ZipCode
XNetworkConnection
ZipCode
VoIPId
Decoder
PhoneNum
PhoneNum
PhoneNum
IPAddress
Sx
Sy
Sc
Sb
Semantic Link sl4
Semantic Link sl3
Sd
FastNC
SlowNC
PhoneNum
Invoiceb : PhoneNum
a : Email
c : ZipCode
Goal β
ABox A
PhoneNum
Goal βABox A
X ∈ {Slow,Fast, ∅}
Semantic Link sl2
Sx, Sy
30/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Automated Computation of Robustness in Composition
Approach
A full Automation of Web service composition?
An agent-based negotiation used to solicit the additionalsemantic descriptions required for robustness .
31/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Automated Computation of Robustness in Composition
Negotiating Robustness of Composition
Agent-based Negotiation as a Process for Achieving Robustness in Composition
Agents represent service providers;
Direct negotiation between agents, no need to involve third party or mediator;
Agents may exchange counter proposals and impose conditions over the use ofservices;
The negotiation process is supported by a negotiation protocol.
Why Yet Another Protocol?
Typical approaches (e.g. Contract-Net, English Auction) give the initiator morecontrol over the negotiation;
The role of participants is limited to providing information/proposals.
In the Proposed Approach
Agents have more control over the negotiation - they can exchange counterproposals;
Agreements may occur at different levels of granularity.
32/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Automated Computation of Robustness in Composition
Protocol for Robust Composition
1 The Initiator agent sends a CFP to other agent(s);2 Participant agent decodes XML encoding and
consults its service providers regarding the MostSpecific Description.
3 If the participant is able to contribute, it willrespond with Propose otherwise Refuse.
4 In Propose, the message contains XML encodingof the proposed Extra Description, which issubsumed by the Most Specific Description. Thismay be accompanied e.g., by cost.
5 On receiving a proposal the initiator agent maydecide to accept the proposal or to iterate theprocess by issuing a revised CFP with newrequired description. The latter is subsumed by theoriginal Most Specific Description and specifies theelements of which are not yet covered by the set ofreceived proposal.
6 The protocol ends when the Initiator agent sendsAccept-Proposal to a set of agent, or when it doesnot issue a new CFP.
33/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Automated Computation of Robustness in Composition
Enabling Automated Negotiation
Agents require specific reasoning and decision making mechanisms (M) that feed intovarious communicative actions (S) in the protocol.
M1: Need for Most Specific Description : Amechanism that enable agents to compute andrealize the need for Most Specific Description.
M2: Proposal Formation : A mechanism foragents to compute required information andgenerate a proposal.
M3: Proposal Evaluation and Ranking : Usesthe well known set-partitioning problem forproposal evaluation.
M4: Notification of Decision : A mechanism tonotify participating agents about the outcome oftheir proposals.
M5: Acknowledgment : A mechanism thatallow participating agents to acknowledge theuse of its information.
34/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Composition Model
Composition Result Modelling
Process Model as a Statechart
Its states refer to services;
Its transitions are labelled with semantic links;
with basic composition constructs.
Legend
Connection
Slow
Output Parameter
Input Parameter
T: Task
s: Service
Semantic Link sl
Network
Connection
Network
s1 s5
s2 s3
ANDBranching
s6
s7
sl15,7
sl12,3
sl11,4
sl15,6
sl14,5
sl16,8
T4
T2 T3 T6
T7
T8T1 T5
sl11,2 sl13,5
s4
s8
OR-Branching
Sequence
sl17,8
35/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Quality Criteria for Semantic Links & Services
q(sli ,j) for Elementary Semantic Links sli ,j
Common Description rate qcd ∈ (0, 1]:
qcd(sli ,j) =|lcs(Out_si , In_sj)|
|H∈〈L,Out_si ,In_sj ,T 〉| + |lcs(Out_si , In_sj)|
Matching Quality qm ∈ (0, 1], valued by SimT (Out_si , In_sj)
(Exact: 1, PlugIn: 34 , Subsume: 1
2 , Intersection: 14 ).
36/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Quality Criteria for Semantic Links & Services
q(sli ,j) for Elementary Semantic Links sli ,j
Common Description rate qcd ∈ (0, 1]:
qcd(sli ,j) =|lcs(Out_si , In_sj)|
|H∈〈L,Out_si ,In_sj ,T 〉| + |lcs(Out_si , In_sj)|
Matching Quality qm ∈ (0, 1], valued by SimT (Out_si , In_sj)
(Exact: 1, PlugIn: 34 , Subsume: 1
2 , Intersection: 14 ).
36/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Quality Criteria for Semantic Links & Services
q(sli ,j) for Elementary Semantic Links sli ,j
Common Description rate qcd ∈ (0, 1]:
qcd(sli ,j) =|lcs(Out_si , In_sj)|
|H∈〈L,Out_si ,In_sj ,T 〉| + |lcs(Out_si , In_sj)|
Matching Quality qm ∈ (0, 1], valued by SimT (Out_si , In_sj)
(Exact: 1, PlugIn: 34 , Subsume: 1
2 , Intersection: 14 ).
q(si ) for Elementary Services si
Execution Price qpr ∈ <+;
Response Time qt ∈ <+.
36/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Quality Criteria for Semantic Links & Services
q(sli ,j) for Elementary Semantic Links sli ,j
Common Description rate qcd ∈ (0, 1]:
qcd(sli ,j) =|lcs(Out_si , In_sj)|
|H∈〈L,Out_si ,In_sj ,T 〉| + |lcs(Out_si , In_sj)|
Matching Quality qm ∈ (0, 1], valued by SimT (Out_si , In_sj)
(Exact: 1, PlugIn: 34 , Subsume: 1
2 , Intersection: 14 ).
q(si ) for Elementary Services si
Execution Price qpr ∈ <+;
Response Time qt ∈ <+.
QoS-extended quality vector of a semantic link sli ,j∗q (sli ,j)
.= (q(si ), q(sli ,j), q(sj ))
36/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Your Turn!!!!
Computation of quality
Computing the semantic quality ql∈{cd ,m} of each semanticlink (Sx , Sy ∈ {SA, S−
A , S+A }).
What do you require to do this?
GoalComposition
Valid Semantic LinksA service such that Sx, Sy ∈ {S−a , Sa, S
+a }
Output Parameters of Web services
Input Parameters of Web services
Invoice
Semantic Link sl1
VideoDecoder
XNetworkConnection
ZipCode
XNetworkConnection
ZipCode
VoIPId
Decoder
PhoneNum
PhoneNum
PhoneNum
IPAddress
Sx
Sy
Sc
Sb
Semantic Link sl4
Semantic Link sl3
Sd
FastNC
SlowNC
PhoneNum
Invoiceb : PhoneNum
a : Email
c : ZipCode
Goal β
ABox A
PhoneNum
Goal βABox A
X ∈ {Slow,Fast, ∅}
Semantic Link sl2
Sx, Sy 37/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Quality Criteria for Composition
Quality Aggregation Rules for Compositions
CompositionQuality Criterion
ConstructSemantic Non Functional
Qcd Qm Qt Qpr
Sequential/ 1|sl|
∑
sl qcd(sl)∏
sl qm(sl)∑
s qt(s) ∑
s qpr(s)AND- Branching maxs qt(s)
OR-Branching∑
sl qcd(sl).psl∑
sl qm(sl).psl∑
s qt(s).ps∑
s qpr (s).ps
Legend
Connection
Slow
Output Parameter
Input Parameter
T: Task
s: Service
Semantic Link sl
Network
Connection
Network
s1 s5
s2 s3
ANDBranching
s6
s7
sl15,7
sl12,3
sl11,4
sl15,6
sl14,5
sl16,8
T4
T2 T3 T6
T7
T8T1 T5
sl11,2 sl13,5
s4
s8
OR-Branching
Sequence
sl17,8
38/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Quality Criteria for Composition
Quality Aggregation Rules for Compositions
CompositionQuality Criterion
ConstructSemantic Non Functional
Qcd Qm Qt Qpr
Sequential/ 1|sl|
∑
sl qcd(sl)∏
sl qm(sl)∑
s qt(s) ∑
s qpr(s)AND- Branching maxs qt(s)
OR-Branching∑
sl qcd(sl).psl∑
sl qm(sl).psl∑
s qt(s).ps∑
s qpr (s).ps
Legend
Connection
Slow
Output Parameter
Input Parameter
T: Task
s: Service
Semantic Link sl
Network
Connection
Network
s1 s5
s2 s3 s6sl12,3 sl15,6 sl16,8
T2 T3 T6
T8T1 T5
sl11,2 sl13,5
s8
Sequence
38/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Quality Criteria for Composition
Quality Aggregation Rules for Compositions
CompositionQuality Criterion
ConstructSemantic Non Functional
Qcd Qm Qt Qpr
Sequential/ 1|sl|
∑
sl qcd(sl)∏
sl qm(sl)∑
s qt(s) ∑
s qpr(s)AND- Branching maxs qt(s)
OR-Branching∑
sl qcd(sl).psl∑
sl qm(sl).psl∑
s qt(s).ps∑
s qpr (s).ps
Output Parameter
Input Parameter
Semantic Link sl
T: Task
s: Service
Legend
sl15,7
s5AND
Branching
s6
s7
sl15,6
T6
T7
T5
38/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Quality Criteria for Composition
Quality Aggregation Rules for Compositions
CompositionQuality Criterion
ConstructSemantic Non Functional
Qcd Qm Qt Qpr
Sequential/ 1|sl|
∑
sl qcd(sl)∏
sl qm(sl)∑
s qt(s) ∑
s qpr(s)AND- Branching maxs qt(s)
OR-Branching∑
sl qcd(sl).psl∑
sl qm(sl).psl∑
s qt(s).ps∑
s qpr (s).ps
Connection
Slow
Output Parameter
Input Parameter
Semantic Link sl
T: Task
s: Service
Network
Connection
Legend
Networksl11,2
s1
s2
sl11,4T4
T2
T1
s4
OR-Branching
38/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Quality Criteria for Composition
Quality Aggregation Rules for Compositions
CompositionQuality Criterion
ConstructSemantic Non Functional
Qcd Qm Qt Qpr
Sequential/ 1|sl|
∑
sl qcd(sl)∏
sl qm(sl)∑
s qt(s) ∑
s qpr(s)AND- Branching maxs qt(s)
OR-Branching∑
sl qcd(sl).psl∑
sl qm(sl).psl∑
s qt(s).ps∑
s qpr (s).ps
A Quality Vector for Web Service Composition
“A” way to differentiate compositions:
Q(c).= (Qcd (c), Qm(c), Qt (c), Qpr (c))
38/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Quality Model
Your Turn!!!!
Computation of quality
Computing the semantic quality Ql∈{cd ,m} of eachcomposition (Sx , Sy ∈ {SA, S−
A , S+A }).
What do you require to do this?
GoalComposition
Valid Semantic LinksA service such that Sx, Sy ∈ {S−a , Sa, S
+a }
Output Parameters of Web services
Input Parameters of Web services
Invoice
Semantic Link sl1
VideoDecoder
XNetworkConnection
ZipCode
XNetworkConnection
ZipCode
VoIPId
Decoder
PhoneNum
PhoneNum
PhoneNum
IPAddress
Sx
Sy
Sc
Sb
Semantic Link sl4
Semantic Link sl3
Sd
FastNC
SlowNC
PhoneNum
Invoiceb : PhoneNum
a : Email
c : ZipCode
Goal β
ABox A
PhoneNum
Goal βABox A
X ∈ {Slow,Fast, ∅}
Semantic Link sl2
Sx, Sy 39/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
CSP
Web Service Composition Driven CSP
CSP Formalization
Formalization as a triple (T , D, C):T is the set of tasks (variables) {T1, T2, ..., Tn};D is the set of domains {D1, D2, ..., Dn} i.e., services;C is the set of constraints i.e., local CL and global CG.
e.g.,1
|slAi,j |
∑
slAi,j
qcd(slAi,j) ≥ v , v ∈ [0, 1]
∑
Ti
qpr (Ti) ≤ v , v ∈ <+
Main Goal to Achieve
An assignment (si , Ti)1≤i≤n i.e., (service, task)with si,1≤i≤n ∈ Di,1≤i≤n;which satisfies all the constraints C.
40/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
CSP
Your Turn!!!!
Complexity
What is the number of potential compositions of n taskswith m potential services per task?
Legend
Connection
Slow
Output Parameter
Input Parameter
T: Task
s: Service
Semantic Link sl
Network
Connection
Network
s1 s5
s2 s3
ANDBranching
s6
s7
sl15,7
sl12,3
sl11,4
sl15,6
sl14,5
sl16,8
T4
T2 T3 T6
T7
T8T1 T5
sl11,2 sl13,5
s4
s8
OR-Branching
Sequence
sl17,8
41/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
CSP
Your Turn!!!!
Computation of quality
Modelling the CSP problem using (Sx , Sy ∈ {SA, S−A , S+
A })?
What do you require to do this?
GoalComposition
Valid Semantic LinksA service such that Sx, Sy ∈ {S−a , Sa, S
+a }
Output Parameters of Web services
Input Parameters of Web services
Invoice
Semantic Link sl1
VideoDecoder
XNetworkConnection
ZipCode
XNetworkConnection
ZipCode
VoIPId
Decoder
PhoneNum
PhoneNum
PhoneNum
IPAddress
Sx
Sy
Sc
Sb
Semantic Link sl4
Semantic Link sl3
Sd
FastNC
SlowNC
PhoneNum
Invoiceb : PhoneNum
a : Email
c : ZipCode
Goal β
ABox A
PhoneNum
Goal βABox A
X ∈ {Slow,Fast, ∅}
Semantic Link sl2
Sx, Sy42/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Scalable Approach
A Stochastic Search Method (1)
Principles
Sacrificing completness (i.e., all solutions) for speed;
Based on a simple idea: computing “a single” solution.
Our Approach
Adaptation of the Hill Climbing algorithm.→ Appropriate for a large number of services.
S. Russell and P. Norvig.Artificial Intelligence: A Modern Approach.
Ed. Prentice-Hall, 1995.
Computational Complexity
CSP based search methods: Exponential!
Stochastic search methods (e.g., Hill Climbing) scale better!43/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Scalable Approach
A Stochastic Search Method (2)
Requirements
An evaluation function f for each composition c:
f (c) =ωcdQ̂cd (c) + ωmQ̂m(c)
ωprQ̂pr (c) + ωt Q̂t(c)
An adjacency function: c1 and c2 are adjacent to eachother if they differ in exactly one assignment (s, T ).
Algorithm in Details
1) Let’s start with a random composition cfinal .2) f -Evaluation of all ci ,1≤i≤n adjacent to cfinal .
If ∃i such that f (cfinal ) ≤ f (ci ) then f (cfinal )← f (ci).
3) Iteration until all constraints are satisfied by cfinal .
4) If no solution, constraints relaxing.
44/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Scalable Approach
Your Turn!!!!
Stochastic Search Method
Let’s elaborate the adjacency function?What do you require to do this?
Compute the best compostion regarding the value of theirevaluation function?
What do you require to do this?
Legend
Connection
Slow
Output Parameter
Input Parameter
T: Task
s: Service
Semantic Link sl
Network
Connection
Network
s1 s5
s2 s3
ANDBranching
s6
s7
sl15,7
sl12,3
sl11,4
sl15,6
sl14,5
sl16,8
T4
T2 T3 T6
T7
T8T1 T5
sl11,2 sl13,5
s4
s8
OR-Branching
Sequence
sl17,8
45/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
A Scalable Approach
Experimentation
Evolution of Constraints SatisfactionThe more tasks, services the more time consuming!
Evolution of Composition Quality
Optimal composition: High Time consuming!
Compositions that satisfy constraints: More scalable!
Search Process vs. DL Reasoning (|T | > 100, |s| > 350)
DL reasoning is the most time consuming process!Large number of potential semantic links.Critical complexity of DL abduction.
Vs. State-of-the-art Approaches (T = 300 |s| > 280)
Adoption of stochastic search method for large domains!No exponential search required.
46/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
CSOP
Composition Optimization Driven CSOP
CSOP Formalization (T , D, C, f )
T is the set of tasks (variables) {T1, T2, ..., Tn};
D is the set of domains {D1, D2, ..., Dn} i.e., services;C is the set of constraints i.e., local CL and global CG;
e.g.,1
|slAi,j |
∑
slAi,j
qcd(slAi,j) ≥ v , v ∈ [0, 1]
∑
Ti
qpr (Ti) ≤ v , v ∈ <+
f is an evaluation function.
Main Goal to Achieve
An assignment (si , Ti)1≤i≤n i.e., (service, task) Problemwith si,1≤i≤n ∈ Di,1≤i≤n;which satisfies all the constraints C;which is optimal in terms of QoS or functional quality.
47/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Local and Naive Global Selection
Local Selection on slAi ,j
Enforcing specific services for both tasks Ti and Tj ;
Quality constraints may be not satisfied, leading to asuboptimal composition.
Naive Global Selection
Exhaustive search of the optimal composition;⇒ Exponential in the number of abstract semantic links.
Our Approach
An integer linear programming IP based global selection, which
further constrains semantic links;
meets a given objective.48/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
IP Based Global Selection
Optimal Composition and IP Problem
The problem of computing an optimal composition is mappedinto an IP problem.
Inputs of the IP Problem
An objective function;
A set of integer variables (restricted to values 0 or 1);
A set of constraints (equalities or inequalities)
where both the objective function and the constraints are linear.
Outputs of the IP Problem
The maximum (or minimum) value of the objective function;
Values of variables at this maximum (minimum).49/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Objective Function
Step 1: Compositions Computation
Computation of Qλ,1≤λ≤pl ,l∈{r ,cd ,m} i.e., quality values of the p potential
compositions.
Step 2: Scaling
Quality values Qλr , Qλ
cd , Qλm are then scaled according to:
∼
Qλ
l =
{
Qλl −Qmin
lQmax
l −Qminl
if Qmaxl −Qmin
l 6= 0l ∈ {r , cd , m}
1 if Qmaxl −Qmin
l = 0
Step 3: Objective Function
max1≤λ≤p
(
∑
l∈{r ,cd ,m}
(∼
Qλ
l × ωl
)
)
50/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Your Turn!!!!
Scaling candidate compositions
Computing the scale based compositions∼
Qλ
l∈{r ,cd ,m}.What do you require to do this?
GoalComposition
Valid Semantic LinksA service such that Sx, Sy ∈ {S−a , Sa, S
+a }
Output Parameters of Web services
Input Parameters of Web services
Invoice
Semantic Link sl1
VideoDecoder
XNetworkConnection
ZipCode
XNetworkConnection
ZipCode
VoIPId
Decoder
PhoneNum
PhoneNum
PhoneNum
IPAddress
Sx
Sy
Sc
Sb
Semantic Link sl4
Semantic Link sl3
Sd
FastNC
SlowNC
PhoneNum
Invoiceb : PhoneNum
a : Email
c : ZipCode
Goal β
ABox A
PhoneNum
Goal βABox A
X ∈ {Slow,Fast, ∅}
Semantic Link sl2
Sx, Sy
51/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Integer Variables
Integer Variables
An integer variable yki ,j ∈ {0, 1} for every candidate link
slk ,1≤k≤ni ,j of an abstract link slAi ,j indicates the selection or
exclusion of link slki ,j in the IP problem
T: Task
Output Parameter
Input Parameter
Candidates Candidates
Legend
s: Candidate Service
Abstract
CandidateAbstract Semantic Links
Semantic Link slAi,j
T1 T2
s2
s2′
s2′′
s1
sl21,2
sl31,2
sl11,2
(y21,2)
(y11,2)
(y31,2)
Integer Variable yki,j
slA1,2
Semantic Link slki,j
52/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Constraints of IP Problem
Allocation Constraint
Only one candidate link is selected for each abstract link slAi ,j .n∑
k=1
yki ,j = 1, ∀slAi ,j
T: Task
Output Parameter
Input Parameter
Candidates Candidates
Legend
s: Candidate Service
Abstract
CandidateAbstract Semantic Links
Semantic Link slAi,j
T1 T2
s2
s2′
s2′′
s1
sl21,2
sl31,2
sl11,2
(y21,2 = 0)
(y11,2 = 1)
(y31,2 = 0)
Integer Variable yki,j
slA1,2
Semantic Link slki,j
53/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Constraints of IP Problem
Incompatibility Constraint
Some semantic links slki ,j and sl lj ,β are incompatible in acomposition.
yki ,j + y l
j ,β ≤ 1, ∀slAi ,j ∀slAj ,β
T: Task
Output Parameter
Input Parameter
Candidates Candidates CandidatesCandidatesCandidates Legend
s: Candidate Service
Abstract
Candidate
Integer Variable yki,j
(y31,2 = 0)
sl11,2 (y11,2 = 0)
sl32,3 (y32,3 = 0)
sl22,3 (y22,3 = 1)sl21,2 (y2
1,2 = 1)
T1 T2 T3
s2 s3s1
slA2,3slA1,2
s′2
s′′2
sl12,3 (y12,3 = 0) Semantic Link slki,j
Semantic Link clAi,j
sl31,2
53/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Constraints of IP Problem
Constraints on Quality values of Compositions
Robustness Constraint for capturing and constraining therobustness quality of a semantic link composition;
Common Description Rate Constraint;
Matching Quality Constraint.
Local Constraints
Such constraints can predicate on properties of a single link(e.g., local robustness).
53/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Flexibility (and Extension) of Constraints
Suggested Constraints (Reminder)
Allocation Constraint;
Incompatibility Constraint;Constraints on Quality values of Compositions:
Robustness Constraint;Common Description Rate Constraint;Matching Quality Constraint.
Local Constraints.
⇒ The method for translating the problem of selecting anoptimal composition into an IP problem is generic.
⇒ Other semantic criteria to value semantic links can beaccommodated.
54/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Your Turn!!!!
Modeling IP based Optimal Web Service Composition
Modeling the Composition optimization problem in an IPProblem.
What do you require to do this?
GoalComposition
Valid Semantic LinksA service such that Sx, Sy ∈ {S−a , Sa, S
+a }
Output Parameters of Web services
Input Parameters of Web services
Invoice
Semantic Link sl1
VideoDecoder
XNetworkConnection
ZipCode
XNetworkConnection
ZipCode
VoIPId
Decoder
PhoneNum
PhoneNum
PhoneNum
IPAddress
Sx
Sy
Sc
Sb
Semantic Link sl4
Semantic Link sl3
Sd
FastNC
SlowNC
PhoneNum
Invoiceb : PhoneNum
a : Email
c : ZipCode
Goal β
ABox A
PhoneNum
Goal βABox A
X ∈ {Slow,Fast, ∅}
Semantic Link sl2
Sx, Sy55/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
IP Based Approach
Computational Complexity and Experimentation
Computational Complexity
The optimization problem is equivalent to an IP problem.⇒ NP-hard!
Experimentation
Exhaustive search based: High computation cost.
IP based: Acceptable computation cost.
0
2000
4000
6000
8000
10000
0 100 200 300 400 500
Number of Abstract Semantic Links in Composition
Global Selection Using Exhaustive SearchGlobal Selection Using IP
Local Optimization Based-Selection
0
1000
2000
3000
4000
5000
6000
7000
8000
0 20 40 60 80 100 120 140
Number of Candidate Semantic Links in Composition
Global Selection Using Exhaustive SearchGlobal Selection Using IP
Local Optimization Based-Selection
56/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
GA Based Approach
A Genetic Algorithm based Method
Principles for computing the optimal solution
simulating the evolution of an initial population until survivalof best fitted compositions satisfying constraints C.
GA Parameters
Genotype.
Initial Population: compositions randomly selected.
Global, Local Constraints: CG, CL.Fitness Function: f (c)
ωcdQ̂cd (c) + ωmQ̂m(c)
ωprQ̂pr (c) + ωt Q̂t(c)− ωpe.
genmaxgen
.∑
l∈{pr ,t,cd,m}
( ∆Q̂l
Q̂maxl (c) − Q̂min
l (c)
)2
Operators on Genotypes: crossover, mutation, selection.
Stopping Criterion: until the constraints are met!57/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
GA Based Approach
Your Turn!!!!
Modeling GA based Optimal Web Service Composition
Modeling the Composition optimization problem in an GAProblem.
What do you require to do this?
T: Task s: Service
s2′
s5
T1
s1
T2
s2
T3 T5 T6
s6
T7
s7
T8
s8
T4
s4
s3
Selected si for Ti
58/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
GA Based Approach
Experimentation
Benefits of Combining QoS and Functional Criteria
Limiting the costs of data integration.
Evolution of Composition Quality (up to |T | = 500, |s| = 500)
Complexity in the number of tasks and services;
Variables: population size and number of generations;
... but could be inappropriate.
GA Process vs. DL Reasoning (up to |T | = 30, |s| > 35)
DL reasoning is the most time consuming process!Large number of potential semantic links.Critical complexity of DL Difference.
Vs. State-of-the-art Approaches
Better fitness values for the optimal composition; 59/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Architecture
The Reference Architecture
Composition
and SelectionDiscoveryService
Reasoning
LevelFunctional
CompositionOptimization
ReasoningCausal Laws
Semantic
ServicesSemantic WebRepository of
Domain Ontology
End User’s Request
Services involvedin Composition
SrWs
Relevant ServicesSWs
S∗Ws
Impl:CPLEX
FoundNot
Contributions
Academics
Industry
sg Parsing
Impl:Fact++
Impl:Naive
CandidateCompositions
ScandidateWSC
Parsing
Services
Impl:jUDDI
Semantic Links
Causal Laws Axioms
Impl:WSML
Impl:GologFormalism
BPEL
RenderingImpl:Perl-based
Golog
Impl:BPEL4WS
Not Found Not Found
Impl:java,perl-based
Function
Objective
Constraints
End User’s
sg := 〈A, β〉Service Goal sg
Impl:JGAP-Lib
Details
60/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Applications
Industrial Scenarios in Use
Motivation, Orientation and Validation
Industrial settings (stateless Web services);
Industrial Transfer through Different Scenarios in
France Telecom AgIS;
European Project (FP6) SPICE;
Network of excellence (FP6) Knowledge Web.
61/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Applications
Industrial Scenarios in Use - An Example
Internet Package
Dynamic and automated configuration of Web services.35 Web services;ALE ontology (305 concepts, 117 properties).
ADSL elegibility
TV over IP
HDTV
Nowdays Solutions
Static/Predefined packages.ADSL Max+ + HDTV.
Open Issue
How to customize commercialoffers in a dynamic way?
The more offers the harder thecomposition task will be.
62/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Results
Experimentation
Main Results for Composition (Scenarios-Dependence!)
AI planning is more time consuming than DL reasoning.
The optimization process takes a negligible time.
Best Practices for using our Approach
Process ParametersComputation Time in ms
(0, 1000] (1000, 2000] (2000, 5000] (5000, 10000]
Semantic LinksNb services 69 74 78 83
orientedNb Inputs,
4 4 4 4Outputs
CompositionNb Services 220 260 350 450
OptimizationNb Candidate
100 100 100 100semantic Link
63/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Conclusions
Contributions1 Analysis of Requirements
Automation, Expressivity, Applicability, Composability,Optmization.
2 SME3-Comp (SeMantic wEb sErvicE) Software:(Robust) Semantic Link, SLM;Automated Composition approaches;Composition Optimization;
3 Achievement in practical and Industrial scenarios;
Lessons LearntExp_Time Problem!Composition’s Complexity Criteria:
Web Service Input/Output Expressivity, Cardinality;Ontology Expressivity .
Composition of thousand of services is not yet a reality.64/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Perspectives
Future Work1 Adding Semantics on Links;2 Investigating in Expressiveness of Web Services;3 Exploring Expressiveness of Composition Constructs;4 Improving Quality of Composition:
Coupling Quality of Service and Semantic Links;Coupling Composition and Discovery.
5 Investigating in further Scenarios, Benchmarks (SWS Challenge).
65/ 66
Introduction Composability Composition Robustness Quality Evaluation Conclusions
Perspectives
Future Work1 Adding Semantics on Links;2 Investigating in Expressiveness of Web Services;3 Exploring Expressiveness of Composition Constructs;4 Improving Quality of Composition:
Coupling Quality of Service and Semantic Links;Coupling Composition and Discovery.
5 Investigating in further Scenarios, Benchmarks (SWS Challenge).
Thanks for your attention!Freddy Lécué
freddy.lecue@manchester.ac.uk
66/ 66
top related