1. Web Science & Technologies
University of Koblenz ▪ Landau, Germany
Validation of User Intentions in
Process Models
Gerd Gröner, Mohsen Asadi, Bardia Mohabatti, Dragan Gasevic,
Fernando Silva Parreiras and Marko Boskovic
2. What is the goal of a particular process?
Goal that should be achieved
several goals
subgoals
dependencies between
goals
?
Process Model:
➔
Operational representation
of activities to achieve a
certain goal
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 2
3. Example: Traveling to Gdansk for CAiSE
quick
journey
✔
comfortable
journey
↯ ✔
cheap
late arrival
journey
goals and dependencies activities and dependencies
among goals between them
➔
Dependencies / relationships might be contradicting!
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 3
4. Problems and Questions
How are goals represented?
Which kinds of
quick
dependencies
journey
are covered in a
process model?
comfortable
journey
late arrival
cheap
journey
How is the influence of a
mapped goal on an activity? How to map / align between goals and
activities?
(What is the meaning?)
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 4
5. Idea
1. Extract and represent relationships of both models
logical view:
quick ➔
similar
journey
relationships
between
elements in both
comfortable
journey
models
late arrival
cheap
journey
2. Explicitly represent mappings between goals and activities
3. Classify inconsistencies between mapped goals and activities
4. Formalization for validation and recognition of inconsistencies
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 5
6. Outline
1. Modeling dimensions
i. Goal models
ii. Process models
2. Realization inconsistencies
3. Modeling principles
4. Validation
5. Discussion and conclusion
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 6
7. Requirements representation
Requirements: goals, functions and constraints of a system
→ Representation: means for understanding user intentions and
how they are related to each other
G5 G
OR OR
Goal models: +
•
Graph with G1 G4
AND
intentional elements (hard AND
goals, soft goals, tasks) G2 G3
links (contributions)
decompositions (AND, IOR,
XOR)
➔
requirements of a
system-to-be
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 7
8. Intentions in Process Models
Requirement perspective: Process model:
• Goals (user intentions) • Control flow perspective
• Relationships (constraints, - activities
dependencies) - ordering through different constructors
preferred
payment
AND AND
no additional 30 days term
fee of credit
mapping
≈
realization of a goal
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 8
9. … Problem
Requirement perspective: Control flow perspective:
preferred
payment
AND AND
no additional 30 days term
fee of credit
mapping
relations not necessarily coincide,
they might even be contradicting
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 9
10. Outline
1. Modeling dimensions
i. Goal models
ii. Process models
2. Realization inconsistencies
3. Modeling principles
4. Validation
5. Discussion and conclusion
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 10
11. Realization Inconsistencies
Intentional relations (IR) Control flow relations /
over intentional elements workflow relations WF over
G1, …,Gm ∈ G activities
A1, …,An ∈ A
WF over A1, …,An ∈ A and IR over G1, …,Gm ∈ G, with target
goal G ∈ G and activities A1, …,An are realizations of G1, …,Gm
A strong inconsistency between WF and IR occurs if there
is no execution combination of activities that leads to the
fulfillment of the target goal G.
A potential inconsistency between WF and IR occurs if
some execution combinations of activities lead to the
fulfillment of G and some do not lead to the fulfillment of G.
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 11
12. Realization Inconsistencies – Example
Strong inconsistency:
preferred
payment
AND AND
no additional
30 days term fee
of credit
Potential inconsistency:
preferred
payment
AND AND
30 days term
little of credit
payment
effort
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 12
14. Outline
1. Modeling dimensions
i. Goal models
ii. User intentions for process models
2. Realization inconsistencies
3. Modeling principles
4. Validation
5. Discussion and conclusion
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 14
15. Description Logics (DLs)
(1) DL Knowledge base (KB)
Concepts C C(x)
Property (role) R R(x,y)
Subclass C⊑D ∀ x (C(x) → D(x))
Negation ¬C ¬ C(x)
Union C⊔D C(x) ∨ D(x)
Intersection C⊓D C(x) ∧ D(x)
Existential
Quantification ∃ P.C ∃ y (P(x,y)∧C(y))
(2) Inference service:
Subsumption: C⊑D?
if KB ⊨ C ⊑ D
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 15
16. Towards a common knowledge base
(atomic) concepts:
- goals
- activities A1
G1
A2
G2
A3
G3
complex concept expressions
- intentional relations of Gi
- control flow relations of Aj
connect concepts (atomic concepts)
of mapped entities
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 16
19. Outline
1. Modeling dimensions
i. Goal models
ii. User intentions for process models
2. Realization inconsistencies
3. Modeling principles
4. Validation
5. Discussion and conclusion
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 19
20. How to detect inconsistencies?
G1 A1
G2 A2
A3
G4
G3
atomic WF by
IR by concept complex
complex equivalence concept
concept
➔
complex concept expressions: logical formulas
➔
comparison of concept expressions
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 20
21. Validation – slightly abstracted
Logical point of view:
G1 M A1
RelG RelA
Mi
Validation principle:
• compare RelG and RelA if there is a mapping between G and A
RelG RelA
?
- coincide / equivalent ?
- contradicting?
- no influence?
• both models are correct on their own
• only mapped elements need to be considered
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 21
22. Relation comparison
Strong inconsistency
''...no execution combination fulfills the target goal …''
➔ KB: no common instance: RelG ⊓ RelA ⊑ ⊥
Potential inconsistency
''… there are some execution combinations that do not lead to the
fulfillment of G...''
➔
an execution combination not necessarily fulfills G
➔ KB: entailment RelA ⊑ RelG does not hold: ¬(RelA ⊑ RelG)
Otherwise
Relations coincide
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 22
23. Relation comparison – concept level
RelG RelA
?
compare RelG and RelA
✔ RelA ⊑ RelG ± ¬(RelA ⊑ RelG) ↯ RelG ⊓ RelA ⊑ ⊥
✔ relations ± depends on ↯ relations contradict
coincide particular execution
➔
detect inconsistency between both models
➔
identify the source (i.e., activity and goal) of an inconsistency
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 23
25. Conclusion
2 views /
perspectives
goals / intentions
activities and
their execution
DTC
XOR mapping
XOR
CCR CRC (realization
of goals)
intentional control flow
relationships relationships
➔
Problem: goals and activities depend on other goals and activities
➔
Mapping imposes to an activity also relationship from its corresponding goal
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 25
26. Conclusion (2)
Approach
1) Specifying realization inconsistencies
2) Formalizing relationships of both models
3) Detection of inconsistencies of mapped goals and activities
Contribution of DLs
Detection of inconsistencies
Potential inconsistency
Strong inconsistency
Pinpointing of sources for inconsistencies
Future Work:
focus on behavioral constraints
(semi-) automatic derivation of process models
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 26
27. Thank you!
WeST Gerd Gröner CAiSE 2012
groener@uni-koblenz.de 27