Tailoring Temporal Description Logics for Reasoning over Temporal Conceptual Models A. Artale 1 R. Kontchakov 2 , V. Ryzhikov 1 , and M. Zakharyaschev 2 1 KRDB Research Centre, Free University of Bozen-Bolzano 2 Department of Comp. Science and Inf. Sys., Birkbeck College, London University of KwaZulu-Natal, Durban, South Africa, 30-09-11 Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
44
Embed
Tailoring Temporal Description Logics for Reasoning over Temporal Conceptual Models
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Tailoring Temporal Description Logics forReasoning over Temporal Conceptual Models
A. Artale1
R. Kontchakov2, V. Ryzhikov1, and M. Zakharyaschev2
1 KRDB Research Centre, Free University of Bozen-Bolzano2 Department of Comp. Science and Inf. Sys., Birkbeck College, London
University of KwaZulu-Natal, Durban, South Africa, 30-09-11
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Motivations
Investigation of the Computational Complexity of reasoningover Temporal Ontologies/Conceptual Models.
Languages considered: Family of Temporally ExtendedDL-Lite languages.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
ERVTThe Temporal Data Model
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
ERVT: The Proposed Temporal Conceptual Model
ERVT is the temporal extended Entity-Relationship model able tocapture Validity Time with the following temporal features:
Timestamping: to distinguish between temporal andatemporal modeling constructs.
Evolution and Transition constraints: to describe how objectscan change their class membership over time. Transitionconstraints presuppose that class migration happens in a fixedamount of time.
Lifespan cardinality constraints: temporal counterparts ofstandard cardinality constraints evaluated over the entireexistence of the object.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
ERVT: The Proposed Temporal Conceptual Model
ERVT is the temporal extended Entity-Relationship model able tocapture Validity Time with the following temporal features:
Timestamping: to distinguish between temporal andatemporal modeling constructs.
Evolution and Transition constraints: to describe how objectscan change their class membership over time. Transitionconstraints presuppose that class migration happens in a fixedamount of time.
Lifespan cardinality constraints: temporal counterparts ofstandard cardinality constraints evaluated over the entireexistence of the object.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
ERVT: The Proposed Temporal Conceptual Model
ERVT is the temporal extended Entity-Relationship model able tocapture Validity Time with the following temporal features:
Timestamping: to distinguish between temporal andatemporal modeling constructs.
Evolution and Transition constraints: to describe how objectscan change their class membership over time. Transitionconstraints presuppose that class migration happens in a fixedamount of time.
Lifespan cardinality constraints: temporal counterparts ofstandard cardinality constraints evaluated over the entireexistence of the object.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
ERVT: The Proposed Temporal Conceptual Model
ERVT is the temporal extended Entity-Relationship model able tocapture Validity Time with the following temporal features:
Timestamping: to distinguish between temporal andatemporal modeling constructs.
Evolution and Transition constraints: to describe how objectscan change their class membership over time. Transitionconstraints presuppose that class migration happens in a fixedamount of time.
Lifespan cardinality constraints: temporal counterparts ofstandard cardinality constraints evaluated over the entireexistence of the object.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
ERVT: A Company Example
Department S InterestGroup
OrganizationalUnit
d
Member S
(1,∞)
org
mbrEmployee S
Name(String)
S
PaySlipNumber(Integer)
S Salary(Integer)
T
Manager T
TopManagerAreaManager
dex−
dev
pex
WorksOn T
(3,∞)
act
emp
Project
ProjectCode(String)
SPropose gp
(0,1)
Ex-Project tex
Managesman
(1,1)
[0,5]
prj
(1,1)
1
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Known Complexity Results for Reasoning over ERVT
Undecidability.
Theorem. Reasoning on the ERVT fragment with bothtimestamping and evolution constraints isundecidable [ :AMAI-05].
Decidability.
Theorem. Reasoning on the ERVT fragment with bothtimestamping and evolution constraints restricted to classes isExpTime [ FWZ:02].Theorem. Reasoning on the ERVT fragment withtimestamping and lifespan cardinalities is 2ExpTime [ LT:07].
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Known Complexity Results for Reasoning over ERVT
Undecidability.
Theorem. Reasoning on the ERVT fragment with bothtimestamping and evolution constraints isundecidable [ :AMAI-05].
Decidability.
Theorem. Reasoning on the ERVT fragment with bothtimestamping and evolution constraints restricted to classes isExpTime [ FWZ:02].Theorem. Reasoning on the ERVT fragment withtimestamping and lifespan cardinalities is 2ExpTime [ LT:07].
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Aims of this Work
Our Aims: Conduct an exhaustive investigation on usefulfragments of ERVT weakening either the atemporal ortemporal component.
Our Results:
We give an exhaustive picture on the complexity of reasoningover temporal extensions of DL-Lite;Based on these results, we show encouraging complexityresults for reasoning over temporal ontologies where practicalreasoning is feasible!
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Aims of this Work
Our Aims: Conduct an exhaustive investigation on usefulfragments of ERVT weakening either the atemporal ortemporal component.
Our Results:
We give an exhaustive picture on the complexity of reasoningover temporal extensions of DL-Lite;Based on these results, we show encouraging complexityresults for reasoning over temporal ontologies where practicalreasoning is feasible!
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
The DL-Lite Languages
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
DL-LiteNbool, DL-LiteNkrom and DL-LiteNcore
DL-LiteNbool. C1 v C2, with:
R −→ P | P−
B −→ A | ≥ n R | ⊥C −→ B | ¬C | C1 u C2
DL-LiteNkrom. B1 v B2, B1 u B2 v ⊥, ¬B1 v B2.
DL-LiteNcore. B1 v B2, B1 u B2 v ⊥.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
The DL-Lite Languages - Complexity Results
Complexity Results [CDLLR:AAAI05, CKZ:JAIR09]:
Satisfiability: NP-complete/NLogSpace/NLogSpace;
Instance Checking (Data Complexity): AC0/AC0/AC0;
Query Answering (Data Complexity): coNP/coNP/AC0.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
DL-Lite – Conceptual Modelling Example
Manager v EmployeeAreaManager v ManagerTopManager v Manager
AreaManager u TopManager v ⊥Manager v AreaManager t TopManager∃WorksFor v Employee∃WorksFor− v Project
Project v ∃WorksFor−
≥ 2 Manages v ⊥≥ 2 Manages− v ⊥
...
Note: We use the shortcut ∃R instead of ≥ 1 R.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
The Family of EER/UML Languages [ CKRZ:ER07]
ERrefDL-LiteNcore
ERboolDL-LiteNkrom
ERfullDL-LiteNbool
ConstructDL-Lite
RepresentationEntities Concept Name: E
+ + + Isa E1 v E2
+ + + Disjointness E1 v ¬E2
– + + Covering E ≡ E1 t E2
Attributes Role Name: A
+ + + Range ∃A− v D
+ + + Multiplicity E v≥ nAE v≤ mA
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
The Family of EER/UML Languages [ CKRZ:ER07]
ERrefDL-LiteNcore
ERboolDL-LiteNkrom
ERfullDL-LiteNbool
ConstructDL-Lite
Representation
RelationshipsConcept Name CR
and n Roles Ui
+ + + TypingCR ≡ ∃Ui
≥ 2 Ui v ⊥∃U−i v Ei
+ + +Cardinality
(Refinement)E v≥ n U−iE v≤ m U−i
– – + Isa —
– – + Disjointness —
– – + Covering —
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
The Temporal DL-LiteLanguages
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Seminal Papers
SCHILD, K., 1993. Combining terminological logics withtense logic. Proc. of the 6th Portuguese Conference on AI.
F. Baader and A. Laux., 1995. Terminological Logics withModal Operators, IJCAI-95.
Wolter, F. and Zakharyaschev, M., 1998, TemporalizingDescription Logics, FroCoS-98.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Complexity for Temporal ALC– Known Results
Temporal operators can be applied to:
Concepts, roles or axioms (they are temporalized);
Concepts, roles or axioms can have a time-invariantinterpretation (they are rigid).
The satisfiability problem has a different complexity dependingfrom the combination between LT L and ALC constructs:
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Complexity: TFPXDL-LiteNbool and Fragments
1 We reduce satisfiability in TFPXDL-LiteNbool KBs tosatisfiability in QT L1, i.e., the one-variable fragment offirst-order temporal logic over (Z, <).
2 We then show how to remove existential quantifiers from suchQT L1 formulas, thus reducing to LT L formulas.
3 Complexity results for temporal extensions of DL-LiteNboolfollow from the corresponding LT L results.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Complexity: TFPXDL-LiteNkrom/core and Fragments
1 We reduce satisfiability in TFPXDL-LiteNkrom/core KBs tosatisfiability in QT L1, i.e., the one-variable fragment offirst-order temporal logic over (Z, <).
2 We then show how to remove existential quantifiers from suchQT L1 formulas, thus reducing to two fragments of LT L, i.e.,propositional temporal logic of binary clauses, i.e., LT Lkromand LT Lcore.
3 We show that:
LT Lkrom is NP-complete;LT Lcore is in PTime.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Conclusions
We showed that:
By dropping ISA between relations (ERbool/DL-LiteNbool) andcovering (ERref/DL-LiteNcore) we obtained bettercomputational behavior for reasoning over temporalschemas/DL-Lite ontologies.
Both ERbool and ERref have been extended with timestamping,evolution and transition constraints, lifespan cardinalities.
DL-LiteNbool, DL-LiteNkrom and DL-LiteNcore have been extendedwith past and future temporal operators, and with theuniversal modality (both over concepts and roles).
We presented a nearly complete picture for reasoning overtemporal CMs/Ontologies/DL-Lite KBs.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Future Work
Few cases involving lifespan cardinalities/universal modalityare still open.
Investigating the problem of temporal queries over temporalontologies/conceptual schemas.
Investigating the possibility to use standard and implementedtemporal reasoners for practical reasoning over temporalschemas.
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
Thanks to...
Enrico Franconi
Carsten Lutz
Christine Parent
Stefano Spaccapietra
David Toman
Frank Wolter
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs
THANK YOU!
Alessandro Artale – KRDB Tailoring Temporal DLs for Reasoning over Temporal CMs