Reasoning about Events on the Semantic Web

Reasoning about Events on the Semantic Web
Ontology Summit 2014 – Track A
Megan Katsumi and Michael Gr¨
uninger
Semantic Technologies Laboratory
Department of Mechanical and Industrial Engineering
University of Toronto
January 23, 2014
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
1 / 15
Introduction
Questions for Ontology Summit 2014
How are Semantic Web / Big Data applications using ontologies?
Which ontologies are being used?
What ontologies are required by these applications?
If ontologies meeting these requirements exist but are not being used,
what are the reasons?
If ontologies meeting these requirements do not exist, how can they
be designed?
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
2 / 15
Motivation
The Role of Reasoning
Currently, most event ontologies on the Semantic Web are employed for
annotation and retrieval
What kinds of reasoning is being done with these event ontologies?
With all of the annotated data, there are many opportunities for
reasoning that are not being addressed.
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
3 / 15
Motivation
Some Motivating Scenarios
Beyond annotation and retrieval:
Emergency response: Are two reported incidents possibly related?
Context-awareness: Are there any potential delays in my area/path of
interest? (construction, accidents, special events...)
Municipal planning: During which time periods are no events scheduled at
the location of interest?
Recreational events: Are there any scheduling conflicts between my events
of interest at Festival-X?
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
4 / 15
Objective
Research Questions
All of these observations raise the questions:
Are Semantic Web ontologies able to support non-trivial reasoning
problems?
If not, why?
Are the existing ontologies simply not designed with enough
semantics to support these applications, or have they reached the
limit of what Semantic Web languages can support?
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
5 / 15
Approach
Semantic Web Event Ontologies
Using the reasoning abilities of PSL as a reference point, our investigation
is focused on a few well known Semantic Web event ontologies:
SEM Core
http://semanticweb.cs.vu.nl/2009/11/sem/
LODE
http://linkedevents.org/ontology/
The Event Ontology
http://purl.org/NET/c4dm/event.owl
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
6 / 15
Approach
Modules of the PSL Ontology
Tactocc
6
Tcomplex
6
Tdisc
6
Tatomic
@
I
@
..
....
..
.
@
.
.
..
.
...
.....
....
...
...
..
....
Tsubactivity
I
@
@
@
@
Tduration
6
...
...
@
. Tocctree
...
....
....
..
....
@
@
..
....
@
Megan Katsumi (MIE)
state
...
....
@......
Tpsl
University of Toronto
core
January 23, 2014
7 / 15
Approach
Process Specification Language
PSL (ISO 18629) is a modular, extensible ontology capturing
concepts required for process specification
There are 300 concepts across 50 extensions of a common core theory
(PSL-Core), each with a set of first-order axioms written using the
Common Logic Interchange Format (ISO 24707).
colore.oor .net/process specification language/
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
8 / 15
Approach
Extending Ontologies
From PSL, SEM, the Event Ontology, and LODE, we created the following
ontologies:
psl.owl (OWL axiomatization of PSL)
sem-x.owl (non-conservative extension of SEM using psl.owl)
event-x.owl (non-conservative extension of EVENT using psl.owl)
lode-x.owl (non-conservative extension of LODE using psl.owl)
This provides us with a spectrum of ontologies in OWL and First-Order
Logic to be evaluated by a set of competency questions derived from
motivating scenarios of potential reasoning applications on the Semantic
Web.
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
9 / 15
Approach
Extending Ontologies
From PSL, SEM, the Event Ontology, and LODE, we created the following
ontologies:
psl.swrl (SWRL axiomatization of PSL)
sem-x.swrl (non-conservative extension of SEM using psl.swrl)
event-x.swrl (non-conservative extension of EVENT using psl.swrl)
lode-x.swrl (non-conservative extension of LODE using psl.swrl)
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
10 / 15
Approach
Relationships among Event Ontology Extensions
psl.clif
psl.swrl
psl.owl
Megan Katsumi (MIE)
sem-r.swrl
event-r.swrl
lode-r.swrl
sem-x.owl
event-x.owl
lode-x.owl
sem.owl
event.owl
lode.owl
University of Toronto
January 23, 2014
11 / 15
Approach
Relationships among Event Ontology Extensions
Using the relationships between PSL and the event ontologies, we can
entail mappings among the event ontologies, and characterize what
semantics are shared across these ontologies.
We also want to use the competency questions to distinguish among
the different extensions of the same ontology – what do we get from
the additional axioms? What role does the ontology language play?
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
12 / 15
Approach
Evaluation of Ontology Extensions
Attempt to
formalize the query
in the ontology’s
signature
Definable?
No
(query) language
restriction?
Yes
Yes
Attempt query
Case 3: retry with
“normalized” (FOL)
version to verify
Successfully
answered?
Yes
No
Case 2: missing
concepts (i.e. insufficient
breadth); done
Case1: done
No
Safety check:
make sure there
are no counterexamples
Counterexample?
Yes
Axioms cannot entail query
b/c it’s inconsistent with the
theory
Resolve issue (error in
query or ontology) and retry)
No
Missing
axioms
Re-try with
extended
version
Successfully
answered?
No
Determine what axioms
(OR: what part of the
domain theory) are
missing
Yes
Done:
Original ontology:
Case 5;
Extension: Case 1
Are the missing
axioms definable in the
current language?
No
Case 4:Translate ontology
to a more expressive
language (e.g. FOL) and
retry to verify
Yes
Case 5: Ontology (and its
identified extension) cannot
satisfy the requirements –
insufficient scope.
Add required axioms and retry
to verify
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
13 / 15
Approach
Possible Outcomes
Successfully
answers
query-X?
Case 1:
Yes
No
Semantics
(axioms)
missing?
Query not
definable?
Case 2: Missing
concept(s)
Megan Katsumi (MIE)
Case 3: Language/
reasoner
expressivity
restrictions
University of Toronto
Case 4: Language
restriction (axioms
not definable, or
necessary parts of
the domain theory
are not definable)
Case 5:
Insufficient
scope (depth)
January 23, 2014
14 / 15
Discussion
Discussion
The development of reasoning applications may serve to promote more use
of ontologies on the web.
We hope that the outcome of this work will also provide guidance on what
is required if we want to be able to perform non-trivial reasoning with
ontologies on the Semantic Web.
Preliminary results indicate that the axiomatizations of the event
ontologies on the Semantic Web are too weak to specify and entail
competency questions extracted from the motivating scenarios.
Identify techniques to prove that we have the maximal subtheory of a
Common Logic ontology in a given ontology language (such as OWL
and SWRL).
Megan Katsumi (MIE)
University of Toronto
January 23, 2014
15 / 15

Download Report