Analyzing Theme, Space and Time: An Ontology-based Approach Matthew Perry, Farshad Hakimpour, Amit Sheth 14 th International Symposium on Advances in Geographic.

Analyzing Theme, Space and Time: An Ontology-based

ApproachMatthew Perry, Farshad Hakimpour, Amit Sheth

14th International Symposium on Advances in Geographic Information Systems, Arlington, VA, Nov. 10 – 11, 2006

Acknowledgement: NSF-ITR-IDM Award #0325464 ‘SemDIS: Discovering Complex Relationships in the Semantic Web’

Outline

• Background• Motivation• Goals and Assumptions• Modeling Approach• Query Operators• Conclusions

Background

• What is an ontology?– Agreed-upon formalization (or conceptual model) of

concepts and relationships in the real world

• Types of ontologies?– General-purpose vs. Domain ontologies

• Parts of an ontology– Classes – types or logical groups of objects– Relationships – how objects relate to each other– Attributes –features and characteristics of objects– Instances – members of Classes who have Attributes and

participate in Relationships

Background (ontologies and RDF/S)

Schema e.g. Student attends University

Data e.g. ‘Matt’ attends ‘University of Georgia’

Representing ontologies and instance data• W3C standards

– Resource Description Framework (RDF)• Language for representing information about

resources• Resources are identified by Uniform Resource

Identifiers (URIs) – globally-unique• Common framework for expressing information

allows exchange and reuse without loss of meaning

rdfs:Class

lsdis:Politician

lsdis:Speech

rdf:Propertyrdfs:Literal

lsdis:gives

lsdis:name

rdfs:domain

rdfs:domainrdfs:range

rdfs:range

lsdis:Politician_123

lsdis:Speech_456

“Franklin Roosevelt”

lsdis:gives

name

lsdis:Person

rdf:typerdfs:subClassOfstatement

Defining Properties (domain and range):<lsdis:gives> <rdfs:domain> <lsdis:Politician> .<lsdis:gives> <rdfs:range> <lsdis:Politician> . Subject Predicate Object

Statement (triple):<lsdis:Politician_123> <lsdis:gives> <lsdis:Speech_456> . Subject Predicate Object

Defining Class/Property Hierarchies:<lsdis:Politician> <rdfs:subClassOf> <lsdis:Person> . Subject Predicate Object

Labeled Graph

Statement (triple):<lsdis:Politician_123> <lsdis:name> “Franklin Roosevelt” . Subject Predicate ObjectDefining Classes:

<lsdis:Person> <rdf:type> <rdfs:Class> . Subject Predicate Object

Defining Properties:<lsdis:gives> <rdf:type> <rdf:Property> . Subject Predicate Object

Motivation

From …..

Finding things

To …..

Finding out about things

Relationships!

Semantic Discovery*

* http://lsdis.cs.uga.edu/projects/semdis

From thematic analytics to spatio-temporal, thematic (STT) analytics (Ex: Bioterrorism)

E5:Terrorist

E6:Attack

E4:Chemical

E10:Doctor

E9:Base

E8:Soldier

E7:Platoon

E1:Soldier

E3:Disease

E2:Symptom

E14:Battle

E12:Platoon

E13:Soldier

E11:Location

assigned_to

participated_in

member_of

carried_out

spotted_at

stationed_at

member_of

sign_ofparticipated_in

causes

member_of

used_in

exhibits

Near in Space

After the Battle [4, 6]

[8, 10]

[0, 10]

[3, 5]

[0, 2]

[0, 2]

Spotted Before and Close in Time

Goals and Assumptions

Goals and Contributions of this work• Define a Domain-independent Ontology which

integrates Spatial and Thematic Knowledge– Allows exploiting the flexibility and extensibility of

Semantic Web data models– Can deal with incompleteness of information on the web

• Incorporate temporal metadata into this model• Identify and formalize basic spatial and temporal

relationship-based query operators which complement current thematic operators of SemDis

Assumptions / Design Decisions• Interested in Relationship Analysis• Do not address issues of scale and

granularity at this time• Exact serialization/representation of

spatial geometry is not a primary concern– RDF XML Literal Type (GML)– Ontology mirroring GML/OGC specification

Our Approach

Upper-level Ontology modeling Theme and Space

OccurrentContinuant

Named_PlaceSpatial_OccurrentDynamic_Entity

Spatial_Region

Occurrent: Events – happen and then don’t existContinuant: Concrete and Abstract Entities – persist over timeNamed_Place: Those entities with static spatial behavior (e.g. building)Dynamic_Entity: Those entities with dynamic spatial behavior (e.g. person)Spatial_Occurrent: Events with concrete spatial locations (e.g. a speech)Spatial_Region: Records exact spatial location (geometry objects,

coordinate system info)

occurred_at located_at

occurred_at: Links Spatial_Occurents to their geographic locationslocated_at: Links Named_Places to their geographic locations

rdfs:subClassOfpropertyFinal Classification of Domain

Classes depends upon the intendedapplication

OccurrentContinuant

Named_Place

Spatial_OccurrentDynamic_Entity

PersonCity

Politician

Soldier

Military_Unit

BattleVehicle

Bombing

Speech

Military_Eventassigned_to

on_crew_of

used_in

gives

participates_in

trains_at

Spatial_Region

located_at occurred_at

Upper-level Ontology

Domain Ontology

rdfs:subClassOf used for integrationrdfs:subClassOfrelationship type

Incorporation of Temporal Information• Use Temporal RDF Graphs defined by

Gutiérrez, et al1

• Focuses on absolute time and considers time as a discrete, linearly-ordered domain

• Associate time intervals with statements which represent the valid-time of the statement– Essentially a quad instead of a triple

1. Claudio Gutiérrez, Carlos A. Hurtado, Alejandro A. Vaisman: Temporal RDF. ESWC 2005: 93-107

Example Temporal Graph: Platoon Membership

E1:Soldier

E3:Platoon E5:Soldier

E2:Platoon

E4:Soldier

assigned_to [1, 10]

assigned_to [11, 20]

assigned_to [5, 15]

assigned_to [5, 15]

Querying in the STT dimensions• Path Query in the thematic dimension

– Thematic Context

• Associate spatial region with a path• Associate temporal interval with a path• Query operators based on properties and

relationships between associated spatial regions and temporal intervals

Thematic Context• Specifies a type of connection between resources in the thematic

dimension of our ontology

‘John Smith’

‘B24#123’‘Bombing#456

’

on_crew_of used_in

PersonMilitary_Vehicl

eBombing

on_crew_of used_in

Schema

Person.on_crew_of.Military_Vehicle.used_in.BombingPath Template

‘John Smith’.on_crew_of.Military_Vehicle.used_in.Bombing

ei INSTANCES(G) and ei = Ci, ei rdf:type Ci, or ei rdf:type Ci’ and Ci’ rdfs:subClassOf Ci

pi PROPERTIES(G) and pi = Pi or pi rdfs:subPropertyOf Pi

Ci CLASSES(G) INSTANCES(G)Pi PROPERTIES(G)

tc = C1.P1.C2.P2.C3…Cn-1.Pn-1.Cn

pt = e1.p1.e2.p2.e3…en-1.pn-1.en

Thematic Context

Thematic Context Instance

Example: ‘John Smith’.on_crew_of.Military_Vehicle.used_in.Bombing

Example: ‘John Smith’.on_crew_of.‘B-24#123’.used_in.‘Bombing#456’

Thematic Context

Thematic Query (ρ-theme)ρ-theme (G, tc) {pt}

Example: find all Bombing events connected to ‘John Smith’ through a vehicle participation context

ρ-theme (G, ‘John Smith’.on_crew_of.Military_Vehicle. used_in.Bombing)

‘John Smith’.on_crew_of.‘B-24#123’.used_in.‘Bombing#456’‘John Smith’.on_crew_of.‘B-24#123’.used_in.‘Bombing#789’

Result

G = temporal RDF Graph, tc = thematic context, pt = thematic context instance

E1:Soldier

E2:Soldier

E3:Soldier

E5:Battle

E4:Address

E6:Address

E7:Battle

occurred_at

occurred_at

located_at located_at

lives_at

lives_at

assigned_to

E8:Military_UnitE8:Military_Unit

assigned_to

participates_in

participates_in

Georeferenced Coordinate Space

(Spatial Regions)

Dynamic EntitiesSpatial OccurrentsNamed Places

Thematic Contexts Linking Non-Spatial Entities to Spatial Entities

ResidencyBattle Participation

E1:Soldier

Deriving Spatial and Temporal Properties of Entities

ρ-spatial_extent (G, {pt}) {(pt, sr)}

Example: Where were the battles in which the ‘101st Airborne Division’ fought?

ρ-spatial_extent (G, ρ-theme (G, ‘101st Airborne Division’. .participates_in.Battle))

‘101st Airborne Division’.participates_in. ‘Operation Market Garden’, ‘Geom#123’

Result

G = temporal RDF Graph, pt = thematic context instance, sr = spatial region

Temporal Properties of Context Instances

Platoon#456 Soldier#789Soldier#123

assigned_to:[3, 12] assigned_to:[6, 20]

Temporal Intersection [6, 12]

Temporal Range [3, 20]

Deriving Spatial and Temporal Properties of Entities

ρ-temporal_intersect ({pt}) {(pt, [t1, t2])}

Example: Which Soldiers were members of the ‘1st Armored Division’ at the same time?

ρ-temporal_intersect (ρ-theme (G, Soldier.assigned_to.‘1st Armored Division’.assigned_to.Soldier))

‘Fred Smith’.assigned_to.’1st Armored Division’.assigned_to. ‘Bill Jones’, [1941:04:15, 1943:02:25]

Result

pt = thematic context instance, [t1, t2] = temporal interval

Querying Relationships Between Entities

(Thematic Context Instance tp, Temporal Interval [ti, tj], Spatial Region sr)

3 Spatial Relationship Operatorsρ-spatial_locateρ-spatial_evalρ-spatial_find

3 Temporal Relationship Operatorsρ-temporal_restrict

ρ-temporal_evalρ-temporal_find

Identify 6 major Spatiotemporal Relationship Querieswhich can be answered by combining previously defined operators

Spatial Relationships Between Entities• Examples:

– Which Military Units have spatial extents which are within 20 miles of (48.45° N, 44.30° E) in the context of Battle participation?

– Which infantry unit’s operational area overlaps the operational area of the 3rd Armored Division?

Quantitative Relationships Qualitative Relationships

Metric Spatial Function msf : S X S [distance]

Spatial Relationship OperatorsSpatial Predicate sp : S X S B

Qualitative Spatial Function qsf : S X S B [overlaps, meets, inside, …]

Metric Spatial Expression <mse> ::= <msf> <comp>

<comp> ::=

Spatial Predicate <sp> ::= <mse> | <qsf> | <sp> <sp> | <sp> <sp>

Spatial Relationship Operators

ρ-spatial_locate ({(pt, sr)}, sr’, sp) {pt}

ρ-spatial_eval ({(pt, sr)}, {(pt’, sr’)}, sp) {(pt, pt’)}

ρ-spatial_find ({(pt, sr)}, {(pt’, sr’)}) {(pt, pt’, qualitative spatial relationship)}

pt = thematic context instance, sr = spatial region, sp = spatial predicate

Example Spatial Relationship Query

Example: Which Military Units’ operational area overlaps the operational area of the ‘3rd Armored Division’?

S1 ρ-spatial_extent (G, ρ-theme (G, ‘3rd Armored Division’. participates_in.Military_Event))

S2 ρ-spatial_extent (G, ρ-theme (G, Military_Unit. participates_in.Military_Event))

ANS ρ-spatial_eval (S1, S2, overlaps(S1, S2))

Temporal Relationships Between Entities• Examples:

– Which Speeches by President Roosevelt were given within one day of a major battle?

– Who were members of the 101st Airborne during November 1944?

Quantitative Relationships Qualitative Relationships

Temporal Relationship Operators

Temporal Predicate tp : I X I B

Temporal Predicate <tp> ::= <mte> | <qtf> | <tp> <tp> | <tp> <tp>

Metric Temporal Expression <mte> ::= <mtf> <comp>

<comp> ::=

Metric Temporal Function mtf : I X I [elapsed time]

Qualitative Temporal Function qtf : I X I B [before, meets, during, …]

Temporal Relationship Operatorsρ-temporal_restrict ({(pt, [ti, tj])}, [tk, tl], tp) {pt}

ρ-temporal_eval ({(pt, [ti, tj])}, {(pt’, [tk, tl])}, tp) {(pt, pt’)}

ρ-temporal_find ({(pt, [ti, tj])}, {(pt’, [tk, tl])}) {union of convex intervals relationship2}

pt = thematic context instance, [tk, tl] = temporal interval, tp = temporal predicate

2. Ladkin, P.B. The logic of time representation, Ph.D. Dissertation, University of California at Berkley, Berkley, CA, 1987

I1 ρ-temporal_intersect (ρ-theme (G, ‘Franklin Roosevelt’. gives.Speech))

I2 ρ-temporal_intersect (ρ-theme (G, Military_Unit. participates_in.Battle))

ANS ρ-temporal_eval (I1, I2, elapsed time (I1, I2) 1 day)

Example Temporal Relationship Query

Example: Which Speeches by President Roosevelt were given within one day of a major Battle?

S1 ρ-spatial_extent (G, ρ-theme (G, ‘101st Airborne Division’. participates_in.Battle))

S2 ρ-spatial_extent (G, ρ-theme (G, ‘1st Armored Division’. participates_in.Battle))

ANS ρ-temporal_intersect (ρ-spatial_eval (S1, S2, distance (S1, S2) 10 miles)

Spatiotemporal Relationship QueriesExample: When did the 101st Airborne Division come within 10 miles of the 1st Armored Division in the context of Battle participation

Conclusions• Defined Basic Domain-independent ontology

integrating Theme and Space and showed how to incorporate temporal information with temporal RDF graphs

• Novelty centers on utilization of named relationships in the thematic dimension

• Link non-spatial entities with spatial entities at the thematic level in order to analyze their spatial properties

• Allows us to take advantage of the flexibility and extensibility of RDF and helps cope with incompleteness of information on the Web

More Information

• SemDis Project:– Google: SemDis

• LSDIS Lab:– Google: LSDIS