christopher wellen m.sc. candidate mcgill university on cognition and computation: an introduction...

28
Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Upload: julianna-ruth-greer

Post on 03-Jan-2016

217 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Christopher WellenM.Sc. CandidateMcGill University

On Cognition and Computation:

An Introduction to Spatial Ontologies

Page 2: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Agenda● Ontologies and GIScience

● Historical Antecedents

● Computational Application areas:

– Interoperability

– Federated Databases● Cognitive Research

– Landscape Categories

– Geography as Science● Critique

● Emergent Debates

Page 3: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

What exactly is Ontology?

● Classic answer: “explicit specification of a conceptualization.” (Gruber, 1993)

● Set of entities, attributes, axioms and relations. ● Field encompasses a large part of GIS research● Geography has no single agreed-upon shared

ontology

Page 4: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

An Ontology of Streams

H. Pundt, Y. Bishr / Computers & Geosciences 28 (2002) 95–102

Page 5: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Suggested Upper Merged Ontology (SUMO)

Phytilia, 2002

Page 6: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Top Level vs. Domain Ontologies

● Domain Ontologies: task, specialty, discipline– Bottom-up

● Top Level Ontologies: reality– Top-Down

Page 7: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Cognitive vs. Computational

Page 8: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Why Ontologies?

● Better design of information systems:– Data sharing and data discovery– Data interoperability and machine readability

● More unified geography:– Conceptualize geography at top level

Page 9: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Ontology Development

● Two General strategies for development:– top down– bottom up

● Web-based interfaces: Ontolingua● Graphical User Interfaces: ConceptVISTA

Page 10: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Ontology Implementation

● Software Components● Database components● Data collection policies

Page 11: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies
Page 12: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Great Britain Historical GIS – Ontology of Administrative

Boundaries

Contains Dates of Creation/Abolition

GeometryStats

Contains Relationship Classes

Page 13: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Historical Antecedents

● Philosophy– Aristotle– Smith and Mark (1998) – geographical domain does

not divide into bounded concepts neatly● Information Science/Artificial Intelligence

– Gruber (1993) – re-usable software components● Semantic Web

– World Wide Web Consortium (W3C) wants to make web machine-readable

Page 14: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Computational Applications

● Interoperability - FieldLog● Federated Databases - GeoNIS

Page 15: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Interoperability - FieldLog (Broderic 2004)

• Scientists like to invent their own language

• Government database managers like everyone to use the same language

• Major conflict in Geological Survey of Canada, which has both types of people

• Solution: FieldLog

Page 16: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Top level classes Cartographic

Geospatial MetadataGeologic

Symbolization Coordinates

Geologic concepts and analysis

References and documentation

Station

Rock

Sample

Analysis

100 50 -87

Drill Hole

ID Lat Lon

A01 51 -86

Site

ID Lat Lon

A01 1 granite

LithologySite Rock Type

A01 1 A

SampleSite Rock Sample

A01 1 A

GeochemSite Rock Sample

User defined User defined classesclasses

User dataUser data

Page 17: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

FieldLog

• User profiles can be created and edited

• Maps can be generated quickly and with minimal complaining (hopefully...)

Page 18: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Federated Databases

DB2

Middleware

DB1

User

Page 19: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Federated Databases

Page 20: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Federated Databases - GeoNIS

● Each user describes their data in terms of a common ontology

● Requests for others' data can be phrased in terms of one's own ontology

● Translators can translate the request for data into others' ontologies as well as their data into the ontology of the user who requested the data

Page 21: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Federated Databases - GeoNIS

● ‘Top-Level’ ontology: Open Geospatial Consortium – data model standard for geospatial data

● Users use relations: synonym (same thing), hypernym (superclass), hyponym (subclass), Topology

● Topology: arc-node, route, NodeRoute, point-event

Page 22: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

(Stoimenov and Djordjevic-Kajan (Stoimenov and Djordjevic-Kajan 2005)2005)

Page 23: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Cognitive Work

● How do people concieve of/categorize space/spatial features?

● Landscape Categories:– Do mountains exist?– Yindjibarndi (Mark and Turk, 2003)

Page 24: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies
Page 25: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies
Page 26: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Critiques

● Little work that synthesizes whole spectrum – mostly cognitive or computational

● Used as a buzzword to give research more exposure● Formal ontologies seen as a technical exercise, little

work done to explore cultural/power differences.● Ontologies seen as too complicated (Schuurmann)● What about fuzzy boundaries?

Page 27: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

Emergent debates

● Top level ontology? What would it look like? Can everyone agree on it?

● Should Ontology development proceed from a top-down or bottom up manner?

● Can Spatial Ontologies be produced independently of temporal ones?

Page 28: Christopher Wellen M.Sc. Candidate McGill University On Cognition and Computation: An Introduction to Spatial Ontologies

What is not quite Ontology?

● Semantics– Methods of representing knowledge (from harvey’s

data sharing paper)– Does not describe relations between features or what

they actually are● Taxonomy

– Classification is not conception. ● Folksonomy

– Flickr photos, messy taxonomy