context sensitivity for networked ontologies igor mozetič, marko grobelnik, damjan bojadžijev...

28
Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

Upload: pauline-west

Post on 17-Jan-2018

218 views

Category:

Documents


0 download

DESCRIPTION

NeOn, Rome, 21 Mar 2006 JSI 3 Overview Formalizing context Cyc Semantic Web C-OWL Probabilistic approaches JSI related work

TRANSCRIPT

Page 1: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

Context sensitivity for networked ontologies

Igor Mozetič, Marko Grobelnik, Damjan BojadžijevJozef Stefan InstituteSlovenia

Page 2: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 2

text context

explicit implicit

global local

text

I’m here who, where

text expl(context)

text_1context_1

+

Page 3: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 3

Overview

Formalizing context Cyc Semantic Web C-OWL Probabilistic approaches JSI related work

Page 4: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 4

McCarthy [1993]

AI: modelling of context and use in automated reasoning

implicit -> explicit ist( context, proposition ) context = collection of assumptions

(generalization of, partially known) entering and exiting,

nesting, lifting, transcending, …

Page 5: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 5

Cyc [Lenat, Guha]

Cyc KB = set of microtheories (Mt) Microtheory = set of axioms

shared assumptions, topic internally consistent localized (more efficient) reasoning preconditions = context in which Mt is applicable

Page 6: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 6

Cyc (example)

ist( NaiveStateChangeMt, isa( ?X, Freezing ) & outputsCreated( ?X, ?Obj ) => stateOfMatter( ?Obj, SolidStateMatter ))

NaiveStateChangeMt domainAssumptions: forAll ?U isa( ?U, StateOfMatterChangeEvent ) => isa( ?U, CreationOrDestructionEvent )

Page 7: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 7

Context for Semantic Web [Guha et al]

AI SWscope, complexity of phenomena

scale (comp. complexity),distributed sources,ease of use

Aggregation from different sources. Issues:• class differences• property type differences• point of view• implicit time• approximations

Page 8: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 8

C-OWL [Giunchiglia et al]:Contextualizing ontologies

Ontologies Contexts

Global, shared modelEncode common viewCombining by import

Global semantics

Local modelsEncode each party’s viewCombining by explicit mappingsLocal Models Semantics

Page 9: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 9

OWL: Global semantics for multiple (networked) ontologies

sharedmodel

Page 10: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 10

OWL: Global semantics for multiple (networked) ontologies

sharedmodel

import

Page 11: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 11

C-OWL: Local model semantics

local models

Page 12: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 12

C-OWL: Mappings

contextualizedontology

context

context

Page 13: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 13

C-OWL ontology is a pair:

OWL ontology (target): concepts individuals roles

mappings (bridge rules): equivalence onto into compatible incompatible

Page 14: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 14

C-OWL example

OWL ontology (target) + mappings (bridge rules)

Page 15: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 15

C-OWL of any use?

Import ontology vs. define context mappings? (diversity as defect vs. feature)

Semantic Web = Web of Semantic links ? (context mappings)

Discovering context mappings = core issue in building Semantic Web ?

Page 16: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 16

JSI related work

Parametric temporal ontology Simultaneous ontologies User profiling Implicit document context (links)

Page 17: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 17

Temporal ontology

Temporal algebra [Allen]: event = temporal interval relations: before, meets, starts, finishes, …

week(now)

day(sun)day(mon) meets meets

starts finishes

Page 18: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 18

Temporal ontology

Temporal algebra [Allen]: event = temporal interval relations: before, meets, starts, finishes, …

week(now) week(now+1)

day(sun)day(mon) meets meets

starts finishes

meets

Page 19: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 19

Temporal ontology

Temporal algebra [Allen]: event = temporal interval relations: before, meets, starts, finishes, …

week(now) week(now+1)

day(sun)day(mon) meets meets

starts finishes

day(now)day(now-1) meets

meets

day(now+1)meets

Page 20: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 20

Temporal reasoning

Temporal algebra [Allen]: event = temporal interval relations: before, meets, starts, finishes, …

week(now) week(now+1)

day(sun)day(mon) meets meets

starts finishes

day(now)day(now-1) meets

meets

equals

day(now+1)meets

?

Page 21: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 21

Temporal reasoning

Temporal algebra [Allen]: event = temporal interval relations: before, meets, starts, finishes, …

week(now) week(now+1)

day(sun)day(mon) meets meets

starts finishes

day(now)day(now-1) meets

meets

equals

day(now+1)meets

day(mon)

starts

Page 22: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 22

Temporal reasoning

Temporal algebra [Allen]: event = temporal interval relations: before, meets, starts, finishes, …

week(now) week(now+1)

day(sun)day(mon) meets meets

starts finishes

day(now)day(now-1) meets

meets

equals

day(now+1)meets

meets day(mon)

starts

Page 23: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 23

Temporal reasoning

Temporal algebra [Allen]: event = temporal interval relations: before, meets, starts, finishes, …

week(now) week(now+1)

day(sun)day(mon) meets meets

starts finishes

day(now)day(now-1) meets

meets

equals

day(now+1)meets

meets day(mon)

starts

equals

Page 24: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 24

Parameterized temporal ontology Parameters:

now order of magnitude past - future

now-1 now now+1 now+2

day

wee

k m

onth

ye

ar

deca

de

now = ?

context

Page 25: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 25

News analysis

earthquake tsunami tsunami tsunami tsunami

News stream:

the same? yet another one?

Page 26: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 26

A temporal model: Tsunami

Earthquake

Tsunami

Search & Rescue Rebuilding

~minutes

~hours

~days ~months

E T S&R

25.dec 26.dec 27.dec 28.dec 29.dec 30.dec 31.dec 1.jan 2.jan 3.jan

Page 27: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 27

News analysis: Temporal model = Context

earthquake tsunami tsunami tsunami tsunami

News stream:

model of tsunami

provides contextfor subsequentevents

Page 28: Context sensitivity for networked ontologies Igor Mozetič, Marko Grobelnik, Damjan Bojadžijev Jozef Stefan Institute Slovenia

NeOn, Rome, 21 Mar 2006

JSI 28

Summary

Parameterized ontology Context determines parameters

when? how long? order of magnitude

Temporal model selected by events provides context