SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf ·...

41
SEMANTIC CONFLICT DETECTION IN META-DATA A RULE-BASED APPROACH KARTHIKEYAN GIRILOGANATHAN ADVISOR: Dr. I. BUDAK ARPINAR COMMITTEE: Dr. AMIT P. SHETH Dr. KHALED M. RASHEED

Transcript of SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf ·...

Page 1: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

SEM

AN

TIC

CON

FLIC

T D

ETE

CTIO

N IN

ME

TA-D

ATA

A

RU

LE-B

ASE

D A

PPRO

ACH

KA

RTH

IKE

YAN

GIR

ILO

GA

NA

THA

N

AD

VIS

OR

: D

r. I.

BU

DA

K A

RPI

NA

RC

OM

MIT

TEE

: Dr.

AM

IT P

. SH

ETH

Dr.

KH

ALE

D M

. RA

SHE

ED

Page 2: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Intro

duct

ion

Mas

sive

am

ount

of d

ata

is a

vaila

ble

on th

e W

ebA

bilit

y to

ann

otat

e, e

xtra

ct, a

nd q

uery

sem

antic

met

a-da

ta h

as

incr

ease

d:S

WE

TO (S

eman

tic W

eb T

echn

olog

y E

valu

atio

n O

ntol

ogy)

:po

pula

ted

with

ove

r 800

,000

ent

ities

and

1.5

mill

ion

expl

icit

rela

tions

hips

bet

wee

n th

em in

RD

F or

OW

LFr

eedo

m (S

emag

ix):

uses

SW

ETO

and

oth

er d

omai

n on

tolo

gies

to s

eman

tical

ly

anno

tate

mill

ions

of d

ocum

ents

or W

eb p

ages

Web

Fou

ntai

n (IB

M):

anno

tate

d an

d di

sam

bigu

ated

dat

a fro

m o

ver a

billi

on

docu

men

ts

Intr

oduc

tion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 3: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Evo

lutio

n of

Met

a-D

ata

[She

th20

03]

Intr

oduc

tion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 4: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Nex

t gen

erat

ion

tool

s w

ill fo

cus

on a

ctio

nabl

e in

form

atio

n (w

ith a

ssoc

iate

d so

urce

s an

d su

ppor

ting

evid

ence

) fro

m

exis

ting

(met

a-)d

ata

Con

cern

s ab

out u

sage

of m

eta-

data

Hig

h qu

ality

(i.e

., re

liabl

e, a

ccur

ate,

and

trus

twor

thy)

sem

antic

m

eta-

data

Ent

ity d

isam

bigu

atio

nIn

cons

iste

ncy

chec

king

in O

WL

Con

flict

det

ectio

n

Met

a-D

ata

Conc

erns

Intr

oduc

tion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 5: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Mot

ivat

ing

Fact

ors

“Rep

rese

ntin

g, id

enti

fyin

g, d

isco

veri

ng, v

alid

atin

g, a

nd

expl

oiti

ng c

ompl

ex r

elat

ions

hips

are

impo

rtan

t is

sues

rel

ated

to

rea

lizin

g th

e fu

ll po

wer

of t

he S

eman

tic

Web

, and

can

hel

p cl

ose

the

gap

betw

een

high

ly s

epar

ated

info

rmat

ion

retr

ieva

lan

d de

cisi

on-m

akin

gst

eps”

[She

th, A

rpin

ar&

Kash

yap

2003

]

“The

Web

is d

ecen

tral

ized

, allo

wing

any

one

to s

ay a

nyth

ing.

As

a re

sult

, dif

fere

nt v

iewp

oint

sm

ay b

e co

ntra

dict

ory,

or

even

fa

lse

info

rmat

ion

may

be

prov

ided

. In

orde

r to

pre

vent

age

nts

from

com

bini

ng in

com

pati

ble

data

or

from

tak

ing

cons

iste

nt

data

and

evo

lvin

g it

into

an

inco

nsis

tent

sta

te, i

t is

impo

rtan

tth

at in

cons

iste

ncie

sca

n be

det

ecte

d au

tom

atic

ally

”[W

3C

2004

]

“…th

ese

prob

lem

s m

anif

est

them

selv

es in

var

ious

way

s,

incl

udin

g po

or r

ecal

l of

avai

labl

e re

sour

ces

and

inco

nsis

tenc

yof

sea

rch

resu

lts.

The

y ar

ise

due

to e

rror

s, o

mis

sion

san

d am

bigu

itie

sin

the

met

adat

a…”[

Curr

ier

& Ba

rton

200

3]

Intr

oduc

tion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 6: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

John

Cla

ura

Bill

Mar

y

fath

erO

fm

arrie

dTo

mot

herO

f

fath

erO

f

mar

riedT

o

fath

erin

Law

Of

CO

NFL

ICT

Sem

antic

Con

flict

Iden

tific

atio

n

Intr

oduc

tion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 7: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Conf

lict i

llust

ratio

n th

roug

h Si

mpl

ifica

tion

CO

NFL

ICT

Intr

oduc

tion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 8: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Met

aDat

aS

tore

Var

ious

alg

orith

ms

for s

eman

tic a

naly

tics

Que

ry in

terfa

ce

Mot

ivat

ing

Scen

ario

EX

TRA

CTO

RS

(F

reed

om)

Sem

DIS

SW

ETO

Intr

oduc

tion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 9: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Out

line

Con

flict

type

s an

d de

finiti

ons

Sim

plifi

catio

n pr

oces

sS

yste

m a

rchi

tect

ure

Exp

erim

enta

l res

ults

Con

clus

ion

and

futu

re w

ork

Intr

oduc

tion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 10: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

‘dam

l:una

mbi

guou

s’ o

r ‘o

wl:i

nver

seFu

nctio

nal-

Pro

perty

’ vio

latio

n

Prop

erty

Ass

ertio

n Co

nflic

ts

‘dam

l:uni

que’

or

‘ow

l:Fun

ctio

nal-P

rope

rty’

viol

atio

n

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 11: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Prop

erty

Ass

ertio

n Co

nflic

ts

‘asy

mm

etric

’ pro

perty

vi

olat

ion

‘dis

join

t’ pr

oper

ty

viol

atio

n

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 12: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Clas

s Ass

ertio

n Co

nflic

ts

Cla

sses

c1

and

c2 a

re

‘dam

l:dis

join

t’ or

‘o

wl:d

isjo

int’

Cla

sses

‘Citi

zen’

and

‘Im

mig

rant

’ are

‘d

aml:d

isjo

int’

or

‘ow

l:dis

join

t’

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 13: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Clas

s Ass

ertio

n Co

nflic

ts

Cla

ss ‘E

mpl

oyee

’ has

a

OW

L or

DA

ML

rest

rictio

n ‘m

axC

ardi

nalit

y’ o

f ‘1’

on

a re

latio

n ‘h

asD

esig

natio

n’

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 14: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Stat

emen

t Ass

ertio

n Co

nflic

ts

We

wan

t to

say

that

a

pers

on c

anno

t be

a su

perio

r and

a fr

iend

to

“Joh

n” a

t the

sam

e tim

e.

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 15: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Non

-Ass

ertio

nalC

onfli

cts

Eith

er th

e su

bjec

t or t

he o

bjec

t alo

ne is

diff

eren

t bet

wee

n tw

o R

DF

tripl

es.

Sub

ject

ive

Con

flict

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 16: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Con

stra

ints

sup

plie

d by

an

expe

rt (e

.g.,

pers

on(x

) ca

n ne

ver

do

actio

n(y)

)E

Con

stra

ints

ex

pres

sed

in

an

onto

logy

(e

.g.,

the

prop

erty

‘b

iolo

gica

lMot

her’

is u

niqu

e)U

The

resu

lt of

sim

plifi

catio

n(S

(T)

s)s

A fu

nctio

n de

notin

g th

e pr

oces

s of

sim

plifi

catio

nS

A s

et o

f trip

les

T

Two

sets

of t

riple

s T1

and

T2 a

re sa

id to

be

in c

onfli

ctif

thei

r sim

plifi

catio

nsS(

T1)

s1 a

nd S

(T2)

s2 a

re m

utua

lly

non-

agre

eabl

e.

Conf

lict D

efin

ition

s

Two

sim

plifi

catio

nss1

and

s2 a

re m

utua

lly n

on-a

gree

able

if ta

ken

toge

ther

they

are

in v

iola

tion

of U

or E

.

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 17: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Will

iam

sC

hris

Rep

ublic

anPa

rtyvo

tedF

orm

embe

rOf

supp

orte

rOf

Sim

plifi

catio

n Ty

pes

An

RD

F tri

ple

is tr

ivia

lly a

sim

plifi

catio

nbe

caus

e it

is

the

mos

t bas

ic p

iece

of k

now

ledg

eC

ompo

sitio

n of

rela

tions

lead

s to

sim

plifi

catio

n

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 18: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Com

posit

ion

of R

elatio

nsC

onsi

der a

set o

f Trip

les T

,Le

t E

-se

t of e

ntiti

es

P-

set o

f rel

atio

ns

Then

, E=

{e1,

e2.

.. en

}P

= {p

1, p

2...

pm}.

Le

tC

-se

t of o

rder

ed re

latio

n tu

ples

that

can

be

com

pose

dto

a si

ngle

rela

tion

R-

set o

f rel

atio

ns o

btai

ned

by su

bstit

utin

g th

e co

mpo

sed

rela

tion

for t

he c

ompo

sabl

ere

latio

ns.

Then

, C=

{(p1

, pk)

, ...,

(pa,

pb,

pc,

...)}

. R

={r

1, r2

... rn

}, w

here

r1, r

2...

rnar

e re

sults

of

the

com

posi

tion.

The

tripl

e (

The

tripl

e ( e

ieirkrk

ejej) i

s a

) is

a si

mpl

ifica

tion

sim

plifi

catio

nif if

rkrk∈∈

R a

nd

R a

nd e

i,ej

ei,e

j ∈∈E

.E

.

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 19: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Stat

emen

t Sim

plifi

catio

n

Ther

e co

uld

be b

ackg

roun

d kn

owle

dge

base

d si

mpl

ifica

tions

of t

he fo

rm:

stat

emen

t 1∧

stat

emen

t 2∧

… st

atem

ent n→

sta

tem

ent t

In th

is c

ase

stat

emen

t tis

a s

impl

ifica

tion.

This

type

of s

impl

ifica

tion

will

dep

end

on e

xper

t kn

owle

dge.

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 20: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Stat

emen

t Sim

plifi

catio

n

Imm

igra

ntIm

mig

rant

Fina

ncia

lOrg

aniz

atio

n

Judi

cial

Org

aniz

atio

n

Bus

ines

sOrg

aniz

atio

n

Per

son

mul

tiple

Dep

osits

asso

ciat

ed

owne

r

wor

ks

unde

rInve

stig

atio

n

Mon

eyLa

unde

ring

susp

ecte

d

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 21: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Def

inin

g St

atem

ent S

impl

ifica

tion

Rules

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 22: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Rule

ML

(Ove

rvie

w)

Exp

lore

rule

sys

tem

s su

itabl

e fo

r the

Web

The

synt

ax (i

n X

ML

and

RD

F fo

rm)

Sem

antic

sTr

acta

bilit

y/ef

ficie

ncy

Tran

sfor

mat

ion

Com

pila

tion

Ena

ble

infe

renc

ing

on W

eb d

ata

& in

terc

hang

e of

rule

s be

twee

n in

telli

gent

sys

tem

s (o

ntol

ogy

inte

grat

ion

etc.

)W

e us

e R

uleM

LA

ny in

fere

nce

engi

ne th

at u

nder

stan

ds R

uleM

Lca

n ev

alua

te o

ur ru

les.

W

e do

not

nee

d to

thin

k ab

out r

epre

sent

atio

n an

d tra

nsla

tion.

http

://w

ww

.rule

ml.o

rg/

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 23: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Our

Rep

rese

ntat

ion

of a

n RD

F tri

ple

<sub

ject

><pr

oper

ty><

obje

ct>

Sta

tem

ent(x

)S

ubje

ct(x

,sub

ject

)P

rope

rty(x

,pro

perty

)O

bjec

t(x,o

bjec

t)

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 24: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

ifst

atem

ent(

x) a

nd s

tate

men

t(y)

and

su

bjec

t(x,

a) a

nd r

elat

ion(

x,re

l1) a

nd

obje

ct(x

,b) a

nd s

ubje

ct(y

,a) a

nd

rela

tion

(y,r

el2)

and

obj

ect(

y,b)

and

disj

oint

(rel

1,re

l2)

then

conf

lict(

x,y)

Conf

lict R

ules

Can

be

clas

sifie

d as

Inte

grity

Con

stra

int R

ules

.

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 25: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Sim

plifi

catio

n Ru

les

Can

be

clas

sifie

d as

Pro

duct

ion

Rul

es

ifst

atem

ent(

x) a

nd s

tate

men

t(y)

and

subj

ect(

x,a)

and

rel

atio

n(x,

rel1)

and

obje

ct(x

,b) a

nd s

ubje

ct(y

,a) a

nd

rela

tion

(y,r

el2)

and

obj

ect(

y,b)

then

newS

tate

men

t(a,

rel

3, b

)

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 26: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Relat

ions

hip

Ont

olog

y

Rel

atio

ns a

re a

t the

hea

rt of

sem

antic

Web

[S

heth

, Arp

inar

& K

ashy

ap20

03]

Rel

atio

ns a

mon

g re

latio

ns n

eed

to b

e sp

ecifi

edH

iera

rchy

of r

elat

ions

is s

imila

r to

a ta

xono

my

Just

as

we

have

mov

ed fr

om ta

xono

my

to

onto

logy

the

idea

is to

hav

e an

ont

olog

y fo

r re

latio

ns a

lso.

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 27: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Relat

ions

hip

Ont

olog

y

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 28: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Edi

ting/

Popu

latin

g th

e Re

latio

nshi

p O

ntol

ogy

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 29: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Tem

plat

e Ba

sed

Rule

Tra

nsfo

rmat

ion

if st

atem

ent(x

) and

stat

emen

t(y) a

ndsu

bjec

t(x,a

) and

rela

tion(

x,re

l1)a

nd o

bjec

t(x,b

) and

su

bjec

t(y,a

) and

rela

tion(

y,re

l2)a

nd o

bjec

t(y,b

) and

di

sjoi

nt(r

el1,

rel2

)the

n co

nflic

t(x,y

)

disj

oint

(http

://fo

o.co

m/te

st#l

ikes

, ht

tp://

foo.

com

/test

#hat

es)

Tran

sfor

m

if st

atem

ent(x

) and

stat

emen

t(y) a

nd

subj

ect(x

,a) a

nd re

latio

n(x,

http

://fo

o.co

m/te

st#l

ikes

)and

obj

ect(x

,b) a

ndsu

bjec

t(y,a

) and

rela

tion(

y, h

ttp://

foo.

com

/test

#hat

es)a

nd o

bjec

t(y,b

) and

disj

oint

(http

://fo

o.co

m/te

st#l

ikes

, http

://fo

o.co

m/te

st#h

ates

)the

n co

nflic

t(x,y

).

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 30: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Sem

antic

Met

adat

a

Ont

olog

y

JEN

A A

PI

SE

RIA

LIZE

R

User Interface

Rel

atio

nshi

p O

ntol

ogy

RULE

SR

uleM

LS

IMPL

IFIC

ATI

ON CO

NFI

DE

R A

PI

MA

ND

AR

AX

AP

I

Fact

sRU

LES

Rul

eML

MA

ND

AR

AX

AP

IC

ON

FLIC

T E

NG

INE

Syst

em A

rchi

tect

ure

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 31: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Alg

orith

m

Sem

antic

Met

adat

a

Ont

olog

y

Find

sta

tem

ents

that

cor

resp

ond

to c

onst

rain

tsC

onve

rt th

ose

stat

emen

ts to

con

flict

rule

sC

onve

rt th

e re

st o

f the

sta

tem

ents

to fa

cts

Rel

atio

nshi

p O

ntol

ogy

Find

rela

tions

that

cor

resp

ond

to c

onst

rain

tsG

ener

ate

Conf

lict r

ules

Find

rela

tions

that

cor

resp

ond

to C

ompo

stiti

onG

ener

ate

Sim

plifi

cati

on ru

les

Con

fider

Ser

ializ

e th

e st

atem

ents

to tr

iple

s an

d as

sign

idto

eac

h tri

ple

Con

vert

tripl

es to

FO

L pr

edic

ates

Sim

plifi

catio

n R

ules

?

Whi

le n

ew s

tate

men

ts p

rodu

ced

App

ly s

impl

ifica

tion

rule

s to

get

new

stm

tsA

ssig

n in

tern

al id

s to

thes

e st

mts

Atta

ch th

e de

rivat

ion

to e

ach

stm

t

yes

CO

NFL

ICT

EN

GIN

E

NO

FAC

TS

Con

flict

Rul

es

List

of S

tate

men

t pai

rs th

at a

re in

con

flict

Der

ivat

ion

avai

labl

e fo

r con

flict

ing

stat

emen

tsD

eriv

atio

n av

aila

ble

if st

mts

gene

rate

d in

tern

ally

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 32: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Man

dara

xA

ope

n so

urce

java

cla

ss li

brar

y fo

r ded

uctio

n ru

les

OO N

ot a

tran

slat

ion

of a

pro

log

inte

rpre

ter f

rom

c to

java

Bas

ed o

n ba

ckw

ard

reas

onin

gE

asy

inte

grat

ion

of v

ario

us d

atab

ases

Sup

port

for W

eb s

ervi

ces,

and

EJB

Rul

es s

peci

fied

as R

uleM

LJe

ns D

ietri

ch (M

asse

y U

nive

rsity

, New

Zea

land

)A

list

of c

ontri

buto

rs a

vaila

ble

at h

ttp://

man

dara

x.so

urce

forg

e.ne

t/

ww

w.m

anda

rax.

org

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 33: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Kno

wled

geba

se w

ith F

acts

and

Rul

es

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 34: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Conf

lict I

dent

ifica

tion

Resu

lts

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 35: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Stat

emen

t Pro

vena

nce

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 36: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Perf

orm

ance

Eva

luat

ion

(1)

with

incr

ease

in n

umbe

r of c

onfli

cts (

500

tripl

es)

Co

nfl

icts

vs

Tim

e

6.03

6466

803

6.11

7612

961

6.15

1044

237

6.19

3546

166.

2061

1288

6

6.27

7261

64

6

6.056.1

6.156.2

6.256.3

010

2030

4050

60

No

of

Co

nfl

icts

Log ( time in milliseconds)

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 37: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Perf

orm

ance

Eva

luat

ion

(2)

with

incr

ease

in n

umbe

r of t

riple

s (10

con

flict

s)Tr

iple

s vs

Tim

e4.651849588

5.955244091

6.508925185

6.730027336

6.808724663

6.875508877

6.89743981

7.030853263

7.075019174

7.049989218

012345678

020

040

060

080

010

0012

00

No

of T

riple

s

Log ( Time in milliseconds)

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 38: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Conc

lusio

n

In th

is w

ork

we

have

:D

efin

ed c

onfli

cts

in s

eman

tic m

eta-

data

and

cl

assi

fied

them

.D

iscu

ssed

a ru

le-b

ased

app

roac

h to

iden

tify

the

conf

licts

.S

how

n th

e us

e of

rela

tions

bet

wee

n re

latio

ns to

si

mpl

ify th

e tri

ples

and

iden

tify

conf

licts

.D

emon

stra

ted

the

appl

icab

ility

of t

he a

ppro

ach

over

a li

mite

d da

ta s

et u

sing

a p

roto

type

.

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 39: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Futu

re W

ork

Our

futu

re w

ork

dire

ctio

ns in

clud

e de

velo

ping

:S

cala

ble

conf

lict i

dent

ifica

tion

tech

niqu

es fo

r lar

ge

amou

nts

of s

eman

tic d

ata

and

conf

lict r

ules

Inve

stig

atio

n of

oth

er ru

le e

valu

atio

n m

etho

ds to

im

prov

e pe

rform

ance

Exp

erim

ent w

ith w

ays

of re

pres

entin

g an

RD

F tri

ple

in

pred

icat

e fo

rm to

com

pare

per

form

ance

A m

echa

nism

for e

xpre

ssin

g, e

valu

atin

g, a

nd a

djus

ting

trust

dyn

amic

ally

bas

ed o

n co

nflic

t det

ectio

n

Intro

duct

ion

Con

flict

Typ

esSi

mpl

ifica

tion

Arc

hite

ctur

eR

esul

tsC

oncl

usio

n

Page 40: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Que

stio

ns?

Page 41: SEMANTIC CONFLICT DETECTION IN META-DATAcobweb.cs.uga.edu/~budak/thesis/gk_talk.pdf · 2010-03-14 · Introduction Massive amount of data is available on the Web Ability to annotate,

Than

k Y

ou