Ontology-based Data Matching and Applications

blabbingunequaledAI and Robotics

Oct 24, 2013 (3 years and 10 months ago)

90 views

25/10/2013

| pag.
1


Dr. Ioana Ciuciu: iciuciu@vub.ac.be

Ontology
-
based Data Matching

and Applications

Overview


Idea


Basic notions


Ontology
-
based Data Matching Framework
(ODMF)


ODMF Algorithms


ODMF Strategies


Applications


EU projects


Prolix


3D Anatomical Human


TAS
3


DIYSE (Do It Yourself Smart Experience)



25/10/2013

| pag.
2

Idea


…last week’s OIS course


Step5: Link your data with other data sources


Manually (e.g. FOAF profiles)


Automated linking algorithms (large data sets)



25/10/2013

| pag.
3

DBpedia

RDF Book Mashup

Harry
Potter and
the Half
Blood
Prince

Harry
Potter and
the Half
Blood
Prince

<
http://dbpedia.org/resource/Harry_Potter_and_the_Half
-
Blood_Prince
> owl:sameAs




<
http://www4.wiwiss.fu
-
berlin.de/bookmashup/books/0747581088
>

Common identifier
: ISBN: 0747581088

Idea


If no common identifier?


Use semantic information


2 resources with properties, relations betw. properties, etc.


Perform matching (semantic level)


Find similarity score

25/10/2013

| pag.
4

Resource 1

Resource 2

Matching


%


Similarity score

Basic notions



Ontology Matching (Alignment)





Data Matching





Ontology
-
based Data Matching

25/10/2013

| pag.
5

Ontology A

Ontology B

Resource 1

Resource 2

Matching


%


Ontology
-
based Data
matching Framework
(ODMF)


ODMF


9 algorithms and 7 strategies in total, among which
WordNet + multilingual terminography, OntoGraM,
GRASIM, LeMaSt, C
-
FOAM are the innovative
strategies


ODMF matching levels


String matching (e.g. SecondString library)


Lexical matching (WordNet based)


Graph matching (ontology based)


25/10/2013

| pag.
6

String Matching Algorithms


String similarity of two objects same super
-
ordinate concept


SecondString

library








+ new ODMF algorithms


ODMF.UnsmoothedJS

(same context, e.g. competency)


ODMF.JaroWrinklerTFIDF

(fuzzy matching)


“hearth”


“heart”

25/10/2013

| pag.
7

Competence 1:

“Obtain and test capillary blood samples”




Competence 2:

“Obtain and test specimens from individuals”

0.4 (MongeElkan)


0.9 (WinklerRescorer)


0.5 (UnsmoothedJS,


JaroWrinklerTFIDF,


TFIDF (Term Frequency Inverse
Document Frequency))


String Matching Algorithms

25/10/2013

| pag.
8

Advantages

Drawbacks

Easy to implement

No complex knowledge resource required

Objects are described in natural language

Different similarity score for different wording (e.g. with Jaro,
Jaro
-
Winkler)


Ex: “Ioana Ciuciu” ≠ “Ciuciu Ioana”

Little explanation of the result to the user

Lexical Matching Algorithms


Semantic similarity of two objects same super
-
ordinate concept


Object descriptions are
terminologically annotated

-

WordNet

25/10/2013

| pag.
9

Competence 1:

“Obtain and test capillary blood samples”

Set1 = {obtain, test, capillary blood sample}




Competence 2:

“Obtain and test specimens from individuals”

Set2 = {obtain, test, specimens, individuals}


Jaccard similarity coefficient
:



-

ODMF versions: Lexical1, Lexical2, Lexical3



2/5 = 0.4

WordNet


Lexical database of English (George A. Miller, Princeton, 1985)


Tool for computational linguistics & natural language processing


Synsets


distinct concepts










25/10/2013

| pag.
10

POS

Unique Strings

Synsets

Nouns

117798

82115

Verbs

11529

13767

Adjectives

21479

18156

Adverbs

4481

3621

Total

155287

117659

WordNet


Examples of semantic
& lexical relations between
synsets

(nouns)


Hypernym



hyponym (e.g. canine
-

dog)


Holonym



meronym

(e.g. car
-

wheel)



In ODMF, we use semantic relations defined in
Wordnet

for
matching two resources


Ex.: ”student” & “pupil” belong to the same
synset




25/10/2013

| pag.
11

Lexical Matching Algorithms

(Jaccard
-
based)


Lexical1


Does not make use of hypernym
-
hyponym relation


Lexical2


Takes the hypernym
-
hyponym relation between terms into account


Knowledge engineer needs to create the type hierarchy between terms


Lexical3


Takes the hypernym
-
hyponym relation between terms into account


Makes use of WordNet as upper ontology (concepts & relations)


Automatically converts information from WordNet to a Categorisation Framework


Knowledge engineer is assisted with possible concept suggestions

25/10/2013

| pag.
12

Example: Competence
description


Let a competence be by definition described as:


Competence description =
Persons Act

or Interact on
Objects

in



Manners

using
Instruments

at
Locations

at
Times


Then for


Ex: Competence: “
Obtain and test capillary blood samples




The annotation could look like


Person


Agent


Action


Obtain capillary blood sample


Test capillary blood sample


Object


Capillary blood sample



25/10/2013

| pag.
13

Example: possible sets of
terms


Lexical1



{
Agent, Obtain capillary blood sample, Test capillary blood sample, Capillary

blood
sample}



Lexical2 (Lexical1+
hypernyms
)



{
Person
, Agent
,
Action
, Obtain
capillary blood sample, Test capillary blood

sample
,
Object
, Capillary
blood sample}





25/10/2013

| pag.
14

Example: possible sets of
terms


Lexical3 (
WordNet
)



Annotation


blood
: the fluid (red in vertebrates) that is pumped by the
heart


capillary
: any of the minute blood vessels connecting arterioles with
venules


obtain
: come into possession
of


sample
: a small part of something intended as representative of the
whole


test1
: determine the presence or properties of (a substance
)


test2
: test or examine for the presence of disease or
infection



Set of terms


{blood, capillary, obtain, sample, test1, test2}


25/10/2013

| pag.
15

Lexical Matching Algorithms

25/10/2013

| pag.
16

Advantages

Drawbacks

The score is less dependent on the wording of the object
descriptions (because synonyms and translation equivalents
will be taken into account using the term base)


The two compared sets of terms produce evidence about the
score to the end user.


The terminological resource (used to annotate object
descriptions) should sufficiently cover
the different domains.

Graph Matching Algorithms


Recall: Every ODMF strategy contains at least one graph algorithm


Idea: using classification information of objects, the relations
between objects, and the properties of objects, to:


Calculate the similarity between two objects (e.g. two competences)


Find related objects for a given object (e.g. find relevant qualifications to improve
a competency)


Graph: semantic graph (terminological ontology)


Vertices: concepts


Edges: semantic relations between concepts


Bidirected (role
-

co
-
role)


2 ODMF graph matching algorithms


Ontology
-
based Graph Matching (OntoGraM)


Graph
-
Aided Similarity Calculation (GRASIM)


25/10/2013

| pag.
17


Ontology
-
based Graph
Matching (OntoGraM)



Two objects: their classification, properties & semantic relations


Domain ontology + application ontology


Rule
-
based reasoning: forward chaining to infer new knowledge,
based on the existing knowledge expressed in the knowledge base


Ex: competency
-
based HRM ontology


competency as reference

25/10/2013

| pag.
18

Person p

Competencies
p
= {c
1
,…c
n
}


Task t

Competencies
t
={c
1
,…,c
m
}


Compare p & t by comparing Competencies
p

& Competencies
t

Ontology
-
based Graph
Matching (OntoGraM)


Use
holonym
-
meronym

relation between competences (or tasks)


Ex: (HSC237
-

Health & Social
Care Standard)



“Obtain and test capillary blood samples”



“Obtain capillary blood”



“Test, record and report on capillary blood sample results ”



If a person has both sub
-
competences, then we can deduce that


he/she has also the competence “
Obtain and test capillary
blood


samples”


2
OntoGram

versions


OntoGram

version 1 (Graph1)


the one described above


OntoGram

version
2
(
Graph2)


Uses extra (fuzzy) relations such as “is slightly similar”, “
is moderately
similar”” or “is very similar”


More accurate results across domains/organizations provided that the relations are applied correctly

25/10/2013

| pag.
19

OntoGraM example


Get the similarity score between


Task = “Agree closure with customer”


Competence: “Process skills” with competence level: Good


Competence:
“Soft
skills” with competence level:
Good


Function = “Junior SAT”


Competence: “Administrative skills & other skills”, with competence level:
Good


Competence:
“Process
skills”, with competence level:
Good


Competence: “Soft skills” with competence level:
Good


Competence:
“Technical
skills” with competence level:
Good

Remark: every competence has a list of sub
-
competences, and so on


Ex: “Process skills” has sub
-

competences {“additional tasks”, “escalations”,

“information”, “products”, “supplier”}

Result:
sim

= 0.5

( acc. to Graph1; 2 competences overlap with the 4 required)



sim

= 0.7593

(acc. to Graph2; intersection: 41 competences; union: 54


competences; all sub
-
competences are taken into account)





25/10/2013

| pag.
20

OntoGraM

25/10/2013

| pag.
21

Advantages

Drawbacks

The similarity of two objects that belong to a different object
type can be calculated with a degree of accuracy comparable
to that
of a human expert.


Extensive evidence for the calculated score may be presented
to the end user (by reporting on the rules that were applied to
calculate the score).

The management of the application ontology requires a

considerable effort by the knowledge engineer
.

Graph
-
Aided

Similarity
Calculation (GRASIM)


Idea [11]:


Compute the shortest path between two sub
-
graphs of a given graph
(Dijkstra, other)


Use Semantic Decision Tables (SDT) to freely & correctly adjust graph
algorithms (assign weights)


Transfer the shortest path into a similarity score



25/10/2013

| pag.
22


25/10/2013

| pag.
23

GRASIM

-

the algorithm

25/10/2013

| pag.
23

Label graph arcs
using SDT

calculate shortest
paths (SP)

Select a shortest
path algorithm

Dijkstra’s
shortest path
algorithm

calculate
Similarity (S)

Tang, Y. et al. (2010):
Towards Freely and Correctly Adjusted Dijkstra's Algorithm with Semantic Decision
Tables for Ontology Based Data Matching
, in Proc.
fo

the 2nd International Conference on Computer and
Automation Engineering "ICCAE 2010, ISBN: 978
-
1
-
4244
-
5586
-
7, Singapore, February 26
-

28, 2010


ODMF Strategies


Combination and composition of Algorithms


Strategy One: Lexon Matching Strategy (LeMaSt)


Strategy Two: Controlled Fully Automated Ontology
Based matching Strategy (C
-
FOAM)

25/10/2013

| pag.
24

Lexon Matching Strategy
(LeMaSt)


LeMaSt


A string matching algorithm


A lexical matching algorithm


Lexon matching


Semantic similarity of the objects descriptions (2 lexon sets)


Object descriptions annotated with lexons

25/10/2013

| pag.
25

Ex: Annotation

Competence :


“Obtain and test capillary blood samples”

Annotated with lexon:


< “Obtain and test capillary blood samples” , “agent”, “obtains”, “obtained by”,

“capillary blood sample”>

Interpreted as:


For the competence “Obtain and test capillary blood samples”,



“an agent obtains a capillary blood sample” &



“a capillary blood sample is obtained by an agent”



Lexon Matching Strategy
(LeMaSt)


Extension of
Jaccard

similarity coefficient:








When S
is calculated, we run
OntoGraM

to calculate the final
matching score


use lexical relations (e.g.
holonym
-
meronym
) between
competences


use semantic constraints (e.g.
cardinaliry

constraint)






25/10/2013

| pag.
26

C


contribution score (calculated with Jaccard similarity scores)

x
i



contribution score of each lexon


x
i
= 1
,

if same head, role, co
-
role & tail


x
i
= 0.5
,

if same head & tail but different role


& co
-
role



x
i
= 0.5
,

if same head or tail and same role &


co
-
role


x
i

= 0.25
,

if same head or tail



S


average score of the lexons

Extended Jaccard similarity
coefficient


pseudo code

FUNCTION FLOAT calculate
-
Score (
Ontology
-
DB)
{

LS1 = Load
lexons

from Ontology
-
DB;

LS2 = Load
Lexons

from Ontology
-
DB;

Union = 0;

WHILE (LS1. Has
Next

()) {

Lexon

lexon1 = (
Lexon
) LS1.next() ;

Lexon

lexon2 =
null

;

Union += 1 ;

Lexon

lexon2MaxOverlap =
null

;

DmaxOverlap

= 0.0 ;

WHILE (LS2. Has
Next

()) {

lexon2 = (
Lexon
) LS2.next
();

IF
(lexon1.Head.equals (lexon2.Head))

AND
(lexon1.Tail.equals (lexon2.Tail))

AND
(lexon1.Role.equals (lexon2.Role))

AND
(lexon1.CoRole.equals (lexon2.CoRole)) {

dMaxOverlap

=
1.0
; //
same

lexon

lexon2MaxOverlap = lexon2;

lexon2MaxOverlap.MaxOverlap =
dMaxOverlap
;

break;

}

}

25/10/2013

| pag.
27

Extended Jaccard similarity
coefficient


pseudo code

ELSE {

dMaxOverlap =
x1
; //e.g. 0.5

if ((lexon2MaxOverlap == null) ||

(lexon2MaxOverlap.dMaxOverlap < dMaxOverlap)) {

lexon2MaxOverlap = lexon2;

lexon2MaxOverlap.dMaxOverlap =
dMaxOverlap;

}

} ELSE IF (lexon1.Role.equals(lexon2.Role)) {

IF (lexon1.CoRole.equals (lexon2.CoRole)) {

DmaxOverlap =
x2
//e.g. 0.5;

IF ((lexon2MaxOverlap == null) ||

(lexon2MaxOverlap.dMaxOverlap <
dMaxOverlap)) {

lexon2MaxOverlap = lexon2;

lexon2MaxOverlap.dMaxOverlap =
dMaxOverlap;

}

}

}

25/10/2013

| pag.
28

ELSE {

DmaxOverlap =
x3

//e.g. 0.25;

IF ((lexon2MaxOverlap == null) ||

(lexon2MaxOverlap.dMaxOverlap < dMaxOverlap)) {

lexon2MaxOverlap = lexon2;

lexon2MaxOverlap.dMaxOverlap = dMaxOverlap;

}

}

...

}


Add

+= dMaxOverlap;

}//while


WHILE (LS2. Has Next ()) {

… //similar

}

Score = Add / Union
;

RETURN Score;

}

Controlled Fully Automated
Ontology Based Matching
Strategy (C
-
FOAM)


25/10/2013

| pag.
29

String
matching

Lexical
matching

Graph matching

(any combination of ≠
graph algorithms)

Controlled Fully Automated
Ontology Based Matching
Strategy (C
-
FOAM)


2 modules: (1)
Interpreter

and (2)
Comparator


25/10/2013

| pag.
30

Similarity
Score

Graph Matching

Ontology
-
based
Comparator

Matching at
lexical level

Matching at
string level

Score +

Penalty

Score +

Penalty

Pre
-
processing by the
Interpreter

“hearty”

“warmhearted”

String matching

Lexical matching

&

“heart” (ontology) + its
annotation set

Controlled Fully Automated
Ontology Based Matching
Strategy (C
-
FOAM)

25/10/2013

| pag.
31

Advantages

Drawbacks

Contains all the advantages of the selected algorithms.

It supports fuzzy inputs from end users


e.g. both “hearty” and “warmhearted” are interpreted

as “heart”

Therefore: provides a lot of freedom to the end users.

Robust (rarely fails)


Contains all the disadvantages of the
selected algorithms

C
-
FOAM performs worse (in terms of complexity) than the
worst among the selected algorithms (since it’s a composition
of these algorithms)

Algorithms are inter
-
dependent (
one
algorithm needs to wait
until another algorithm finishes the calculation.)

ODMF Applications


3 EU projects


FP6 Prolix


Competency matching (learning & training in Competency Management)


FP6 3D Anatomical Human (3DAH)


Anatomic data matching (human anatomy: images, videos, text, etc.)


FP7 TAS
3

(Trusted Architecture for Securely Shared Services)


Security attributes matching (security and privacy)

25/10/2013

| pag.
32

Prolix


competency
matching


25/10/2013

| pag.
33

Term1

Role

Co
-
role

Term2

person

empathise

Is empathised by

person

person

manage

Is managed

by

emotion

person

respect

Is respected by

context

person

Interact with

Interact with

person

person

Deal with

Is dealt by

Communication

simplicity

is char. of

has char. of

Communication

clarity

Is char. of

Has char. of

Communication


Straightforward & Heart

Term1

Role

Co
-
role

Term2

employee

describe

Is described by

Process model

employee

define

Is defined by

Role

employee

define

Is defined by

Function

employee

distinguish

Is distinguished

by

Service

employee

describe

Is described by

Practice

employee

describe

Is described by

Service

employee

define

Is defined by

Service

Course ITIL1

GRASIM result
-

example


Matching a competence (“heart”) with a learning material (“Problem
Solving and decision making”)

25/10/2013

| pag.
34

3DAH


anatomic data
matching


Virtual teacher


3 components:


Knowledge Base


Anatomy Browser


Controlled Fully Automated Ontology
-
based Data Matching Strategy (C
-
FOAM)


Match: anatomical data (images, videos, books, etc.) with user
knowledge (captured from Computer
-
Human Interactions)


Purpose


Evaluate students


Retrieve and deliver personalized suggestions on the learning materials in order
to improve the students’ skills





25/10/2013

| pag.
35

Virtual teacher


the
Knowledge Base


25/10/2013

| pag.
36

10/25/2013



Lexons
:
Extensor

Hallucis

Longus Tendon

Extensor Hallucis
Longus Tendon

part of

part

Tendons Of Lower Leg

Extensor Hallucis
Longus Tendon

affected by

affects

Rupture

Extensor Hallucis
Longus Tendon

reconstucted with

reconstructs

Gracilis Tendon
Autograft

Extensor Hallucis
Longus Tendon

repaired by

repairs

Sugery

Extensor
Hallucis

Longus

Tendon

replaced by

replaces

Accessory Tendon

Lexons
:
Acetabular

Labrum

Acetabular

Labrum

part of

part

Cartilage of Hip

Acetabular

Labrum

stabilizes

stabilized by

Hip Joint

Hip Arthroscopy

repairs

repaired by

Acetabular

Labrum

Hip Arthroscopy

is a

is

Surgery

Arthroscopy

diagnoses

diagnosed by

Acetabular

Labral

Tear

Acetabular

Labrum

stabilizes

stabilized by

Hip Joint

Acetabular Labral Tear

affects

affected by

Acetabular Labrum

Acetabular Labral Tear

produces

produced by

High Stresses

Acetabular Labral Tear

produces

produced by

Joint Deterioration

Acetabular Labral Tear

produced by

produces

Direct Trauma

Extensor Hallucis Longus Tendon
(image)

Acetabular labrum (video)

25/10/2013

| pag.
37

Virtual Teacher


the
Anatomy Browser


3DAH Viewer [UNIGE&CRS4]


Knowledge Interaction Framework


Queries submitted to the CMS


Online knowledge retrieval

10/25/2013

http://3dah.miralab.ch/index.php?option=com_remository&Itemid=78&func=finishdown&id=107


25/10/2013

| pag.
38

Virtual Teacher


C
-
FOAM

10/25/2013


Ω


(musculoskeletal ontology)




overlapping rate of L
i

and L
i


L


lexon set
describing labeled
concepts (
e.g.
“Patella”
)

L’


lexon set
describing the
learning materials
(
e.g. “Imaging of the
dysplasia”
)

C
-
FOAM


-

student input


-

concept label linked to the
interpretation of

We denote
-

synonym set of

3 possible situations:


eLearning Scenario


8 steps

10/25/2013

1.

The

viewer

shows

highlighted

zone


2.

The

student

gives

input


3.



labrum

Black box

-

Matching engine works

-

Finds labrum


Acetabular labrum




x labrum




y labrum, ...

40% correct

calculate

4
.

5.

Final score 70%

6.

-
>5 Questions

7. The computer shows the correct
answers for answer ≠ 100%


8. The computer finds the learning
materials



“Hip Arthroscopy”

“Acetabular Labral Tears”

“Operative Hip Arthroscopy”

Acetabular
labrum

Results


10/25/2013

Image (correct
answer)

Student
’s input

卣ore

Message

卵gges瑥d
ma瑥rials

Remarks


patella

patella

1

perfect

-


String matching


patella

patela

0,98

[TYPO] typo error

-


String matching


patella

kneecap

0,75

[SYNONYM] patella

Publ1, publ2,…

Lexical matching


patella

knee

0,69

[SYNONYM+TYPO]
kneecap

Publ1, publ2,…

Lexical matching


gluteus
medius

gluteus

0,71

[TYPO] typo error

Video1, publ1,
publ2,…

String matching


acetabular

labrum

labrum

0,71

[TYPO] typo error

Video1, Publ1,
Book1,…

String matching

Extensor
hallucis

longus

tendon

hallucis

longus

tendon

0,87

[TYPO] typo error

Publ1,

Publ2,…

String matching


popliteus

plantaris

0,30

[ONTOLOGY] is
-
a

Video1, Publ1,
Page1, Web1,…

Role
-
co
-
role
matching

popliteus

disease

0,11

[ONTOLOGY]
lexon

term

Video1, Publ1,
Page1, Web1,…


Lexon

matching

Total score

73

Maximum
score

100

Interpretation

10/25/2013


Patella


4 situations


Patela

-

0,98 error spelling


Patella
-

1


Kneecap
-

0,75 synonym


Knee
-

0,69 synonym + error spelling


C
-
FOAM


JaroWinkler

(
pate
l
a
-
>pate
ll
a)


WordNet

(kneepan)


LeMaSt

They both use LeMaSt (graph
algorithm)

Combined
(advanced
C
-
FOAm)

pate
l
a pate
ll
a



kneecap patella

Jaro

Wrinkler

Word

Net

lexons

Learning
materials

1

2

TAS
3



security policy
matching


Semantic interoperability


Between SR & SP


Security Policy Ontology (SecPODE)


Subjects, Actions, Targets


(Domination) Relation lookup (between 2 terms originating from
different security policies)


C
-
FOAM matching strategy


Context: authorization architecture

25/10/2013

| pag.
42

Ontology
-
based Interoperation
Service (OBIS)


OBIS


WS


OBIS main method:





Val:

-
3

“I don’t know SP”
=
J
O
=
“I don’t know SR”
=
J
1
=
“SR is
=
not=牥lated=to=卐


0

“SR is less general than
=
⡤o浩nated=b礩y卐


1

“SR is equivalent to SP”
=
O
=
“SR is more general than (dominates) SP”
=
OBIS


Example of security concept dominance:


C
1

C
2

Level

of

dominance

PlacementService

PlacementService

1

(equivalent)

Email

PersonalDetails

0

(less

general)

French

Arabic

-
1

(not

related)

OBIS Architecture

controlled

mapping

Use Case Scenario


Employability Demonstrator

Use Case Scenario

Use Case Scenario

OBIS Role

Security Concept Matching

To check whether
the placement coordinator (PC) can
access
Betty’s
personal data by calculating the relation
between the placement
coordinator
role

in the request
and the
role
in Alice’s policy (to determine whether a
foreign role dominates the local role in the credential
validation policy).


Role

matching for CVS.


Translator
:


SR
:
Placement Co
-
ordinator

=
PlacementService


SP
: Placement Service
= PlacementService


Path finder:
1 “equivalence”



Attribute is valid



To determine whether the
resource

to be accessed by the
service requester (
e.g

Betty’s Email)
is more specific
than
the
resource protected by the security policy (e.g.
Betty’s
PersonalDetails
).

Resource

matching for PDP.

Translator
:

SR
:
Email
=
Email

SP
:
PersonalDetails

=
PersonalDetails


Path finder:
0 “less general”



Access granted

Use Case Scenario

25/10/2013

| pag.
49

1

2

3

4

5

6

6

9

9

7

7

8

8

10



Service requester = Placement Co
-
ordinator



Resource = Email

OBIS for credential
validation

25/10/2013

| pag.
50







Step1: CVS invokes OBIS


Step2: OBIS performs the (ontology
-
based data) matching


Step3: OBIS returns Val =
1 (i.e. ‘equivalence’)
i.e.
‘equivalence’)

CVS

OBIS

1: (‘Placement Co
-
ordinator’,
‘Placement Service, Val)

2: matching (

‘Placement Co
-
ordinator’,

‘Placement Service’)

3: Val = 1

OBIS for credential
validation

25/10/2013

| pag.
51

OBIS for access control

25/10/2013

| pag.
52





Previous step: credentials


validated)


Step1: PDP invokes OBIS


Step2: OBIS performs the (ontology
-
based data) matching


Step3: OBIS returns Val = 0 (i.e. ‘less general’)

PDP

(PERMIS)

OBIS

1: (‘Email’, ‘PersonalDetails’, Val)

2: matching (

‘Email’,

‘PersonalDetails’)

3: Val = 0

OBIS for access control

25/10/2013

| pag.
53

Final remarks


ODMF foundations & Applications


Please refer to the bibliography for details


ODMF Evaluation Methodology


Not covered in this course


Please refer to [7] for details


25/10/2013

| pag.
54

Questions





iciuciu@vub.ac.be



yan.tang@vub.ac.be


25/10/2013

| pag.
55

References


Ontology Matching

[
1
]

Jain,

P
.
,

Hitzler
,

P
.
,

Sheth
,

A
.
P
.
,

Verma
,

K
.
,

Yeh
,

P
.
Z
.
,

Ontology

Alignment

for

LOD,

Proceedings

of

ISWC’
10
,

Springer
-
Verlag
,

Heidelberg
.

Link
:

http
:
//
knoesis
.
wright
.
edu/library/publications/iswc
10
_paper
218
.
pdf

[
2
]

Interoperability

and

Supply

Chain
.

Link
:

http
:
//eil
.
stanford
.
edu/supply_chai n
/

[
3
]

Knowledge

Web

FP
6
-
507482
,

Link
:

http
:
//knowledgeweb
.
semanticweb
.
org/semanticportal/sewView/frames
.
html


Data Matching

[
4
]

Elmagarmid
,

A
.
K
.
,

Ipeirotis
,

P
.
G
.
,

Verykios
,

V
.
S
.
,

Duplicate

Record

Detection
:

A

Survey,

IEEE

Transactions

on

Knowledge

and

Data

Engineering,

Vol
.

19
(
1
),

2007
.

Link
:

http
:
//
www
.
cs
.
purdue
.
edu/homes/ake/pub/TKDE
-
0240
-
0605
-
1
.
pdf

[
5
]

Software
:

http
:
//www
.
cs
.
umd
.
edu/projects/li nqs/ddupe
/

[
6
]

Software
:

http
:
//sourceforge
.
net/projects/simmetrics/


Ontology
-
based Data Matching

[
7
]

Y
.

Tang,

R
.

Meersman
,

I
.
G
.

Ciuciu
,

E
.

Leenarts
,

K
.

Pudney
.

Towards

Evaluating

Ontology

Based

Data

Matchi ng

Strategies
.

Proceedings

of

4
th

IEEE

Research

Challenges

in

Information

Science

(RCIS’
2010
),

2010
.

Link
:

http
:
//starlab
.
vub
.
ac
.
be/website/files/TEObDMS
.
pdf

[8]
http://secondstring.sourceforge.net
/

[9] Cohen, W.W.,
Ravikumar
, P., Fienberg, S.E., A Comparison of String Distance Metrics for Name
-
Matching Tasks,
IIWeb

2003.
Link:
http://
secondstring.sourceforge.net/doc/iiweb03.pdf

[10]
WordNet
:
http://wordnet.princeton.edu
/

25/10/2013

| pag.
56

25/10/2013

| pag.
57

References


Applications

[11]
Tang, Y.
Towards

Evaluating

GRASIM for
Ontology
-
based

Data
Matching
, in proc. of the 9th international
conference

on ontologies,
databases
, and applications for
semantics

(ODBASE'2010), Springer
Verlaag
, LNCS 6427,
2010. Link:
http://starlab.vub.ac.be/website/files/Towards%20Evaluating%20GRASIM%20for%20Ontology
-
based%20Data.pdf

[12]

Ciuciu
, I., Tang
,
Y., A
personalized and collaborative eLearning materials recommendation scenario using
ontology
-
based data matching strategies, Proceedings of On The Move Federated Conferences and Workshops
(OTM’2010),
SeDeS

workshop, vol. LNCS 6428, 575
-
584, Crete, Greece, 2010.
Link:
http
://starlab.vub.ac.be/website/files/64280575.pdf

[13]

Ciuciu
, I.,
Reul
,
Q
., Zhao,
G
.,
Meersman
,
R
., Chadwick, D.,
Hibbert
,
M
., Ontology
-
based
interoperation for securely
shared services, Proceedings of 4
th

IEEE International Conference on New Technologies, Mobility and Security, ISBN:
978
-
1
-
4244
-
8705
-
9, Paris, France, February 2011
. Link:
http://starlab.vub.ac.be/website/files/PID1066536.pdf