Traversal and relations discovery among business entities and ...

farmpaintlickInternet και Εφαρμογές Web

21 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

89 εμφανίσεις

Dejan

Lavbič

Dejan.Lavbic@fri.uni
-
lj.si

http://www.lavbic.net

University of Ljubljana,

Faculty of Computer and Information Science,

SLOVENIA

Agenda


Motivation

» semantic integration » problem of trust


Problem


Trust

and semantic integration of data

»
modelling trust


SocioLeaks
case study

» technology » ontologies »
example case study


Conclusions

Motivation
(1)


semantic integration

of various data sources
that include information about
business
entities

and
people


t
he problem of
trust

as a method of dealing
with uncertainty


especially when dealing with online personal
identity


government registers vs. online social networks,
newspaper archives etc.

Motivation

(2)

Identify person from keyword
and display known properties.

Sources


Wikipedia


Freebase




Problem
(1)


Lack of
semantically integrated information

about online
personal identity

with the
purpose of:


coping with
corruption

in crossing the frontiers
of legislation,


fraud detection

in banks, insurance companies
and other public institutions,


pattern discovery

and
identification
.

Problem

(2)


current approaches

deal with integration of
information from several data sources and
omit

or
don't fully address

the aspect of

trust
,


main focus

on

personal information from
social networks

which are
not very reliable

as users for various reasons tend to give
false information.

Trust and semantic integration
(1)

Definition of trust


T
rust

is …


a measurable belief that utilizes personal
experiences


experiences of others or possibly combined
experiences, to make
trustworthy decisions

about an
entity


a
trustworthy decision is assumed to be a transitive
process such that there is
a web of trust network

in
which a link between two entities means that a
trustworthy decision has been made and the
quantitative value of that trust has been evaluated
.

Trust and semantic integration

(2)

Modelling trust

(1)


our approach is based on
RDF language

(extends to RDFS, OWL etc.),


different types of trust

can be defined for
each entity


data source trust

𝑇

𝑒


entity trust

𝑇

𝑒
, which further

consist of


schema level entity trust

𝑇

𝑥
𝑒


instance level entity trust

𝑇


𝑥
𝑒


Trust and semantic integration

(3)

Modelling trust

(2)


trust of entity e

»
𝑇
𝑒
=
𝑇

𝑒

𝑇

𝑒


entity trust

»
𝑇

𝑒
=
𝑇

𝑥
𝑒
+
𝑎

𝑇

𝑥

𝑇

𝑥


schema level entity trust

»
𝑇

𝑥
𝑒


instance level entity trust

»
𝑇

𝑥
𝑒
=

+

+
+




degree
of incorporation of users' votes

»
𝑎
=

0
;

+
+


<

𝑎𝑥

+
,



+
+


;

+
+





Trust and semantic integration

(4)

Modelling trust

(3)


trust of entity e

»
𝑇
𝑒
=
𝑇

𝑒

𝑇

𝑒


entity trust

»
𝑇

𝑒
=
𝑇

𝑥
𝑒
+
𝑎

𝑇

𝑥

𝑇

𝑥


schema level entity trust

»
𝑇

𝑥
𝑒

Trust and semantic integration
(5)

Modelling trust » example
(4)


What degree of confidence does the information
about the instance of a class Person represent?


𝑇
𝑃𝑒 
=
𝑇
𝑃
=

𝑑
𝑃

×


𝑑
𝑃

+



𝑃

×



𝑃

+



𝑃

×



𝑃

+


𝑆
𝑃

𝑑
𝑃

+



𝑃

+



𝑃

+

1


𝑇
𝑑
𝑃
=
0
,
5
×
0
,
7
1
=
35%


𝑇

𝑃
=
0
,
92
×
0
,
9
×
0
,
5
1

41%


𝑇

𝑃
=
𝑇
𝐸 𝑖𝑦
=
1
×
1
×
1
1
=
100%



𝑇
𝑃
=
0
,
35
×
1
+
0
,
41
×
1
+
1
×
1
+
0
,
8
1
+
1
+
1
+
1


𝑇
𝑃
=
2
,
56
4

𝟔𝟒%

SocioLeaks case study

(1)

Technology


Open source technologies that support
current W3C standards

in Semantic Web
and linked
-
data applications


Apache Jena
framework

SocioLeaks case study

(2)

Ontologies

SocioLeaks case study
(3)

Prototype example
(1)

Traversal is performed by specifying entry point of 1 or 2 entities.

Defining the length of property paths to follow.

Considering the
time dimension.

The trust level threshold.

Filtering

of entities and relations.

SocioLeaks case study
(4)

Prototype example
(2)

Conclusion


Proposed the use of Semantic Web
technologies for
semantic integration

of
data about
business entities

and
people

coupled with trust layer
.


Several layers of trust


data source, schema
level entity and instance level entity.


Enables filtering the data based on the user
preference
.


Application of the approach is feasible in several
cases


banks, insurance companies etc.

Discussion


Thank you for your attention!



Questions, comments and critiques are
more
than welcome
!






»

http://www.lavbic.net


»

Dejan.Lavbic@fri.uni
-
lj.si


»

@dlavbic