TheHiddenU

A Social Nexus for Privacy-assured Personalisation Brokerage

G. Kappel, J. Schönböck, M. Wimmer

Vienna University of Technology, Favoritenstrasse 9-11, A-1040 Vienna, Austria

G. Kotsis, A. Kusel, B. Pröll, W. Retschitzegger, W. Schwinger, R. R. Wagner

Johannes Kepler University Linz, Altenbergerstrasse 69, A-4040 Linz, Austria

S. Lechner

Netural Communication, Europaplatz 4, A-4020 Linz, Austria

Keywords: Social Networks, Personalization, Profiles, Integration, Service Brokerage.

Abstract: Social networks have seen enormous growth over the past few years, providing also a powerful new channel

for distributing personalized services. Personalization, however, is exacerbated because social content is

scattered across different social networks serving specific human needs and social networkers are

particularly reluctant to share social content with service providers, if not under their full control. This paper

sketches TheHiddenU, a social nexus exploiting semantic techniques for integrating, profiling and

privatising social content, thereby providing the technical prerequisites for personalized brokerage, a new,

sustainable business model in the Social Web.

1 INTRODUCTION

Social networks have seen enormous growth being

already used by more than two-third of the world’s

Internet population (Nielsen, 2009). Thus, service

providers like shopping sites, travel portals, or news

sites are urgently forced to exploit social networks

as additional distribution channel for providing

highly personalized services to social networkers.

Currently, however, both social networkers and

service providers are in an undesirable situation

(Halpin, 2009). Firstly, social content is scattered

over several social networks each serving specific

human needs like sharing friendships, business

contacts or music interests. Secondly, these isolated

views bear the pitfall of untargeted services.

Thirdly, social networkers have a great sense of

ownership being reluctant to share social content

with service providers for personalization, at least if

it is obscure and beyond their control, how their

social content is being exploited (Kobsa, 2007b).

Let's consider an online media retailer intending

a recommendation service for birthday gifts on time,

matching the interests of friends while fitting the

donor’s economic situation – if only they could

know. At the same time, social networkers might use

facebook

to connect to friends, Last.fm

to serve

music interests, and XING

for business issues – if

only they could oversee.

TheHiddenU system proposed herein is intended

to act as a social nexus, integrating scattered social

content into a common social profile and, on basis of

that, explicating formerly hidden knowledge by

means of profiling. At the same time, TheHiddenU

can act as a personalization broker. This means that

social networkers are enabled to share, in a privacy-

assured way, necessary profile information with the

service provider to receive, e.g., gift

recommendations.

The paper is structured as follows. The next

section describes TheHiddenU’s architecture and

* funded by Austrian Federal Ministry of Transport, Innovatio

and Technology under FIT-IT Semantic Systems (819577,

819584).

1 www.facebook.com

2 www.lastfm.com

3 www.xin

.com

158

Kappel G., Schönböck J., Wimmer M., Kotsis G., Kusel A., Pröll B., Retschitzegger W., Schwinger W., R. Wagner R. and Lechner S. (2010).

TheHiddenU - A Social Nexus for Privacy-assured Personalisation Brokerage.

In Proceedings of the 12th International Conference on Enterprise Information Systems - Software Agents and Internet Computing, pages 158-162

DOI: 10.5220/0003016201580162

 SciTePress

Integrator

Adaptor Profiler

Service Provider

Directory

BrokerageSandbox

RSS

API

XML

ResolveDatamodel

Heterogeneity

ResolveTechnical

Heterogeneity

Isolated

Profiles

Repository

Isolated Profiles

Enriched Profiles

Integrated Profiles

Integrated

Profile

Enriched

Profile

RDF

GenerationGeneration

Discovery

Subscription

Service

Descriptions

Subscribers

Service

Bazaar

Social

Networkers

DashBoard

Personalized

Services

RetrievalRetrieval

Provenance Information

Validate

UserInterest

Service Provision

BrokerBoard

AdminBoard

TheHiddenU Provider

DataCleansing

SchemaMapping

DataFusion

Query

Reasoning

Provision

Provenance Layer

Privacy Policies

Service

Providers

…

Service

Providers

…

External

Knowledge Sources

…

Social

Networks

…

Figure 1: System Architecture of TheHiddenU.

approach.

Afterwards the envisioned prototypical

implementation is discussed, proposing concrete

technologies, which seem to be suitable for our

purposes. Finally, a comparison of related

approaches and first ideas concerning the evaluation

of TheHiddenU’s feasibility conclude the paper.

2 ARCHITECTURE AND

APPROACH

The architecture of TheHiddenU, as depicted in

Figure 1, is divided into two major components, a

Sandbox which is under control of the social

networker for integrating, profiling and privatizing

social content, and a Brokerage area for connecting

service providers to gain personalized services. In

the following, these two components are described

in more detail in a process-oriented manner.

2.1 Hybrid Approach for Social

Content Integration

First of all, TheHiddenU provider connects social

networks via the AdminBoard by adaptors resolving

technical and data model heterogeneities, resulting

in Isolated Profiles for each source stored in a

Repository. Further on, the Integrator component

allows for a hybrid integration involving both,

schema level and instance level. For this,

TheHiddenU provider can apply data cleansing

operators to resolve lexical heterogeneities,

mapping operators to resolve structural ones and

data fusion operators for dealing with duplicates and

conflicts to achieve an Integrated Profile. Social

networkers, in turn, can view, verify and change

their integrated profile via the DashBoard.

2.2 Semantic-based Profiling &

Brokerage

Service providers and social networkers will be

provided with an extensible library of reusable

profiler components and a language for their

composition. First, input components allow for

querying recorded profile and provenance data and

external sources. Second, reasoning components are

foreseen for inferring new knowledge like a “best-

friend” relationship and for refining and augmenting

existing knowledge like computing the strength of

an "interested-in” relationship or associating favorite

music tracks to external music categories. Third,

provision components are provided for materializing

the newly derived knowledge, resulting in a

Semantically Enriched Profile.

For brokerage, service providers register their

services via the BrokerBoard in the Service Provider

Directory, indicating those parts of the social

content necessary for a service to operate (e.g., the

gift's occasion) and optional ones (e.g., a price

range), together with possible provision channels

(e.g., eMail or widget). Social networkers can

discover and subscribe services on basis of interest-

driven and collaborative filtering using the

DashBoard. The ServiceBazaar realizes a

negotiation process, allowing social networkers to

balance between privacy concerns and

TheHiddenU - A Social Nexus for Privacy-assured Personalisation Brokerage

159

personalization needs. Succeeding service provision

takes place in a pseudonymous way.

2.3 User-empowered Provenance &

Privacy-control

The ProvenanceLayer automatically records all

information about integration and profiling activities

within the Repository. By means of a

ProvenanceViewer which is part of the DashBoard,

social networkers are supported to understand which

social content is stored, how it is computed, and how

it is used. Privacy concerns of social networkers are

made explicit in terms of adaptable privacy policies

with respect to social content (i.e. social content

policies) and with respect to service providers (i.e.

service provider policies).

3 PROTOTYPICAL

IMPLEMENTATION

The rationale behind the selection of techniques

intended for the implementation of TheHiddenU is

mainly based on our experience in the areas of

ontology-driven situation-awareness (Baumgartner

et al. 2010), model-driven engineering (Wimmer et

al. 2009) and aspect-oriented software engineering

(Wimmer et al. 2010). To start with, the central

repository will build on existing RDF triple stores

like Jena

. To provide a unified and extensible

representation of social content, an ontology will be

developed, incorporating the core concepts prevalent

in user models (Kobsa, 2007a) (e.g., personal data,

user knowledge, capabilities, interests, and

relationships) and in social ontologies (Sutterer et

al. 2008) and standards like OpenSocial

, FOAF

(Friend of a Friend) allowing to describe persons,

their activities, and relationships, SIOC

(Semantically-Interlinked Online Communities)

contributing interactions and resources, and

RELATIONSHIP

, specializing relationships (e.g.,

family, professional). The adaptors acquiring

knowledge from social networks, should be based

on existing approaches e.g., the Linking Open Data

Project

and the Data Portability Project

. To

realize our user interfaces, DashBoard, AdminBoard

and BrokerBoard, the open source Ajax framework

ICEfaces

will be used.

3.1 Social Content Integration

For realizing the Integrator, we will apply model-

driven techniques, in particular the Eclipse Modeling

Framework (EMF)

and the Graphical Modeling

Framework (GMF)

, allowing a graphical

representation of the integration operators in a

mashup-based manner. For the definition of new

integration operators a domain specific language

should be offered whereby, e.g., the Epsilon Merge

Language

would be well suited. Finally, the

resulting mix of code snippets must be transformed

e.g., on basis of OpenArchitectureWare

into our

runtime environment Jena allowing for a proper mix

of SPARQL and Java code.

3.2 Profiling and Brokerage

Profiling will be based on the same technologies as

used for the Integrator since we envision again a

mashup-based approach, naturally differing in the

realized operator logic and the underlying execution

platform. Reasoning operators should make use of

both, declarative techniques on basis of generic

rules (e.g., subsumption) as well as user-defined

ones and imperative techniques in case declarative

reasoning is not flexible enough. In this respect,

SWRL

seems to be promising, allowing the

combination of OWL with RuleML on basis of

inference engines like Pellet

or AllegroGraph

providing declarative reasoning with RDFS++ and

logic programming with Prolog. For service

brokerage, we will base on existing semantic web

standards like UDDI and OWL-S.

3.3 Provenance and Privacy

For introducing provenance recording into the

integration and profiler components, Higher-Order

Transformations (HOTs) expressed by the ATLAS

Transformation Language

will be used, the

Provenance Viewer will be again based on

EMF/GMF. Concerning privacy, the approaches

ofPersonis (Kappel et al. 2009) and Kobsa (Kobsa,

4 jena.sourceforge.net

www.opensocial.org

www.foaf-project.org

sioc-project.org

vocab.org/relationship.html

tinyurl.com/OpenDataProject

www.dataportability.org

11 www.icefaces.org

12 www.eclipse.org/emf

13 www.eclipse.org/gmf

14 www.eclipse.org/gmt/epsilon

15 owa.itemis.de

16 ruleml.org/swrl

17 clarkparsia.com/pellet

18 www.franz.com/agraph

www.eipse.org/m2m/atl

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

160

2007b) are of foremost interest since Personis puts

forward the privacy principle of scrutability and

Kobsa focuses on dynamic, adaptable privacy

policies.

4 RELATED WORK

There is already plethora of related approaches in

the areas of integrating, profiling and privatizing

social content, which is briefly reviewed in the

following. For a more detailed comparison of

TheHiddenU to related work it is referred to (Kappel

et al. 2009).

4.1 Social Content Integration

Recently, commercial meta social networks like

Power.com

, snag

or OrSiSo

emerged, providing

a single-point-of-access to other social networks but

miss an integrated view resolving heterogeneities as

intended by TheHiddenU. Research projects already

build on an ontology-based infrastructure, like (Gosh

et al. 2009), (Golemati et al. 2007), where a social

profile is built up by aggregating content spread over

multiple services, or (Noor et al. 2009), where tags

of a certain social networker are analyzed to build up

a social profile for recommendations. In contrast to

TheHiddenU, however, brokerage is not an issue.

Regarding integration, in the last 30 years, numerous

approaches where brought forward in the areas of

data engineering (Bleiholder et al. 2008),

(Elmagarmid et al. 2007), (Rahm et al. 2000),

ontology engineering (Baumgartner et al. 2010),

(Le-Phuoc et al. 2009) and model engineering

(Bleiholder et al. 2008), (Elmagarmid et al. 2007),

which are of high value for our approach.

Nevertheless, the majority of them address the

whole spectrum of integration tasks necessary in

TheHiddenU in an isolated way, not least since data

cleansing and fusion deal with instances whereas

schema mapping deals with types.

4.2 Profiling and Brokerage

Regarding the profiling aspect of TheHiddenU, first

of all, research in the area of personalization and

recommender systems (Bonhard et al. 2007), (Kobsa

et al. 2007a) is of high relevance. A crucial

challenge will be, however, to reason with

incomplete and vague information since it can

neither be predicted which information social

networkers have entered, nor which social networks

they have joined. In (Lukasiewicz et al. 2009),

formal and practical means for automated reasoning

with incomplete and vague information are

proposed, which will be adapted to our context.

Concerning existing reasoning techniques, in

(Huang et al. 2008) an overview on different

approaches is given, whereby rule-based reasoning

for dynamic and incomplete knowledge best fits our

requirements. More widely related to our approach

are social network analysis methods which focus on

exploring general properties of a whole social

network using graph theory (Ereteo et al. 2008),

whereas we intend to reason about a single social

networker. Concerning the brokerage aspect,

relevant work was brought forward in the coalescing

areas of Service Oriented Architecture (SOA) and

semantic web as well as more distant, also in the

area of agent-based and self-organized systems

(Dustdar et al. 2009).

4.3 Provenance and Privacy

Concerning provenance, research primarily focused

on describing the origins of data (Tan, 2007) in

terms of workflow-oriented approaches providing a

coarse-grained focus on data processing, data-

oriented approaches aiming at a fine-grained view

and approaches incorporating social aspects (Harth

et al. 2007). Only few explicit models, formally

representing provenance have been published, being

a recent research topic especially in combination

with privacy aspects (Kobsa, 2007b). Approaches

for retrieval and visualization of provenance most

often do not go beyond simple provenance queries

(Simmhan et al. 2005), although being

overwhelming for users (Chapman et al. 2007).

Finally, privacy approaches in the realm of

personalization have been postulated recently, only

(Kobsa, 2007b).

5 EVALUATION METHOD

The feasibility of TheHiddenU will be evaluated in

three phases. Firstly, the proposed integration

approach will be analyzed in the course of

experiments, establishing in particular a test-set for

content integration by acquiring real world data of

volunteer social networkers, supplemented by

synthetic data containing diverse heterogeneities.

Secondly, expressivity and productivity gained by

our profiling language will be evaluated in an

20 ww.power.com

ww.dapper.net/applicaitons/Snag

22 www.orsiso.com

TheHiddenU - A Social Nexus for Privacy-assured Personalisation Brokerage

161

empirical study, incorporating about 100 students of

our semantic web course. Finally, case studies will

be conducted together with our demonstrators, the

eMedia retailer Thalia

and the eRecruiting

company IVENTA

, in order to evaluate the

TheHiddenU in a real world setting. In particular,

three exemplary social networks will be integrated

by the system provider, dedicated profilers will be

developed by our demonstrators to personalize their

services and volunteers will refine their integrated

profile and subscribe to services. On basis of

structured interviews, it will be explored, first, to

which extent personalization and service provision

met the volunteer’s requirements, second which

social content is shared with the service provider to

analyze privacy attitudes and third, to what extent

the provided provenance information procured

privacy control over the volunteers social content.

REFERENCES

Baumgartner, N., Gottesheim, W., Mitsch, S,

Retschitzegger, W., Schwinger, W.: BeAware!

Situation Awareness, the Ontology-Driven Way.

accepted for DKE Journal, Elsevier, 2010.

Bizer, C., Heath, T., Berners-Lee, T.: Linked Data - The

Story So Far. Int. Journal on Semantic Web and

Information Systems (IJSWIS), 2009

Bleiholder, J., Naumann, F.: Data fusion. ACM Computing

Surveys, 41(1), 2008

Bonhard P., et al.: The devil you know knows best – How

online recommendations can benefit from social

networking. Proc. of the 21st British CHI Group

Annual Conf. on HCI, 2007

Chapman, A., Jagadish, H. V.: Issues in Building Practical

Provenance Systems. IEEE Data Eng. Bull., 30 (4),

2007

Dustdar, S., et al.: Self-Adaptation Techniques for

Complex Service-oriented Systems. Proc. of 5

Int.

Conf. on Next Generation Web Services Practices,

IEEE, 2009

Elmagarmid, A. K., Ipeirotis, P. G., Verykios, V. S.:

Duplicate Record Detection: A Survey. IEEE

Transactions on Knowledge and Data Engineering, 19

(1), 2007

Ereteo, G., et al.: A State of the Art on Social Network

Analysis and its Applications on a Semantic Web.

Proc of the 1

Social Data on the Web Workshop

(SDoW’08), Karlsruhe, Oct. 2008

Ghosh, R., Dekhil, M.: Mashups for semantic user

profiles. Proc. of the 17th Int. Conf. on World Wide

Web (WWW’08), ACM, 2008

Golemati, M. et al.: Creating an Ontology for the User

Profile: Method and Application. Proc. of the 1

Int.

Conf. on Research Challenges in Information Science

(RCIS’07), 2007

Halpin, H.: Ten Theses on the Future of Social

Networking. Proc. of W3C Workshop on the Future of

Social Networking, Barcelona, 2009

Harth, A., Polleres, A., Decker, S.: Towards a Social

Provenance Model for the Web. Proc. of Workshop

on Principles of Provenance (PrOPr’07), 2007

Huang, Z. et al.: A Survey of Web Scale Reasoning. FP7

Deliverable, LarKC/2008/D4.1, 2008

Kappel, G., Kotsis, G., Kusel, A., Lechner, S., Pröll, B.,

Retschitzegger, W., Schönböck, J., Schwinger, W.,

Wimmer, M.: TheHiddenU - A Social Nexus for

Privacy-Assured Personalisation Brokerage, Technical

Report, University Linz, 2009,

http://www.ifs.jku.at/download/TheHiddenU.pdf

Kay, J.: Scrutable Adaptation: Because We Can and Must.

Proc. of 4th Int. Conf. on Adaptive Hypermedia and

Adaptive Web-Based Systems, Springer, LNCS 4018,

2006

Kobsa, A.: Generic User Modeling Systems. The Adaptive

Web, Springer, LNCS 4321, 2007a

Kobsa, A.: Privacy-enhanced personalization.

Communication of the. ACM 50(8): 24-33, 2007b

Lukasiewicz, T., Straccia, U.: Description Logic Programs

under Probabilistic Uncertainty and Fuzzy Vagueness.

Journal of Approximate Reasoning, 50(6), 2009

Nielsen: Global Faces and Networked Places – Report on

Social Networking’s New Global Fingerprint.

http://blog.nielsen.com/nielsenwire/wp-content/

uploads/2009/03/nielsen_globalfaces_mar09.pdf, 2009

Noor, S., Martinez, K.: Using Social Data as Context for

Making Recommendations: An Ontology based

Approach. Proc. of the Workshop on Context,

Information And Ontologies (CIAO’09), 2009

Le-Phuoc, D. et al.: Rapid prototyping of semantic mash-

ups through semantic web pipes. Proc. of the 18

Int.

Conf. on World Wide Web (WWW '09), ACM, 2009

Rahm, E., et al.: Data Cleaning: Problems & Current

Approaches. IEEE Data Eng. Bull., 23(4), 2000

Simmhan, Y. L. et al.: A survey of data provenance in e-

science. SIGMOD Rec. 34(3), 2005

Sutterer, M. et al.: UPOS: User profile ontology with

situation-dependent preferences support. Proc. of the

Int. Conf. on Advances in Computer-Human

Interaction, IEEE, 2008

Tan, W.: Provenance in Databases: Past, Current, and

Future. IEEE Data Eng. Bull., 30(4), 2007

Wimmer, M., Kusel, A., Schoenboeck, J., Kappel, G.,

Retschitzegger, W., Schwinger, W.: Reviving QVT

Relations: Model-based Debugging using Colored

Petri Nets. Proc. of ACM/IEEE 12

Int. Conf. on MDE

Languages and Systems (MODELS’09), Denver, 2009

Wimmer, M., Schauerhuber, A., Retschitzegger, W.,

Schwinger, W., Kappel, G., Kapsammer, E.: A Survey

on UML-based Aspect-Oriented Design. accepted for

ACM Computing Surveys, 2010.

23 www.thalia.at

www.inventa.at

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

162