User-centered Social Network Proﬁles Integration

Xuan Truong Vu, Pierre Morizet-Mahoudeaux and Marie-H

ene Abel

UMR-CNRS 7253, Universit

e de Technologie de Compi

egne, Compi

egne, France

Keywords:

Online Social Networks, Proﬁles Aggregation, User Proﬁling, Recommendation, FOAF.

Abstract:

Large scale online social networks (OSNs) such as Facebook, Twitter, LinkedIn, have become an important

part of our every day life. Users are connected to multiple OSNs in which they maintain their different proﬁles

including a lot of personal and social information. The number of friends of a given user may grow so rapidly

that it becomes impossible to manage all updates from friends’ proﬁles and to ﬁlter relevant new information.

We present a FOAF-based proﬁles aggregation model, which is able to align different user proﬁles available

on different OSNs into aggregated proﬁles within a single triple store. The aggregated proﬁles are then linked

together by friend connection. We illustrate the applicability by the presentation of some applications which

can provide users with some effective help for information searching.

1 INTRODUCTION

In recent years, social Web sites, Social networking

sites and Social media sites, have become extremely

popular (Kim et al., 2010), among which Facebook,

Twitter, and LinkedIn are the most well known ex-

amples. These OSNs attract millions of connected

users for building relationships, staying connected to

friends, family members and work colleagues, ﬁnd-

ing people who have had similar experiences, and dis-

cussing common topics of interest.

OSNs’ main function is allowing users to set up

visible proﬁles and to link to other individuals’ pro-

ﬁles. The user proﬁle is a unique page where one

can type oneself and display an articulated list of

”Friends” (Boyd and Ellison, 2007). This page may

include frames, where different kinds of information

can appear such as user activities (i.e. posts, statuses,

tags, messages, etc.). A user may be connected to

multiple OSNs and have a lot of friends, who are not

necessarily the same, on each OSN. Thus, numer-

ous personal and social information are available on

OSNs.

The number of Friends may grow so rapidly that

it becomes impossible to sort and ﬁlter all heteroge-

neous information from Friends proﬁles according to

the user’s current interest. To cope with this prob-

lem some cross-system tools have been developed to

give users useful information such as recommenda-

tion systems (Abel et al., 2011), social search services

(Zhou et al., 2012).

In this paper, we explore this idea of the aggrega-

tion of user proﬁles with the objective to provide the

user with relevant information aggregated from the

user’s social networks, which are directly related to

his/her current interest. We have ﬁrst based our work

on the Friend of a Friend ontology (FOAF) to model

users. This model will be later extended to other stan-

dards such as the Relationship Vocabulary or the Db-

pedia ontology, to enrich the aggregated proﬁles.

The paper is organised as follows. In the next sec-

tion we present work related to the aggregation of user

proﬁles from OSNs. Then we present our general so-

cial user model and its FOAF-based proﬁles aggrega-

tion. We will introduce a use case to illustrate the ap-

plicability of aggregated proﬁles. An implementation

is also presented in the Application section as a proof

of the concepts. Finally, we conclude and present our

future work.

2 RELATED WORK

Over the last few years, the increasing growth of

OSNs leads the distribution of a lot of user data within

isolated data silos. A new research challenge then

emerged, seeking solutions for sharing and reusing

user data available across OSNs.

Abel et al. (2010) have shown that FOAF can be

used as a domain speciﬁc vocabulary for aggregating

the users proﬁles from OSNs. The gathered proﬁles

have been aligned to FOAF by means of hand-crafted

473

Vu X., Morizet-Mahoudeaux P. and Abel M..

User-centered Social Network Proﬁles Integration.

DOI: 10.5220/0004353904730476

In Proceedings of the 9th International Conference on Web Information Systems and Technologies (WEBIST-2013), pages 473-476

ISBN: 978-989-8565-54-9

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

rules, however they have only investigated a reduced

number of properties (e.g. name, photos, homepage,

etc.).

A new user model, Social Web User Model, sup-

posed to be adapted to the needs of the Social Web

applications, has been introduced by Plumbaum et al.

(2011). The model is intended to include the most fre-

quent user dimensions and attributes available in 17

social applications. However, the social relationship

aspect has not been considered.

Other interesting works focus on some speciﬁc

aspects of user proﬁles : (1) Modeling user inter-

ests (Abel et al., 2011; Orlandi et al., 2012) by

combing user information proﬁling and the Semantic

Web (especially using DBpedia) ; (2) Utilizing user

preferences for collaborative recommender systems

(Shapira et al., 2012) ; (3) Modelling user expertises

and weighting user relationships for social search en-

gine (Zhou et al., 2012), (Vu and Baid, 2012).

In our approach, we extend the proﬁle aggrega-

tion proposed in these works to larger public domain

applications. We propose therefore a basic common

proﬁle model, which can be extended to a more com-

plete proﬁle, in as much as the user has provided an

access authorisation to certain protected data. Finally,

we intend to link the aggregated proﬁles together in

order to implement more useful applications.

3 PROFILES INTEGRATION

In this section, we ﬁrst introduce our general social

user model which aims at aggregating social user pro-

ﬁles. We then present the FOAF ontology upon which

we have based our user proﬁling. A use case is also

included to illustrate the model usefulness.

3.1 General Social User Model

We have studied the most frequent proﬁles proper-

ties handled respectively by the top social networks :

Facebook, Twitter, LinkedIn, Google+ and OpenSo-

cial. We have thus organised them into six dimen-

sions, which are listed below :

• Personal Characteristics includes a large range of

personal information such as name, current city,

email, gender, birthday, photo, etc.

• Friends includes connections established between

an OSN member and other members.

• Interests could be a topic (e.g. Social Networks)

or a speciﬁc entity (e.g. WEBIST 2013) that the

user is interested in.

• Groups contains information about groups, based

on attended school, hobby, interest, cause, profes-

sion, etc., in which the user has been involved.

• Studies and Works describe respectively the

school and academic experience and the profes-

sional experience of the user.

• User-created contents (UCCs) denotes contents

posted by users on OSNs. (Kim et al., 2010).

Based on this analysis, we have built a general so-

cial user model which aims at facilitating the user pro-

ﬁle information aggregation. The model covers the

ﬁve ﬁrst dimensions previously described. The UCCs

dimension is not included since it only contains raw

data. It is possible however, to extract user’s current

interests from UCCs to enrich the Interests dimension

over time (Abel et al., 2011). Our user model fol-

lows new trends from the Semantic Web approaches

in social user modelling (Noor and Martinez, 2009;

Orlandi et al., 2012). The advantage is that the model

is not static and can be easily extended.

For the present time, we have mainly used FOAF

for proﬁling users and linked data (i.e. URI), web

resources (i.e. URL), for referencing the entities of

interest such as school, location, interest, which we

explain in the next section.

3.2 FOAF-based User Proﬁling

and Aggregating

FOAF (Brickley and Mille, 2005) makes it possible

to build and manage a structured representation of

users, and the links between them. A FOAF user is

described through different properties.

The FOAF basic (name, gender, age, birthday,

location, email, photo) dimension is the same as

our User personal characteristics dimension. The

”foaf:knows” property allows to describe Friends

connections. Our Interests dimension can be repre-

sented by the ”foaf:interests” or ”foaf:topic interest”

property. The ”foaf:member” property enables to as-

sociate a user to a ”foaf:group”. The user’s studies

and works are not really speciﬁed in FOAF, but it

is possible to deﬁne the places of studies and works

thanks to two properties ”foaf:schoolHomepage” and

”foaf:workplaceHomepage” respectively. The Figure

1 shows that FOAF can handle our user proﬁling in a

very simple way, yet representative.

Therefore, based on FOAF, our user model can ag-

gregate user proﬁles available on OSNs to create an

aggregated proﬁle. Each of gathered information is

mapped to a speciﬁc FOAF property by means of a

set of hand-crafted rules (e.g. Table 1).

WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies

474

Figure 1: The user #me, named Xuan Truong Vu and based

in Compi

egne, France, has linked his Facebook and Twitter

accounts to his aggregated proﬁle. He is interested in the

Social networks and the WEBIST conference. He knows

two other users : #friend1 who is also at the UTC university

and #friend2 who is in the same town.

Table 1: Facebook - Twitter - FOAF mapping rules.

Facebook User In-

formation

FOAF Person

Property

Twitter User Pro-

ﬁle Information

User.name foaf:name User.name

User.username foaf:nick User.username

User.gender foaf:gender

User.birthday foaf:birthday

User.photo foaf:img User.proﬁle image

User.location foaf:based near User.location

User.friends foaf:knows User.friends

User.groups foaf:member User.lists

User.likes foaf:topic interest

The model cannot yet keep trace of neither the

provenance nor the adding time of any information.

Moreover, there may be conﬂictual values for a given

property. In this case, this is the user who decides

later, which information should be kept or deleted.

The friends of a user are initially saved as reduced

instances of the [foaf:Person] class (it only contains a

name and a proﬁle page).Two aggregated proﬁles will

be also linked, if their respective users are friends on

at least one social network.

3.3 Use Case

We illustrate, in this subsection, the applicability of

our model.

3.3.1 Topic-related Friends Searching

The function required is to suggest from amongst all

of a user’s contacts, the friends who can give use-

ful information about a given topic. For example,

a user intends to attend WEBIST 2013, in Aachen,

Germany. The user searches for ”Aachen, Germany”

and ”WEBIST 2013” (the application might also be

able to detect this need for information by matching

a message posted by the user on OSNs to some pre-

deﬁned pattern). Upon the query reception, the ap-

plication browses all proﬁles of the user’s friends and

looks whether they are linked to any entity the label

of which contains ”Aachen, Germany” or ”WEBIST

2013”. The application outputs : (1) two contacts also

express their interest in the conference. Moreover,

one of them is a work colleague and the other share

with the user a same interest for Social Networks ;

(2) one contact is based near Aachen Germany. The

ﬁrst kind of information pushes for meetings before,

and during the conference . The second information,

though it is peripheral to the conference, is useful for

planning the trip.

3.3.2 Topic-related Information Watching

Conversely to the preceding feature, the aim here is to

watch over the user’s network in order to ﬁlter infor-

mation that the user may be interested in. For exam-

ple, the user is interested in Web and some of his/her

contacts share the same passion. The user has anno-

tated these contacts with a series of tags (e.g. web,

technology, system) so that the application follows

their proﬁles and ﬁlters any information matching the

tags. Suppose that one of these contacts posts on

OSNs the link to the WEBIST 2013 conference, then

the application will detect it and send a notiﬁcation to

the user.

4 APPLICATION

We present in this section a prototype of our FOAF-

based user proﬁles aggregation. The prototype has

been tested with Facebook and Twitter, which are the

two most popular OSNs. Both of them makes it pos-

sible for their users to grant selected third-party ap-

plications an access to user data via their own APIs.

With respect to this policy, the prototype always asks

users for permission to access their proﬁles.

We have used different aggregators. Each of them

is dedicated to a speciﬁc social network and manages

the authentication protocols as well as data collection.

Only the basic, interests, and friends information are

collected. Data are translated into triples before being

stored in a triple store type OpenLink Virtuoso. We

have constructed some generic queries to implement

the features described in the Use case.

- Searching for friends who are based near Aachen

SELECT ?friend WHERE {

<#me> foaf:knows ?friend .

?friend foaf:based_near ?town.

?town gn:name ’Aachen’. }

- Searching for friends who are interested by WEBIST 2013

User-centeredSocialNetworkProfilesIntegration

475

SELECT ?friend WHERE {

?friend foaf:topic_interest ?interest.

?interest dc:label ’WEBIST 2013’.

FILTER ( bif:exists ((

SELECT * WHERE {<#me> foaf:knows ?friend})))}

where all aggregated proﬁles that are linked to an en-

tity the label of which contains ’WEBIST2013’, are

ﬁrst retrieved and then only friend proﬁles are re-

turned.

- Searching for web-related events from right friends

SELECT ?interest WHERE {

<#friend1> foaf:topic_interest ?interest.

?interest dc:label ?label.

FILTER (REGEX (?label, ’web’, ’i’)).

FILTER (REGEX (?label, ’conference’, ’i’)).}

The web-based service consists of a personal user

interface which permits to connect to ones Facebook,

Twitter accounts and visualise ones aggregated proﬁle

with three views basic, interest, and friends. The user

can also search for friends thanks to a keyword-based

search feature. The user’s query is translated into a

SPARQL query as cited above. For the present time,

the prototype is only able to search from aggregated

information.

We have tested our prototype with several real

users. The size of their merged friends list varies from

300 to more than 1000 connections.

5 CONCLUSIONS AND FUTURE

WORK

In this paper, we have presented a primary social

user aggregation based on the FOAF ontology. The

FOAF-based user proﬁling can (1) represent users of

OSNs, especially social aspects such as interests and

friends, (2) support the aggregation of user proﬁles

from OSNs, (3) and link aggregated proﬁles together

so that advanced searches could be possible. Our

ﬁrst prototype, implemented for Facebook and Twit-

ter, has shown the applicability of the aggregated pro-

ﬁles. The user is able visualise his/her aggregated

proﬁle and search for friends using keywords.

In our future work, we will increase the number of

supported OSNs and extend the FOAF-based model

to better describe users. Moreover, we plan to uti-

lize global ontologies like DBpedia and Wordnet as

a generic cross-domain interest model to enrich and

classify the user’s interests through different OSNs. It

could then be possible to develop more advanced per-

sonal recommendation applications in order to evalu-

ate the actual beneﬁts for end-users.

ACKNOWLEDGEMENTS

Part of this work has been developed in cooperation

with the 50A Company

who is funding this work.

REFERENCES

Abel, F., Gao, Q., Houben, G.-J., and Tao, K. (2011). An-

alyzing temporal dynamics in twitter proﬁles for per-

sonalized recommendations in the social web. In Pro-

ceedings of the ACM WebSci’11, Koblenz, Germany.

Abel, F., Henze, N., Herder, E., and Krause, D. (2010).

Linkage, aggregation, alignment and enrichment of

public user proﬁles with mypes. In Proceedings of the

6th International Conference on Semantic Systems,

number 11 in I-SEMANTICS ’10, pages 11:1–11:8,

New York, NY, USA. ACM.

Boyd, D. M. and Ellison, N. B. (2007). Social net-

work sites: Deﬁnition, history, and scholarship.

Journal of Computer-Mediated Communication. doi:

10.1111/j.1083-6101.2007.00393.x, 13:210–230.

Brickley, D. and Mille, L. (2005). Foaf vocabulary speciﬁ-

cation. http://xmlns.com/foaf/0.1/.

Kim, W., Jeong, O.-R., and Lee, S.-W. (2010). On social

web sites. Information System, 35(2):215–236. Spec

Section: Context-Oriented Information Integration.

Noor, S. and Martinez, K. (2009). Using social data as con-

text for making recommendations: an ontology based

approach. In Proceedings of the 1st Workshop on Con-

text, Information and Ontologies, CIAO’09, pages 1–

7, NY, USA. ACM.

Orlandi, F., Breslin, J., and Passant, A. (2012). Aggregated,

interoperable and multi-domain user proﬁles for the

social web. In Proceedings of the 8th International

Conference on Semantic Systems - I-SEMANTICS ’12,

page 41, New York, New York, USA. ACM Press.

Plumbaum, T., Wu, S., De Luca, E., and Albayrak, S.

(2011). User modeling for the social semantic web. In

Proceedings of the 2nd Workshop on Semantic Person-

alized Information Management: Retrieval and Rec-

ommendation, in conjunction with ISWC 2011.

Shapira, B., Rokach, L., and Freilikhman, S. (2012). Face-

book single and cross domain data for recommenda-

tion systems. User Modeling User-Adapted Interac-

tion.

Vu, T. and Baid, A. (2012). Ask, don’t search: A social

help engine for online social network mobile users. In

2012 35th IEEE Sarnoff Symposium, pages 1–5. Ieee.

Zhou, M., Zhang, W., Smith, B., Varga, E., Farias, M., and

Badenes, H. (2012). Finding someone in my social di-

rectory whom i do not fully remember or barely know.

In Proceedings of the 2012 ACM international confer-

ence on Intelligent User Interfaces, pages 203–206,

New York, NY, USA. ACM.

http://www.50a.fr/

WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies

476