![](bg2.png)
two phases: constructing an overlay model by us-
ing concepts and hierarchical information from ex-
ternal knowledge bases and creating links from the
constructed user model concepts to supported ontolo-
gies. The former phase outputs a semantically en-
hanced user model whereas the latter enables interop-
erability between applications which use the proposed
system for personalization. Moreover, fuzzy member-
ship values are computed for each interest and context
item in the user model.
The semantically enhanced user profile which is
enriched with fuzziness values are stored by utilizing
fuzzy hypergraph data structure. Fuzzy hypergraph
representation enables extraction of partial user pro-
file in the requested domains and output formats be-
sides answering user modeling queries such as the de-
gree of the user’s interest for the given concepts. By
extracting partial profiles by specifying domains, the
proposed system can be used for personalization pur-
poses in multi application environments.
2 RELATED WORK
In our study, we aim to exploit online traces of the
user on social networking and tagging environments
in order to construct the user model. Moreover, we
propose to mine the social web in a context-aware
manner and compute fuzziness values for the discov-
ered information about the individual during aggrega-
tion and semantic enrichment of partial profiles which
are obtained from different knowledge sources. The
constructed user model is able to extract partial user
profiles for specified domains in supported ontology
formats in order to provide personalization for multi
application environments. Therefore, our research
is related to cross system personalization, ubiquitous
user modeling process for multi application environ-
ments and fuzzy user modeling.
Cross system personalization is formulated in
(Mehta, 2009) and proved to be effective in cold start
problem in addition to providing a more robust user
profile. The nature of individual user profiles dis-
tributed on the social web is analyzed in (Abel et al.,
2011). In our study, we not only consider explicitly
stated form based information in social networks, but
also activities performed such as sharing or comment-
ing on a video about a certain topic and clicking the
‘like’ button on a sports team page etc. Moreover, we
consider check-in declarations on Facebook profiles
for current context of the user and events for his/her
possible future context. A generic user modeling li-
brary for the social semantic web which allows for
generating profiles that summarize the given stream
of messages according to domain and application spe-
cific requirements is proposed in (Gao et al., 2011).
Similarly, we aim to tailor the constructed user model
in accordance with the needs of the requester appli-
cations. Furthermore, we intend to manage whole life
cycle of an individual’s user model by considering not
only the construction of the profile but also the neces-
sary information updates to the profile.
In a multi-application environment, there are two
scenarios of constructing and consuming user pro-
files. In the first scenario, each application may con-
struct a partial user model and the challenge is reusing
built partial user models amongst applications. The
second scenario which we adopt, separates the user
model constructor and consumer applications. (Vi-
viani et al., 2010) classifies user modeling approaches
for multi application environments as standardization
based and mediation based user modeling. In stan-
dardization approaches, all participating applications
in the environment which are consuming the profile
are required to support the same user model. We pro-
pose a hybrid solution by constructing a user model
which is dynamically mapped to several well known
ontologies during construction phase. The proposed
user model is capable of exporting the required por-
tion of the profile partially in the form of the ontology
supported by the consumer application.
In (Kavcic, 2004), the uncertainty in the user’s
knowledge description is dealt with a fuzzy user
model in adapting educational hypermedia domain.
The uncertainty arises from vague boundaries be-
tween known and unknown concepts whereas in our
study the uncertainty is the problem of determin-
ing set memberships of the user profile items. In
(Vanekov and Vojts, 2009) partial preferences of the
user are combined by using monotone aggregation
function and stored in an ontology structure. How-
ever, in our system we are trying to determine the
confidence of the user profile item instead of setting a
preference ordering between profile items.
When the user model is semantically enhanced
and fuzzy membership values are taken into account,
more sophisticated user model structures are required,
since pairwise relations is not able to represent higher
order relations amongst concepts. (Ghoshal et al.,
2009) models folksonomies as tripartite graph struc-
tures.However, tripartite graphs are not able to rep-
resent relations with order 4 or higher. In order to
address this problem, (Tan et al., 2011) models higher
order relations in the social network as a unified hy-
pergraph and considers recommendation as a ranking
problem on the constructed hypergraph. Influenced
by this idea we employ unified fuzzy hypergraph(Roy
H. Goetschel, 1995) structure which is able to model
UCASFUM:AUbiquitousContext-awareSemanticFuzzyUserModelingSystem
279