Health Assistance for Immigrants - Smart Information Systems: Computational Intelligence for Real-Life Applications - page 81

person differ significantly and thus should be reflected in the information that is
tailored to the users' needs.
Tailoring information based on these user contexts requires storage of user-centric
information. In order to easily connect this information with ontology-based knowl-
edge bases, i.e., for providing a personalized healthcare system, a promising approach
is to model information using user-centric (semantic) ontologies. Semantic-based
user ontologies are for instance:

￿ FOAF: The Friend-of-a-Friendmodel is anRDF-based usermodelmainly designed

for the web. It defines demographic data such as name, age, and friendship rela-
tions. Common applications of FOAF are social networks services [ 3 ].

￿ SWUM: The Semantic Web User Model defines a comprehensive user model

designed for the needs of the modern social semantic web. It models information
about the users' demographics, friendship, etc. [ 31 ].

￿ GUMO: The General User Modelling Ontology tries to provide a user model

covering all aspects of life. GUMO models health related information such as
blood pressure or temperature [ 15 ].

In this work, we introduce an approach to map users' demands (as expressed by
the search query) with the computer-based knowledge basis under consideration of
the context that is stored within a semantic-based user ontology, thus presenting a
healthcare information system that can provide answers based on context.

3.2.4 Multilingual Semantic Information Management

As argued above, onemain challenge in the described scenario is to deal with different
languages that might be relevant to provide health-oriented information, e.g., the
difference between the patients' mother tongue ( Language A ) and the language used
by the physician ( Language B ). Multilingual information management has been
studied extensively in the literature. A straightforward approach to cross-lingual
information management is either a direct translation approach [ 27 ]orusinganinter-
lingual mapping like EuroWordNet [ 43 ], which have been shown to work well at the
CLIR task [ 6 ]. Imprecise translations are acceptable for the retrieval performance,
as the document search itself is more important than disambiguation of individual
translated terms [ 16 ]. Other approaches map individual documents into a higher
dimension semantic feature space that is uniform for different languages. Thus, it
is possible to map similar documents to nearby points in the feature space, even
if they do not share a language. Sorg and Cimiano [ 34 ] have exploited the explicit
links between related Wikipedia articles of different languages to map documents
to a Wikipedia feature space in which documents are considered similar when they
are semantically similar to the same set of articles. A similar approach is to map
documents to multilingual ontology concepts, which can be represented as points in
a feature vector space [ 12 ]. In this work, we employ such ontology mapping method
to link concepts expressed in different languages.