Health Assistance for Immigrants - Smart Information Systems: Computational Intelligence for Real-Life Applications - page 84

Fig. 3.2 The GID health ontology
3.3.2 User Model
As explained above, personalization services rely on the creation of user models
to store personal information in a user profile. User profiles are then exploited to
adapt information accordingly. In the health scenario, the most important factors are
the users' demographic details (i.e., age, sex) and their personal context (e.g., their
hometown, language knowledge and pre-existing medical conditions). In order to
receive personalized health information, users of the GID system are required to
provide the above information.
3.3.3 Multilingual Semantic Search
As explained above, concepts in the health ontology are enriched with multilingual
labels. When a user enters a search query, we process the entered text by lowercas-
ing it and pruning unusual characters. The search terms are then used as a query
to retrieve matching concept labels contained in the Health Ontology. We do not
perform language-specific stemming or remove stop-words of the search input since
this enables us to easily extend the GID system with additional languages without
requiring any changes to program code. Rather, we use a fuzzy search based on the
open-source information retrieval system Lucene 7 to cover slight alterations in term
surface forms. Taking the multilingual search query “ Hamilelikte hangi besinleri
yememeliyim? ” (Translation: “ Which food must I not eat when I am pregnant? ”)
7 http://lucene.apache.org/ .