Personalized Information Access Using Semantic Knowledge - Smart Information Systems: Computational Intelligence for Real-Life Applications - page 206

user profile and the therein-specified user taste, because the Freebase data contains
general domain information, and for users with a common taste more or less uni-
versal knowledge is added. Adding the artist “Madonna” does not make sense for
user profiles already containing a lot of pop artists; it only leads to more general pro-
files less tailored to individual user preferences. Different strategies to overcome this
problem are conceivable. On the one hand, our approach needs to weight the edge
types in a more user-centric way. A user may like an artist because of a certain song
but does not like the complete discography; or a user might like the artist because of
the social engagement of that artist and not because of the music. Therefore, more
contextual information about users is needed, enabling a context-sensitive weight-
ing of the information used for the profile enrichment. The increasing popularity of
Social Semantic Web approaches and standards like FOAF 9 can be one important
step in this direction [ 8 , 9 ]. On the other hand, semantic datasets themselves have
to be enriched with more meta-information about the data. General quality and sig-
nificance information like prominence nodes and weighted relations can improve
semantic algorithms to better compute the importance of paths between nodes. An
artist that made hundreds of bad albums may have a high number of links to, e.g.,
a genre node, but is not an important artist for this genre while another artist made
only one or two albums but defined a genre. In this case, a significant weightage for
the artists can improve the quality and performance of semantic algorithms.
Future steps are the evaluation of a focused enrichment, e.g., only using artist
or genres information, based on the context of the user. Another direction is to
implement an advanced weighting model (e.g., based on prominence, context, user
groups or interaction time (e.g., [ 11 ])) as an overlay for the Freebase dataset, and to
implement alternative network models (e.g., based on a low-rank approximation for
the adjacency matrix of a relationship set [ 18 ]).

Acknowledgments This work was funded by the Federal Ministry of Economic Affairs and Energy
(BMWi) under funding reference number KF2392305KM0.

References

1. G. Adomavicius, A. Tuzhilin, Toward the next generation of recommender systems: a survey
of the state-of-the-art and possible extensions. Trans. Knowl. Data Eng. 17 , 734-749 (2005)
2. S.S. Anand, P. Kearney, M. Shapcott, Generating semantically enriched user profiles for
webpersonalization. ACM Trans. Internet Technol. 7 , 26 (2007)
3. S.S. Anand, B. Mobasher, Introduction to intelligent techniques for web personalization. ACM
Trans. Internet Technol. 7 , 4 (2007)
4. S. Berkovsky, T. Kuflik, F. Ricci, Mediation of user models for enhanced personalization in
recommender systems. User Model. User-Adapt. Interact. 18 (3), 245-286 (2008)
5. S. Berkovsky, T. Kuflik, F. Ricci, Cross-representation mediation of user models. User Model.
User-Adapt. Interact. 19 (1-2), 35-63 (2009)
6. C. Bizer, T. Heath, T. Berners-Lee, Linked data—the story so far. Int. J. Semant. Web Inf. Syst.
5 (3), 1-22 (2009)

9

http://www.foaf-project.org/ .