Semantic Movie Recommendations - Smart Information Systems: Computational Intelligence for Real-Life Applications - page 143

and edges), there are no problems arising from the processing of natural language
text (such as handling of spelling mistakes and ambiguous queries). In addition, the
developed semantic system provides human readable explanations for suggestions
based on the sub-graph consider while computing the relevance of items.
The creation of a semantic recommender system is a complex process that
reveals several challenges: Starting from selecting and integrating appropriate knowl-
edge sources to aggregating heterogeneous data in a unified graph to learning
scenario-optimized recommender models able to cope with the complexity of data
and ensuring a fast response time. In this section we discussed different approaches
and showed that a semantic movie recommender system can be successfully learned
using the developed approach. Since semantic systems represent knowledge based on
graphs, the knowledge processing is independent from natural language descriptions.
This allows system designers separating natural language methods from the process-
ing of facts. Furthermore, the support for additional languages can be added by
integrating labels for new languages to the existing nodes.
Coming back to the initial scenario, the presented semantic movie recommender
system provides a powerful solution for the problems that Marc and Clara see in
traditional recommender systems. The semantic movie recommender system inte-
grates ratings and content-based knowledge provided by huge knowledge stores.
This allows the recommender system to consider fine-grained preferences about
favorite composers, actors, and producers. In addition, the recommender system can
suggest high-quality movies, still not known to everyone. By considering the age
classification of movies (retrieved from knowledge basis for the movie domain), the
parent's concerns are encountered that the recommended movies are suitable for the
children.
Summing up, the presented semantic recommender system allows us to overcome
the shortcomings of traditional recommender systems. The graph-based represen-
tation of knowledge enables the aggregation of different types of knowledge and
the integration of knowledge from many heterogeneous sources. Based on compre-
hensive knowledge graphs better recommendations can be computed considering
several different facets. This ensures highly useful, serendipitous recommendations.
Explanations computed based on the knowledge graph improve the transparency
of the recommendation process and the user's acceptance of the recommender
system.
Acknowledgments This research was supported by the Deutsche Forschungsgemeinschaft, DFG,
project number AL 561/11-1.
References
1. G. Adomavicius, A. Tuzhilin, Toward the next generation of recommender systems: a survey
of the state-of-the-art and possible extentions. IEEE Trans. Knowl. Data Eng. 17 (6) (2005)
2. S. Auer, C. Bizer, G. Kobilarov, J. Lehmann, R. Cyganiak, Z. Ives, DBpedia: a nucleus for a
web of open data, in The Semantic Web , Lecture Notes in Computer Science, vol. 4825, ed.