Document Type

Conference Proceeding

Publication Date



Recent developments in recommendation algorithms have focused on integrating Linked Open Data to augment traditional algorithms with background knowledge. These developments recognize that the integration of Linked Open Data may or better performance, particularly in cold start cases. In this paper, we explore if and how a specific type of Linked Open Data, namely hierarchical knowledge, may be utilized for recommendation systems. We propose a content-based recommendation approaches that adapts a spreading activation algorithm over the DBpedia category structure to identify entities of interest to the user. Evaluation of the algorithm over the Movielens dataset demonstrates that our method yields more accurate recommendations compared to a previously proposed taxonomy driven approach for recommendations.


Presented at the 4th Workshop on Knowledge Discovery and Data Mining Meets Linked Open Data, Portoroz, Slovenia, May 31, 2015.