Abstract
In this paper, we present the LODBookRec application. LODBookRec builds on top of Linked Open Data (LOD) knowledge about literature domain and provides information retrieval GUI to conveniently present this knowledge to the end users.
As such, LODBookRec aims to contribute towards better utilization of LOD and provide a suitable platform for on-line evaluation of information retrieval methods, especially recommender systems. LODBookRec contains a basic search GUI and several recommendation methods, with the primary focus on item-based recommendations.
The results of offline evaluation indicates that content-based recommendations utilizing attribute-based similarity of books provides best item-based recommendations w.r.t. recommendation relevance for both highly popular books as well as long-tail books. However, modification of the original algorithm via maximal margin relevance increases diversity of the recommended lists with a modest relevance penalties.