TPDL 2013: 272-278
Abstract. Despite the vast amount on works on personalizing keyword search on unstructured data (i.e. web pages), there is not much work done handling RDF data. In this paper we present our first cut approach on personalizing keyword query results on RDF data. We adopt the well known Ranking SVM approach, by training ranking functions with RDF-specific training features. The training utilizes historical user feedback, in the form of ratings on the searched items. In order to do so, we join netflix and dbpedia datasets, obtaining a dataset where we can simulate personalized search scenarios for a number of discrete users. Our evaluation shows that our approach outperforms the baseline and, in cases, it scores very close to the ground truth.