Affiliations: Corporate Technology, Siemens AG, Otto-Hahn-Ring 6, 81739 Munich, Germany. E-mail: {yihuang,volker.tresp}@siemens.com | Department of Computer Science, Ludwig Maximilian University, Oettingenstraße 67, 80538 Munich, Germany. E-mail: {nickel,kriegel}@dbs.ifi.lmu.de | Karlsruhe Institute of Technology, Kaiserstraße 12, 76131 Karlsruhe, Germany. E-mail: rettinger@kit.de
Note: [] Corresponding author.
Abstract: Increasingly, data is published in the form of semantic graphs. The most notable example is the Linked Open Data (LOD) initiative where an increasing number of data sources are published in the Semantic Web's Resource Description Framework and where the various data sources are linked to reference one another. In this paper we apply machine learning to semantic graph data and argue that scalability and robustness can be achieved via an urn-based statistical sampling scheme. We apply the urn model to the SUNS framework which is based on multivariate prediction. We argue that multivariate prediction approaches are most suitable for dealing with the resulting high-dimensional sparse data matrix. Within the statistical framework, the approach scales up to large domains and is able to deal with highly sparse relationship data. We summarize experimental results using a friend-of-a-friend data set and a data set derived from DBpedia. In more detail, we describe novel experiments on disease gene prioritization using LOD data sources. The experiments confirm the ease-of-use, the scalability and the good performance of the approach.
Keywords: Statistical machine learning, Linked Open Data, Semantic Web, statistical relational learning, linked life data
DOI: 10.3233/SW-130100
Journal: Semantic Web, vol. 5, no. 1, pp. 5-22, 2014
