Type Prediction in RDF Knowledge Bases Using Hierarchical Multilabel Classification
André Melo, Heiko Paulheim, Johanna Völker · 2016
Large Semantic Web knowledge bases are often noisy, incorrect, and incomplete with respect to type information. Automatic type prediction can help reduce such incompleteness, and, as previous works show, statistical methods are well-suited for this kind of data. Since most Semantic Web knowledge bases come with an ontology defining a type hierarchy, in this paper, we rephrase the type prediction problem as a hierarchical multilabel classification problem. We propose SLCN, a modification of the local classifier per node approach, which performs feature selection, instance sampling, and class balancing for each local classifier. Our approach improves scalability, facilitating its application on large Semantic Web datasets with high-dimensional feature and label spaces. We compare the performance of our proposed method with a state-of-the-art type prediction approach and popular hierarchical multilabel classifiers, and report on experiments with large-scale RDF datasets.