Unsupervised learning of link specifications: deterministic vs. non-deterministic

Link Discovery has been shown to be of utter importance for the Linked Data Web. In previous works, several supervised approaches have been developed for learning link specifications out of labelled data. Most recently, genetic programming has also been utilized to learn link specifications in an unsupervised fashion by optimizing a parametrized pseudo-F-measure. The questions underlying this evaluation paper are twofold: First, how well do pseudo-F-measures predict the real accuracy of non-deterministic and deterministic approaches across different types of datasets? Second, how do deterministic approaches compare to non-deterministic approaches? To answer these questions, we evaluated linear and Boolean classifiers against classifiers computed by using genetic programming on six different data sets. We also studied the correlation between two different pseudo-F-measures and the real F-measures achieved by the classifiers at hand. Our evaluation suggests that pseudoF-measures behave differently on the synthetic and real data sets.