EAGLE: Efficient Active Learning of Link Specifications Using Genetic Programming

With the growth of the Linked Data Web, time-efficient approaches for computing links between data sources have become indispensable. Most Link Discovery frameworks implement approaches that require two main computational steps. First, a link specification has to be explicated by the user. Then, this specification must be executed. While several approaches for the time-efficient execution of link specifications have been developed over the last few years, the discovery of accurate link specifications remains a tedious problem. In this paper, we present EAGLE, an active learning approach based on genetic programming. EAGLE generates highly accurate link specifications while reducing the annotation burden for the user. We present EAGLE and the framework within which it is implemented. We evaluate EAGLE against batch learning on three different data sets and show that it can detect specifications with an F-measure superior to 90% while requiring a small number of questions.