KBLRN : End-to-End Learning of Knowledge Base Representations with\n Latent, Relational, and Numerical Features

Abstract

We present KBLRN, a framework for end-to-end learning of knowledge base\nrepresentations from latent, relational, and numerical features. KBLRN\nintegrates feature types with a novel combination of neural representation\nlearning and probabilistic product of experts models. To the best of our\nknowledge, KBLRN is the first approach that learns representations of knowledge\nbases by integrating latent, relational, and numerical features. We show that\ninstances of KBLRN outperform existing methods on a range of knowledge base\ncompletion tasks. We contribute a novel data sets enriching commonly used\nknowledge base completion benchmarks with numerical features. The data sets are\navailable under a permissive BSD-3 license. We also investigate the impact\nnumerical features have on the KB completion performance of KBLRN.\n

Ähnliche Arbeiten