Corpus-based Comparison of Distributional Models of Language and Knowledge Graphs