Walk Extraction Strategies for Node Embeddings with RDF2Vec in Knowledge Graphs

As KGs are symbolic constructs, specialized techniques have to be applied in\norder to make them compatible with data mining techniques. RDF2Vec is an\nunsupervised technique that can create task-agnostic numerical representations\nof the nodes in a KG by extending successful language modelling techniques. The\noriginal work proposed the Weisfeiler-Lehman (WL) kernel to improve the quality\nof the representations. However, in this work, we show both formally and\nempirically that the WL kernel does little to improve walk embeddings in the\ncontext of a single KG. As an alternative to the WL kernel, we propose five\ndifferent strategies to extract information complementary to basic random\nwalks. We compare these walks on several benchmark datasets to show that the\n\\emph{n-gram} strategy performs best on average on node classification tasks\nand that tuning the walk strategy can result in improved predictive\nperformances.\n