While RDF data is graph shaped by nature, most traditional data mining and machine learning software expect data as vectors of features, each of which are either binary, numerical or nominal. In order to apply many of the existing data mining tools to the KG, an initial propositionalization of the corresponding graph is required.

The embedding is an effective yet efficient way to solve the graph propositionalization problem. It converts the graph data into a low dimensional space in which the graph structural information and graph properties are maximumly preserved.

We contributed to two embedding techniques. For each technique you will find an overview of the approach, which tests have been executed, information related to the datasets used during the tests and some notes. The precomputed embeddings can also be downloaded from these pages.

We also have an open source embedding evaluation framework