Subject-object information in word embeddings (subj-obj-embeds)

This set of datasets was initially built for the experiment to find out wether embeddings encoded the information suggesting a word is more likely to be the subject or object. The datasets were extracted from German TüBa-D/Z treebank with UD and Hamburg dependency annotations (for German) and the Lassy small treebank (for Dutch).