LexVec

We present a novel, minimally supervised method of generating word embedding evaluation datasets for a large number of languages. Our approach utilizes existing dependency treebanks and parsers in order to …...