Rotations and Interpretability of Word Embeddings: the Case of the Russian Language

Alexey Zobnin

Consider a continuous word embedding model. Usually, the cosines between word vectors are used as a measure of similarity of words. These cosines do not change under orthogonal tra…