Autoencoder-Speaker-Embeddings We find audio feature vectors using a variational autoencoder, upon which we apply speaker/language classifier along with domain adaptation to unlearn unnecessary features