Phonetically-Aware Embeddings, Wide Residual Networks with Time-Delay Neural Networks and Self Attention Models for the 2018 NIST Speaker Recognition Evaluation | Publicación