Variational Disentangled Attention and Regularization for Visual Dialog | IEEE Conference Publication | IEEE Xplore