Fig. 1From: Deep neural networks for automatic speech processing: a survey from large corpora to limited dataIllustration of multi-task architecture. The output of the encoder is given to each decoder to produce the prediction for each ti taskBack to article page