A Neural Machine Translation Model for Arabic Dialects That Utilizes Multitask Learning (MTL)

<div>The encoder-decoder architecture for the neural machine translation (NMT). Here, the source sentence A B C is translated into a target sentence W X Y Z. At each step, an evolving real-valued vector summarizes the state of the encoder (pink) and decoder (blue).</div>

Computational Intelligence and Neuroscience

fig1

Figure 1

Figure 1: A Neural Machine Translation Model for Arabic Dialects That Utilizes Multitask Learning (MTL) 