English to Urdu: Optimizing Sequence Learning in Neural Machine Translation

Neural machine translation is a new approach for machine translation, which translates source language sentences into the target language. Source to target words mapping is the key task of attention mechanism in neural machine translation. The number of attention-based seq2seq encoder-decoder models have been proposed in the literature. But these models have a nonlinearity problem with attention in English to Urdu translation. A seq2seq encoder-decoder model named Convolutional English to Urdu Translation (CEUT) has been proposed in this paper. The model reduces nonlinearity in attention word mapping and achieves 29.94 BLEU score in English to Urdu translation.