Time Convolutional Network-Transformer based Chinese English Translation Model

The growth of Neural Machine Translation (NMT) by utilizing neural networks has progressively become typical method in industry and academia. The previous ChineseEnglish NMT models usually choose neural network architecture based on attention mechanism. However, it faces challenging issue to existing model for learning and translating from one language to other language and vice versa simultaneously. To overcome this problem a Time Convolutional Network-Transformer (TCN-Transformer) method is proposed for improving end-to-end Chinese-English translation. The dataset utilized for research is obtained from GitHub and back translation technique is performed to prevent meaningless sentence while reverse translation. After that, a Seq2Seq model and an improved transformer is utilized for endto end translation of both Chinese and English sentences. The experimental results stipulate the better translation of words efficiently with $95.4 \%$ of precision, $95.3 \%$ of accuracy, $95.3 \%$ of Recall and $42.4 \%$ of BLEU which is superior than existing methods like Transformer, Two-way Gated Recurrent Unit (Two-way GRU), WD-Transformer, Self-Attention-Transformer (SA-Transformer) and Random Decision Forest (RDF).