Research on Tibetan-Chinese Neural Machine Translation Based on GRU

The paper discusses the Tibetan-Chinese neural machine translation model based on LSTM and GRU, uses the Tibetan Dolma Tibetan word-splitting tool provided by the Chinese Academy of Sciences to split the Tibetan language as the input of the translation system, and analyzes it in comparison with the traditional recurrent neural network model. After experiments on the same Tibetan-Chinese dataset, it is shown that the BLEU value evaluated based on the LSTM NMT model is more than 5 BLEUS higher than that based on the traditional RNN NMT, and the BLEU value evaluated based on the GRU NMT model is more than 6 BLEUS higher than that based on the LSTM NMT.