Compared with other neural encoders, such as BiLSTM and CNN, ... we use the pretrained sequence-to-sequence NMT models1 provided by fairseq for translation.
確定! 回上一頁