English Classical Translation under the Knowledge Difference Based on Transformer

When performing machine translation on poems in English ancient literature and classics, this paper adopted a two-stage translation method. In the first stage, the Transformer model was used to translate English poems into vernacular translations. In the second stage, an encoder and a decoder constructed with a long short-term memory (LSTM) were used to convert the vernacular translations into Chinese poems. Meanwhile, a back-translation strategy was adopted when training the encoder and decoder in the second stage. After that, simulation experiments were carried out. In the experiments, the two-stage algorithm was compared with the multilingual bidirectional and auto-regressive transformers (mBART), traditional LSTM, and traditional Transformer models. The findings suggest that the translation algorithm can accurately translate English poetry and align the translated text more closely with the stylistic characteristics of Chinese poetry.