Deep Learning-Based Automatic Cultural Translation Method for English Tourism

The general LSTM-based encoder-decoder model has the problems of not being able to mine the sentence semantics and translate long text sequences. This study presents a neural machine translation model utilizing LSTM with improved attention, incorporating multi-head attention and multi-skipping attention mechanisms into the LSTM baseline model. By adding multi-head attention computation, the syntactic information in different subspaces can be mined, and then attention can be paid to the semantic information in the sentence sequences, and then multiple attentions are computed on each head separately, which can effectively deal with the long-distance dependency problem and perform better in the translation of long sentences. The proposed model is analysed and compared using the WMT17 Chinese and English datasets, newsdev2017 and newstest2017, and the results show that the proposed model improves the BLEU score of the automatic translation of Tourism English Culture and solves the problem of low scores in long sentence translation.

@article{Liu2025,
  title     = {Deep Learning-Based Automatic Cultural Translation Method for English Tourism},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {16},
  number    = {1},
  year      = {2025},
  publisher = {The Science and Information Organization},
  author    = {Jianguo Liu and Ruohan Liu},
  doi       = {10.14569/IJACSA.2025.0160162},
  url       = {https://doi.org/10.14569/IJACSA.2025.0160162}
}