Word Alignment Based Transformer Model for XML Structured Documentation Translation

In the context of globalization, the development of localized translation technology for enterprise online documents is crucial for business promotion. The enterprise online documents are represented by semi-structured text documents with markup tags, while the mainstream neural machine translation methods focus on only the plain text translation. In this research, a Word Alignment based Transformer Model was proposed for markup language translation. Experiments conducted on the Salesforce XML English-Chinese datasets, and the result demonstrated that adding a word alignment model to the translation model can improve the translation model's performance in translating text with makup tags.