This repository is the pytorch implementation for the paper: Go From the General to the Particular: Multi-Domain Translation with Domain Transformation Networks.

The project is based on the fairseq. (Please get familar with the fairseq project first)

• PyTorch version >= 1.0.0
• Python version 3.6

First, preprocess your training corpus. Use BPE (byte-pair-encoding) to segment text into subword units and please follow https://github.com/rsennrich/subword-nmt for further details. In addition, please place the validation and test sets of respective domains to the directory valid_test_txt, which is included in the directory of binary files. For training and inference, please refer to our paper and scripts for more details.

For any questions, please email to the first author.