Mask-Align
Mask-Align copied to clipboard
Operation about bpe
Did you apply bpe in your train data? What do you mean by "We used a joint source and target Byte Pair Encoding (BPE) (Sennrich et al., 2016) with 40k merge operations." in your artical sector 3.1?
I tried bpe training and the result is not as satisfactory as original corpus. Have you tried this before and do you have some comments about this?