Hello,
As mentioned in this page ( https://opennmt.net/OpenNMT/data/word_features/ ), can I use 3 additional features for target language, e.g., token | lemma | pos | case… and only token for source language OR vice versa? Does openNMT-py supports this for MT training? Then, how will be this be specified in onmt_preprocess commands and onmt_train commands?
Thanks in advance.
Word features are only supported on the source side for now.
See: