Pluggable Neural Machine Translation Models via Memory-augmented Adapters

arXiv:2307.06029v2 Announce Type: replace
Abstract: Although neural machine translation (NMT) models perform well in the general domain, it remains rather challenging to control their generation behavior to satisfy the requirement of different users. Given the expensive training cost and the data scarcity challenge of learning a new model from scratch for each user requirement, we propose a memory-augmented adapter to steer pretrained NMT models in a pluggable manner. Specifically, we construct a multi-granular memory based on the user-provided text samples …

abstract arxiv behavior challenge control cost cs.cl data domain general machine machine translation memory neural machine translation scratch training translation type via