Pointer-Generator Networks for Low-Resource Machine Translation: Don't Copy That!

arXiv:2403.10963v1 Announce Type: new
Abstract: While Transformer-based neural machine translation (NMT) is very effective in high-resource settings, many languages lack the necessary large parallel corpora to benefit from it. In the context of low-resource (LR) MT between two closely-related languages, a natural intuition is to seek benefits from structural "shortcuts", such as copying subwords from the source to the target, given that such language pairs often share a considerable number of identical words, cognates, and borrowings. We test Pointer-Generator Networks …

abstract arxiv benefit benefits context copy cs.cl generator intuition languages low machine machine translation natural networks neural machine translation transformer translation type