Supervising one Transformer attention head with attention-derived or GIZA++ alignments, with full target-sentence context for the alignment loss, improves word alignment accuracy without hurting translation BLEU.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Jointly Learning to Align and Translate with Transformer Models
Supervising one Transformer attention head with attention-derived or GIZA++ alignments, with full target-sentence context for the alignment loss, improves word alignment accuracy without hurting translation BLEU.