Soft alignment

Appears in 1 paper

Attention's approach: each target word is generated using a *weighted blend* of multiple source words, not a hard assignment to one.

As used in Paper 07 — Neural Machine Translation by Jointly Learning to Align and Translate →

Attention's approach: each target word is generated using a weighted blend of multiple source words, not a hard assignment to one. More expressive than hard alignment, can represent genuine ambiguity (one English phrase mapping to multiple French words with shared influence).