English

Confidence through Attention

Computation and Language 2017-10-11 v1

Abstract

Attention distributions of the generated translations are a useful bi-product of attention-based recurrent neural network translation models and can be treated as soft alignments between the input and output tokens. In this work, we use attention distributions as a confidence metric for output translations. We present two strategies of using the attention distributions: filtering out bad translations from a large back-translated corpus, and selecting the best translation in a hybrid setup of two different translation systems. While manual evaluation indicated only a weak correlation between our confidence score and human judgments, the use-cases showed improvements of up to 2.22 BLEU points for filtering and 0.99 points for hybrid translation, tested on English<->German and English<->Latvian translation.

Keywords

Cite

@article{arxiv.1710.03743,
  title  = {Confidence through Attention},
  author = {Matīss Rikters and Mark Fishel},
  journal= {arXiv preprint arXiv:1710.03743},
  year   = {2017}
}
R2 v1 2026-06-22T22:09:13.827Z