English

Cued@wmt19:ewc&lms

Computation and Language 2019-06-14 v1

Abstract

Two techniques provide the fabric of the Cambridge University Engineering Department's (CUED) entry to the WMT19 evaluation campaign: elastic weight consolidation (EWC) and different forms of language modelling (LMs). We report substantial gains by fine-tuning very strong baselines on former WMT test sets using a combination of checkpoint averaging and EWC. A sentence-level Transformer LM and a document-level LM based on a modified Transformer architecture yield further gains. As in previous years, we also extract nn-gram probabilities from SMT lattices which can be seen as a source-conditioned nn-gram LM.

Keywords

Cite

@article{arxiv.1906.05447,
  title  = {Cued@wmt19:ewc&lms},
  author = {Felix Stahlberg and Danielle Saunders and Adria de Gispert and Bill Byrne},
  journal= {arXiv preprint arXiv:1906.05447},
  year   = {2019}
}

Comments

WMT2019 system description (University of Cambridge)

R2 v1 2026-06-23T09:52:14.049Z