English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Previous studies show that intermediate supervision signals benefit variousNatural Language Processing tasks. However, it is not clear whether there existintermediate signals that benefit Neural Machine Translation (NMT). Borrowingtechniques from Statistical Machine Translation, we propose intermediatesignals which are intermediate sequences from the "source-like" structure tothe "target-like" structure. Such intermediate sequences introduce an inductivebias that reflects a domain-agnostic principle of translation, which reducesspurious correlations that are harmful to out-of-domain generalisation.Furthermore, we introduce a full-permutation multi-task learning to alleviatethe spurious causal relations from intermediate sequences to the target, whichresults from exposure bias. The Minimum Bayes Risk decoding algorithm is usedto pick the best candidate translation from all permutations to further improvethe performance. Experiments show that the introduced intermediate signals caneffectively improve the domain robustness of NMT and reduces the amount ofhallucinations on out-of-domain translation. Further analysis shows that ourmethods are especially promising in low-resource scenarios.

Progressive Translation: Improving Domain Robustness of Neural Machine Translation with Intermediate Sequences