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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Multilingual neural machine translation can translate unseen language pairsduring training, i.e. zero-shot translation. However, the zero-shot translationis always unstable. Although prior works attributed the instability to thedomination of central language, e.g. English, we supplement this viewpoint withthe strict dependence of non-centered languages. In this work, we propose asimple, lightweight yet effective language-specific modeling method by adaptingto non-centered languages and combining the shared information and thelanguage-specific information to counteract the instability of zero-shottranslation. Experiments with Transformer on IWSLT17, Europarl, TED talks, andOPUS-100 datasets show that our method not only performs better than strongbaselines in centered data conditions but also can easily fit non-centered dataconditions. By further investigating the layer attribution, we show that ourproposed method can disentangle the coupled representation in the correctdirection.

Adapting to Non-Centered Languages for Zero-shot Multilingual Translation