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

Non-autoregressive translation (NAT) models generate multiple tokens in oneforward pass and is highly efficient at inference stage compared withautoregressive translation (AT) methods. However, NAT models often suffer fromthe multimodality problem, i.e., generating duplicated tokens or missingtokens. In this paper, we propose two novel methods to address this issue, theLook-Around (LA) strategy and the Vocabulary Attention (VA) mechanism. TheLook-Around strategy predicts the neighbor tokens in order to predict thecurrent token, and the Vocabulary Attention models long-term token dependenciesinside the decoder by attending the whole vocabulary for each position toacquire knowledge of which token is about to generate. %We also propose adynamic bidirectional decoding approach to accelerate the inference process ofthe LAVA model while preserving the high-quality of the generated output. Ourproposed model uses significantly less time during inference compared withautoregressive models and most other NAT models. Our experiments on fourbenchmarks (WMT14 En$\rightarrow$De, WMT14 De$\rightarrow$En, WMT16Ro$\rightarrow$En and IWSLT14 De$\rightarrow$En) show that the proposed modelachieves competitive performance compared with the state-of-the-artnon-autoregressive and autoregressive models while significantly reducing thetime cost in inference phase.

LAVA NAT: A Non-Autoregressive Translation Model with Look-Around Decoding and Vocabulary Attention