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

Undirected neural sequence models such as BERT (Devlin et al., 2019) havereceived renewed interest due to their success on discriminative naturallanguage understanding tasks such as question-answering and natural languageinference. The problem of generating sequences directly from these models hasreceived relatively little attention, in part because generating fromundirected models departs significantly from conventional monotonic generationin directed sequence models. We investigate this problem by proposing ageneralized model of sequence generation that unifies decoding in directed andundirected models. The proposed framework models the process of generationrather than the resulting sequence, and under this framework, we derive variousneural sequence models as special cases, such as autoregressive,semi-autoregressive, and refinement-based non-autoregressive models. Thisunification enables us to adapt decoding algorithms originally developed fordirected sequence models to undirected sequence models. We demonstrate this byevaluating various handcrafted and learned decoding strategies on a BERT-likemachine translation model (Lample & Conneau, 2019). The proposed approachachieves constant-time translation results on par with linear-time translationresults from the same undirected sequence model, while both are competitivewith the state-of-the-art on WMT'14 English-German translation.

A Generalized Framework of Sequence Generation with Application to Undirected Sequence Models