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

This article describes an efficient end-to-end speech translation (E2E-ST)framework based on non-autoregressive (NAR) models. End-to-end speechtranslation models have several advantages over traditional cascade systemssuch as inference latency reduction. However, conventional AR decoding methodsare not fast enough because each token is generated incrementally. NAR models,however, can accelerate the decoding speed by generating multiple tokens inparallel on the basis of the token-wise conditional independence assumption. Wepropose a unified NAR E2E-ST framework called Orthros, which has an NAR decoderand an auxiliary shallow AR decoder on top of the shared encoder. The auxiliaryshallow AR decoder selects the best hypothesis by rescoring multiple candidatesgenerated from the NAR decoder in parallel (parallel AR rescoring). We adoptconditional masked language model (CMLM) and a connectionist temporalclassification (CTC)-based model as NAR decoders for Orthros, referred to asOrthros-CMLM and Orthros-CTC, respectively. We also propose two trainingmethods to enhance the CMLM decoder. Experimental evaluations on threebenchmark datasets with six language directions demonstrated that Orthrosachieved large improvements in translation quality with a very small overheadcompared with the baseline NAR model. Moreover, the Conformer encoderarchitecture enabled large quality improvements, especially for CTC-basedmodels. Orthros-CTC with the Conformer encoder increased decoding speed by3.63x on CPU with translation quality comparable to that of an AR model.

Non-autoregressive End-to-end Speech Translation with Parallel Autoregressive Rescoring