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

Speech recognition models often obtain degraded performance when tested onspeech with unseen accents. Domain-adversarial training (DAT) and multi-tasklearning (MTL) are two common approaches for building accent-robust ASR models.ASR models using accent embeddings is another approach for improving robustnessto accents. In this study, we perform systematic comparisons of DAT and MTLapproaches using a large volume of English accent corpus (4000 hours of USEnglish speech and 1244 hours of 20 non-US-English accents speech). We exploreembeddings trained under supervised and unsupervised settings: a separateembedding matrix trained using accent labels, and embeddings extracted from afine-tuned wav2vec model. We find that our DAT model trained with supervisedembeddings achieves the best performance overall and consistently providesbenefits for all testing datasets, and our MTL model trained with wav2vecembeddings are helpful learning accent-invariant features and improvingnovel/unseen accents. We also illustrate that wav2vec embeddings have moreadvantages for building accent-robust ASR when no accent labels are availablefor training supervised embeddings.

Accent-Robust Automatic Speech Recognition Using Supervised and Unsupervised Wav2vec Embeddings