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

Despite huge improvements in automatic speech recognition (ASR) employingneural networks, ASR systems still suffer from a lack of robustness andgeneralizability issues due to domain shifting. This is mainly becauseprincipal corpus design criteria are often not identified and examinedadequately while compiling ASR datasets. In this study, we investigate therobustness of the state-of-the-art transfer learning approaches such asself-supervised wav2vec 2.0 and weakly supervised Whisper as well as fullysupervised convolutional neural networks (CNNs) for multi-domain ASR. We alsodemonstrate the significance of domain selection while building a corpus byassessing these models on a novel multi-domain Bangladeshi Bangla ASRevaluation benchmark - BanSpeech, which contains approximately 6.52 hours ofhuman-annotated speech and 8085 utterances from 13 distinct domains. SUBAK.KO,a mostly read speech corpus for the morphologically rich language Bangla, hasbeen used to train the ASR systems. Experimental evaluation reveals thatself-supervised cross-lingual pre-training is the best strategy compared toweak supervision and full supervision to tackle the multi-domain ASR task.Moreover, the ASR models trained on SUBAK.KO face difficulty recognizing speechfrom domains with mostly spontaneous speech. The BanSpeech will be publiclyavailable to meet the need for a challenging evaluation benchmark for BanglaASR.

Investigating self-supervised, weakly supervised and fully supervised  training approaches for multi-domain automatic speech recognition: a study on  Bangladeshi Bangla