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

Self-supervised learning (SSL) is a long-standing goal for speech processing,since it utilizes large-scale unlabeled data and avoids extensive humanlabeling. Recent years witness great successes in applying self-supervisedlearning in speech recognition, while limited exploration was attempted inapplying SSL for modeling speaker characteristics. In this paper, we aim toimprove the existing SSL framework for speaker representation learning. Twomethods are introduced for enhancing the unsupervised speaker informationextraction. First, we apply the multi-task learning to the current SSLframework, where we integrate the utterance-wise contrastive loss with the SSLobjective function. Second, for better speaker discrimination, we propose anutterance mixing strategy for data augmentation, where additional overlappedutterances are created unsupervisely and incorporate during training. Weintegrate the proposed methods into the HuBERT framework. Experiment results onSUPERB benchmark show that the proposed system achieves state-of-the-artperformance in universal representation learning, especially for speakeridentification oriented tasks. An ablation study is performed verifying theefficacy of each proposed method. Finally, we scale up training dataset to 94thousand hours public audio data and achieve further performance improvement inall SUPERB tasks.

UniSpeech-SAT: Universal Speech Representation Learning with Speaker Aware Pre-Training