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 pre-training methods based on contrastive learning orregression tasks can utilize more unlabeled data to improve the performance ofautomatic speech recognition (ASR). However, the robustness impact of combiningthe two pre-training tasks and constructing different negative samples forcontrastive learning still remains unclear. In this paper, we propose anoise-robust data2vec for self-supervised speech representation learning byjointly optimizing the contrastive learning and regression tasks in thepre-training stage. Furthermore, we present two improved methods to facilitatecontrastive learning. More specifically, we first propose to constructpatch-based non-semantic negative samples to boost the noise robustness of thepre-training model, which is achieved by dividing the features into patches atdifferent sizes (i.e., so-called negative samples). Second, by analyzing thedistribution of positive and negative samples, we propose to remove the easilydistinguishable negative samples to improve the discriminative capacity forpre-training models. Experimental results on the CHiME-4 dataset show that ourmethod is able to improve the performance of the pre-trained model in noisyscenarios. We find that joint training of the contrastive learning andregression tasks can avoid the model collapse to some extent compared to onlytraining the regression task.

Robust Data2vec: Noise-robust Speech Representation Learning for ASR by Combining Regression and Improved Contrastive Learning