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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Continued improvements in machine learning techniques offer exciting newopportunities through the use of larger models and larger training datasets.However, there is a growing need to offer these new capabilities on-boardlow-powered devices such as smartphones, wearables and other embeddedenvironments where only low memory is available. Towards this, we considermethods to reduce the model size of Conformer-based speech recognition modelswhich typically require models with greater than 100M parameters down to just$5$M parameters while minimizing impact on model quality. Such a model allowsus to achieve always-on ambient speech recognition on edge devices withlow-memory neural processors. We propose model weight reuse at different levelswithin our model architecture: (i) repeating full conformer block layers, (ii)sharing specific conformer modules across layers, (iii) sharing sub-componentsper conformer module, and (iv) sharing decomposed sub-component weights afterlow-rank decomposition. By sharing weights at different levels of our model, wecan retain the full model in-memory while increasing the number of virtualtransformations applied to the input. Through a series of ablation studies andevaluations, we find that with weight sharing and a low-rank architecture, wecan achieve a WER of 2.84 and 2.94 for Librispeech dev-clean and test-cleanrespectively with a $5$M parameter model.

Sharing Low Rank Conformer Weights for Tiny Always-On Ambient Speech Recognition Models