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

Automatic speech recognition (ASR) models with low-footprint are increasinglybeing deployed on edge devices for conversational agents, which enhancesprivacy. We study the problem of federated continual incremental learning forrecurrent neural network-transducer (RNN-T) ASR models in the privacy-enhancingscheme of learning on-device, without access to ground truth human transcriptsor machine transcriptions from a stronger ASR model. In particular, we studythe performance of a self-learning based scheme, with a paired teacher modelupdated through an exponential moving average of ASR models. Further, wepropose using possibly noisy weak-supervision signals such as feedback scoresand natural language understanding semantics determined from user behavioracross multiple turns in a session of interactions with the conversationalagent. These signals are leveraged in a multi-task policy-gradient trainingapproach to improve the performance of self-learning for ASR. Finally, we showhow catastrophic forgetting can be mitigated by combining on-device learningwith a memory-replay approach using selected historical datasets. Theseinnovations allow for 10% relative improvement in WER on new use cases withminimal degradation on other test sets in the absence of strong-supervisionsignals such as ground-truth transcriptions.

Federated Self-Learning with Weak Supervision for Speech Recognition