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

End-to-end (E2E) automatic speech recognition models like Recurrent NeuralNetworks Transducer (RNN-T) are becoming a popular choice for streaming ASRapplications like voice assistants. While E2E models are very effective atlearning representation of the training data they are trained on, theiraccuracy on unseen domains remains a challenging problem. Additionally, thesemodels require paired audio and text training data, are computationallyexpensive and are difficult to adapt towards the fast evolving nature ofconversational speech. In this work, we explore a contextual biasing approachusing likelihood-ratio that leverages text data sources to adapt RNN-T model tonew domains and entities. We show that this method is effective in improvingrare words recognition, and results in a relative improvement of 10% in 1-bestword error rate (WER) and 10% in n-best Oracle WER (n=8) on multipleout-of-domain datasets without any degradation on a general dataset. We alsoshow that complementing the contextual biasing adaptation with adaptation of asecond-pass rescoring model gives additive WER improvements.

A Likelihood Ratio based Domain Adaptation Method for E2E Models