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/

Zendy Plus

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

Zendy Tools

Zendy Open

Bootstrapping speech recognition on limited data resources has been an areaof active research for long. The recent transition to all-neural models andend-to-end (E2E) training brought along particular challenges as these modelsare known to be data hungry, but also came with opportunities aroundlanguage-agnostic representations derived from multilingual data as well asshared word-piece output representations across languages that share script androots. We investigate here the effectiveness of different strategies tobootstrap an RNN-Transducer (RNN-T) based automatic speech recognition (ASR)system in the low resource regime, while exploiting the abundant resourcesavailable in other languages as well as the synthetic audio from atext-to-speech (TTS) engine. Our experiments demonstrate that transfer learningfrom a multilingual model, using a post-ASR text-to-text mapping and syntheticaudio deliver additive improvements, allowing us to bootstrap a model for a newlanguage with a fraction of the data that would otherwise be needed. The bestsystem achieved a 46% relative word error rate (WER) reduction compared to themonolingual baseline, among which 25% relative WER improvement is attributed tothe post-ASR text-to-text mappings and the TTS synthetic data.

Bootstrap an end-to-end ASR system by multilingual training, transfer learning, text-to-text mapping and synthetic audio