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

Multi-talker speech recognition (MT-ASR) has been shown to improve ASRperformance on speech containing overlapping utterances from more than onespeaker. Multi-talker models have typically been trained from scratch usingsimulated or actual overlapping speech datasets. On the other hand, the trendin ASR has been to train foundation models using massive datasets collectedfrom a wide variety of task domains. Given the scale of these models and theirability to generalize well across a variety of domains, it makes sense toconsider scenarios where a foundation model is augmented with multi-talkercapability. This paper presents an MT-ASR model formed by combining awell-trained foundation model with a multi-talker mask model in a cascadedRNN-T encoder configuration. Experimental results show that the cascadeconfiguration provides improved WER on overlapping speech utterances withrespect to a baseline multi-talker model without sacrificing performanceachievable by the foundation model on non-overlapping utterances.

Cascaded encoders for fine-tuning ASR models on overlapped speech