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

Joint optimization of multi-channel front-end and automatic speechrecognition (ASR) has attracted much interest. While promising results havebeen reported for various tasks, past studies on its meeting transcriptionapplication were limited to small scale experiments. It is still unclearwhether such a joint framework can be beneficial for a more practical setupwhere a massive amount of single channel training data can be leveraged forbuilding a strong ASR back-end. In this work, we present our investigation onthe joint modeling of a mask-based beamformer andAttention-Encoder-Decoder-based ASR in the setting where we have 75k hours ofsingle-channel data and a relatively small amount of real multi-channel datafor model training. We explore effective training procedures, including acomparison of simulated and real multi-channel training data. To guide therecognition towards a target speaker and deal with overlapped speech, we alsoexplore various combinations of bias information, such as direction of arrivalsand speaker profiles. We propose an effective location bias integration methodcalled deep concatenation for the beamformer network. In our evaluation onvarious meeting recordings, we show that the proposed framework achieves asubstantial word error rate reduction.

Exploring End-to-End Multi-channel ASR with Bias Information for Meeting Transcription