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

Although modern automatic speech recognition (ASR) systems can achieve highperformance, they may produce errors that weaken readers' experience and doharm to downstream tasks. To improve the accuracy and reliability of ASRhypotheses, we propose a cross-modal post-processing system for speechrecognizers, which 1) fuses acoustic features and textual features fromdifferent modalities, 2) joints a confidence estimator and an error correctorin multi-task learning fashion and 3) unifies error correction and utterancerejection modules. Compared with single-modal or single-task models, ourproposed system is proved to be more effective and efficient. Experiment resultshows that our post-processing system leads to more than 10% relative reductionof character error rate (CER) for both single-speaker and multi-speaker speechon our industrial ASR system, with about 1.7ms latency for each token, whichensures that extra latency introduced by post-processing is acceptable instreaming speech recognition.

Cross-Modal ASR Post-Processing System for Error Correction and Utterance Rejection