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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Error correction models form an important part of Automatic SpeechRecognition (ASR) post-processing to improve the readability and quality oftranscriptions. Most prior works use the 1-best ASR hypothesis as input andtherefore can only perform correction by leveraging the context within onesentence. In this work, we propose a novel N-best T5 model for this task, whichis fine-tuned from a T5 model and utilizes ASR N-best lists as model input. Bytransferring knowledge from the pre-trained language model and obtaining richerinformation from the ASR decoding space, the proposed approach outperforms astrong Conformer-Transducer baseline. Another issue with standard errorcorrection is that the generation process is not well-guided. To address this aconstrained decoding process, either based on the N-best list or an ASRlattice, is used which allows additional information to be propagated.

N-best T5: Robust ASR Error Correction using Multiple Input Hypotheses and Constrained Decoding Space