search

AI4Bharat / NPTEL2020-Indian-English-Speech-Dataset

NPTEL2020: Speech2Text dataset for Indian-English Accent
72 stars 20 forks source link

New SOTA result for NPTEL2020 #11

Open bigdatabaracus opened 3 years ago

bigdatabaracus commented 3 years ago

Dear people at NPTEL2020. We have ran the Pure Set againts our ASR api ( the www.speechly.com api). To our surplice we got the best WER result ( 0.2103 ) so far (at least from the list that is in the README) and we'd like to post it to the list in the README of this repo. How should we go about this? Can I do a PR?

Thank you very much!

CTO, Co-Founder, Speechly http://www.speechly.com

GokulNC commented 2 years ago

Hi @bigdatabaracus Thanks for reporting! Sorry for delayed response, did not notice this issue.

Inorder to ensure your results are reproducible, can you please help us with the following:

  1. Please send us a PR in this benchmarking repo for running your service: https://github.com/narVidhai/Speech-Transcription-Benchmarking
  2. To ensure that the comparison is consistent, since we do not have the API keys for Speechly to test it, can you please send us your transcripts to: gokulnc at ai4bharat dot org

Thanks!