0
$\begingroup$

So i have the task to study the feasability of setting up a Speech-To-Text engine in a production environnement, and i've been researching on this topic, so I tried Google's Speech-To-Text API and there is a quota on how many transcribtions you can use daily => That's a no-no. Second alternative, CMU Sphinx ( Open Source pretrained model for that purpose ), just tested it, and the output is messy for the LiveSpeech feature ( Perhaps i need to adjust some parameters first ). Third option, using the available deep learning tools to create my own SpeechRecognition engine ( I have tons of recorded conversations of people speaking French ).

Question : What should i use for my case? building an engine from scratch using neural networks seems complicated for a beginner in ML like me, using CMU Sphinx seems easy, but i have a feeling it's not. Any advice on what approach should i follow?

$\endgroup$
0
$\begingroup$

This would be a gift to the world. I certainly hope someone is working on it even as we speak (ha ha). Let's both keep looking.

$\endgroup$
  • 1
    $\begingroup$ Actually, that gift is being done by the guys at Mozilla. github.com/mozilla/DeepSpeech is an open source speech to text engine that is released in English, and other languages will be released as soon as they have enough data. $\endgroup$ – Blenz Aug 27 '19 at 8:29
  • $\begingroup$ you can either use their architecture to train your own model with your own vocals, or use their pre-trained models which have been benchmarked, and are working at levels near Google's and other giant tech companies. $\endgroup$ – Blenz Aug 27 '19 at 8:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.