Boosting Speech-to-Text API accuracy in Google Cloud | C2C Community

Boosting Speech-to-Text API accuracy in Google Cloud

  • 14 September 2022
  • 2 replies
  • 8 views

Userlevel 7
Badge +13

Google cloud Speech -to-text support for 73 distinct languages, and 137 different local variants over 125 languages, the Speech-to-Text API allows you to quickly and accurately convert audio to text. In this video, Anu Shrivastava, Developer program engineer  walk you through the best tips and tricks for lowering your word error rate when using the Speech-to-Text API to transcribe audio files into text. Watch along to learn how you can boost your automated speech recognition accuracy without having to train your own custom model.

Click on the video below to watch it in detail:

 

Chapters:

0:00 - Intro

0:48 - What is Speech-to-Text API?

1:11 - Speech-to-Text API quickstart demo

1:47 - How do you measure output accuracy?

2:53 - Tips for checking accuracy at scale

3:49 - Improving accuracy with the Speech Adaption API

6:42 - Wrap up

 

Extra Credit


2 replies

Userlevel 7
Badge +56

​When I was working as a developer I used to play a lot with Speech-to-Text.

I hope to see, soon, Speech-to-Translated-Text in real time.

This would be great. Don't you think so @malamin?

Userlevel 7
Badge +13

Yes, it would be great @ilias . Speech-to-Translated-Text and Speech-to-Translate-Speech will change communication at different levels.

As an example:

I'm a patient speaking with a German doctor. We're both fluent in English. However, our accents differ, making communication difficult.

In real time.
When I speak with a German doctor, it will be automatically converted into German. It should be like the doctor's native language and accent that he used when speaking with his family and friends.

 

On the other hand, when a German doctor speaks to me, his voice changes to sound like mine. As a result, it will remove communication barriers and bring joy to many use cases.

 

Reply