Abstract: We present a web platform for Automatic Speech Recognition built on top of Kaldi toolkit, which supports both batch and on-line recognition and has an annotation interface for transcription of the gathered speech recordings. By providing such service we want to offer the developers an easy way to use voice control in their applications. Also we want to collect and publish new speech data for further research.
I will also mention how we deploy and run the CloudASR platform in a robust and scalable way using technology similar to how Google and Facebook run their software platforms. Demo of the platform is available at demo.cloudasr.com.