There is Microsoft's Azure speech-to-text API: https://azure.microsoft.com/en-us/services/cognitive-services/speech-to-text/ https://azure.microsoft.com/en-us/s

There is Microsoft's Azure speech-to-text API: https://azure.microsoft.com/en-us/services/cognitive-services/speech-to-text/ https://azure.microsoft.com/en-us/services/cognitive-service... But MSR's research system from that paper is probably quite different from what's deployed in Cortana/Azure, and Switchboard is a very different task from podcast transcription.