AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Phone Ames Model Is a New Approach to Speech Recognition
i learned about phone ames after joining with s i l. I'd seen the name, but i hadn't dug into it. It's trying to do is take in unlabelled speech and generate phonams corresponding to that audio. They did this in a sort of generative adversarial way. So they took text that was arty, you know, it was good text. And then they tried to see if a discriminator could tell the difference between the sort of good phones that were actually from actual text.