Events & Conferences2 years ago
More-inclusive speech recognition with cross-utterance rescoring
Automatic-speech-recognition (ASR) models, which convert speech to text in voice agents, typically have two stages. The first stage involves a deep neural network that maps acoustic...
