Advances in Speech Recognition.md

Advances in Speech Recognition

@ WWDC 19

Support for macOS

• Supports both aAppKit and iPad apps

• 50+ languages supported

• Requires privacy approval

• Must have Siri enabled

On-Device Speech Recognition

• Speech is private, stays on-device
• Network connection not required
• No cellular data consumption

	Server	On-device
Accuracy	Best	Good
Limits	1 minute max audio duration Limited requests per day	None
Languages	50+	10+

전에 Speech Recognition을 사용해 보았을 때는 on-device recognition이 지원되지 않았었는데 지금은 지원이 된다. 하지만 여전히 한국어는 포함되어 있지 않다. 영어, 스페인어, 이탈리아어, 브라질식 포르투갈어, 러시아어, 터키어, 중국어가 지원된다.

guard let recognizer = SFSpeechRecognizer() else {
  return
}

if !recoognizer.isAvailable {
  return
}

let request = SFSpeechURLRecognitionRequesT(url: url)

if speechRecongizer.supportsOnDeviceRecognition {
  requuest.requiresOnDeviceRecognition = true
}

Speech Recognition Results: iOS 10

• Transcription
• Alternative interpretations
• Confidence levels
• Timing information

Speech Recognition Results: iOS 13

• Speaking rate
• Average pause duration
• Voice Analytics features

Summary

• Access to speech recognition in macOS
• Run speech recognition on-device
• Access to rich voice analytics features