Hey,Siri. Set an alarm for 6:30 AM tomorrow morning.
Okay. I set an alarm for 6:30 AM.
Boy, speech recognition(语音识别) technology sure has come a long way. Have you ever stopped to think about how it works?
Actually, I have. When you trigger Siri by Speaking, you speech is converted from audio to its corresponding (相当的 相应的 一致的 通信的) textual form in the back-end of the system.
I see. Developing such software must have been extremely challenging.
And time-consuming. Siri is able to cater to various accents because engineers train machine learning models on large transcribed (转录 转写) datasets comprised of voice samples from large groups of people.
That's fascinating.
Yeah,but what I find fascinating is that Apple servers run Natural Language Processing, or NLP, algorithms to understand the intent of what the user is trying to say.
Which enables NLP engines to distinguish exactly what someone is saying.
You got it. But doing so is a mammoth(巨大的) task because of the many different ways you can say one thing.
I imagine contextual understanding must be another major aspect of Siri's machine learning technology then?
Sure is. That's why more and more we're able to speak to Siri like any other human. Hey Siri, play "Yellow Submarine" by the Beatles.