Student Blog‎ > ‎

How Alexa Responds

posted Oct 10, 2018, 6:29 AM by Sameena L Ghogawala ‎(2020)‎   [ updated Oct 10, 2018, 6:32 AM ]

Amazon released a new device called Echo on November 4, 2014. Now as we yell at things in our houses they may start responding, with the use of simple wake words like; Alexa, Echo, or computer. This Amazon device can play games, turn the lights on, give you real time updates on weather and traffic, along with many other uses.

But cracking open one of these may surprise you. It is mostly speakers and microphone equipment, the secret to Echo’s smarts is ‘her’ connection to a cloud computing service called, Alexa voice services (AVS). This is Alexa’s ‘brains’.

Asking this home speaker to tell you a story often results in getting to be apart of an interactive story with different noises to transport you to the scene. Every time you call Alexa’s name, voice speech recognition alerts the device to turn on. After this point, everything you say is recorded and sent to AVS. This service has a remarkable ability to decode your voice into commands.  It can hear what you say, choose a response and send that back to your speaker at home.

Nothing Alexa says is random, it is likely that many other homes have heard the same things.  AVS can spit out trained responses, depending on the command. For weather and traffic, your home speaker sends your voice to this cloud system, then the location of the speaker all before you hear an answer, and maybe even a reminder to wear sunscreen. Alexa’s voice is manufactured to say specific words and sound out letters of the words not in ‘her’ vocabulary. Your voice recording is very important to the growth of Alexa’s ‘mind’.

Every time you have to repeat the same question twice, or Alexa cannot find an answer, those voice recordings get organized and categorized to help find Alexa's flaws. As those are tended to, the system is updated in order to help Alexa work as smoothly as possible. Because of the fairly simple build of the home speakers updating them is simple. It does not require a new device but just a change in AVS that automatically applies to every device it is connected to.