Anyone who has ever used an Alexa-enabled device such as Echo Dot or any other Echo range of devices will know that you need to wake up the virtual assistant by calling out her name “Alexa” before it starts recording and understanding your request. But what if, you were to learn that your Echo device will keep a track of whatever you speak even if you did not summon the assistant? Creepy, isn’t it?
Amazon has filed a patent application with the US Patent & Trademark Office which details about a system that can capture and process portions of a spoken command that may occur before a wake word such as “Alexa”, BuzzFeed News reported.
The patent application which was made public last week gives details of Amazon’s plans to augment the capabilities of the virtual assistant’s voice recognition technology. At present, Echo devices or any other Alexa-enabled devices cannot understand a command when the wake word comes later in the sentence. But the patent application suggests that sentences such as “Play some music, Alexa” and “Play some music, Alexa. The Beatles, please” may be useful for ordering the virtual assistant.
“While such phrasings may be natural for a user, current speech processing systems are not configured to handle commands that are not preceded by a wakeword,” the inventors Kurt Wesley Piersol and Gabriel Beddingfield wrote in the patent.
In the current scenario, all Alexa-enabled devices only record the users if a wake word is detected. It also sends the audio to Amazon servers. In case this new technology is developed and implemented, Alexa-enabled devices would most likely be keeping a track of what users say near their device.
Even though Amazon has written in the patent application that all of the captured dialogues will not be sent to Amazon servers and “thus addressing privacy concerns associated with an “always-on” speech processing system.” the company added that it may record a limited amount of speech (for example 10-30 seconds) at a time to avoid capturing too much speech and causing privacy concerns.
However, implementing this new technology may be difficult, as according to the patent application, current systems for processing speech are not configured to handle commands that are not preceded by a wake word.