Amazon Alexa Starts Guessing What You Want to Ask; Gets Care Hub to Keep an Eye on Elderly
Alexa is using a deep learning-based trigger model to suggest a latent goal to users.
Advertisement
By Jagmeet Singh | Updated: 12 November 2020 13:23 IST
Highlights
Amazon Alexa is inferring customer’s latent goals
The company is using “sophisticated algorithms” for the update
Amazon is bringing Care Hub to Alexa in the US
Amazon is updating Alexa toward bringing natural interaction experiences
Amazon Alexa is becoming smarter and is coming with an added capability to guess what users would like to ask after hearing an initial question. The new change is called “inferring customers' latent goals” and is coming as a result of various “sophisticated algorithms” that help Alexa predict subsequent requests. In addition to the latest update, Amazon has announced that it is bringing elderly-focused Care Hub to Alexa to let users offer remote caregiving to their aging loved ones. The Care Hub for Alexa was announced by Amazon at its product event in September.
With the capability to infer latent goals, Amazon Alexa will be able to ask follow up questions on the basis of what you've asked initially. This is another step towards a natural interaction experience with the voice assistant.
So, when you ask Alexa about the time required for steeping tea, you will be answered saying, “Five minutes is a good place to start,” and then it will ask a follow-up: “Would you like me to set a timer for five minutes?”
Amazon explains in a blog post that it uses a deep learning-based trigger model that considers factors including the text of the customer's current session with Alexa and whether the customer has engaged with Alexa's multi-skill suggestions in the past. This helps Alexa to suggest a latent goal using a discovery model.
“Over time, the discovery model improves its prediction through active learning, which identifies sample interactions that would be particularly informative during future fine-tuning,” the company said.
Once the discovery model starts suggesting some latent goals, Alexa uses a semantic-role labeling model to look for named entities and other arguments from the current conversation. It also uses bandit learning that leverages machine learning to track whether recommendations are helpful to users.
Amazon has already enabled the new experience to Alexa in English for users in the US. Skill developers have also been provided with the Name-Free Interaction Toolkit to make their skills more visible to the discovery model.
In addition to the new update, Amazon has made the Care Hub available on Alexa to let users monitor people the well-being of people in need of assistance, especially elderly family members. The new solution requires the care receivers to have an Echo or Alexa-based device, while the person providing support needs to have the Alexa app on their phone.
An invite has to be sent from care receiver's Alexa app to the caregiver. Once that invite is accepted, the Care Hub will start providing alerts and an activity feed on the caregiver's Alexa app. It will also let the caregiver speak with the receiver by ‘tap to call' or ‘drop in' features. Similarly, there is an option to let care receivers get help from their caregivers by saying, “Alexa, call for help.” This will trigger the voice assistant to call, text, and send a push notification to the caregiver's phone.
Alongside providing alerts and the ability to monitor the health of people who need it, the Care Hub is touted to maintain their privacy. Thus, it doesn't allow caregivers to look into what exact songs or podcasts the receivers are listening to or what they said to Alexa while accessing their activity feed.
Amazon is aiming “to help simplify the remote caregiving experience for both the person providing support and their aging loved one” with the Care Hub. The new offering is, however, initially limited to the US.
Amazon Trims Jobs in Devices and Services Unit15 May 2025
Search
