TechRadar

What is Alexa Conversation Mode and how does it work?

TechRadar logo TechRadar 25.12.2021 13:33:17 Alistair Charlton
a desk with a laptop computer sitting on top of a table: echo show 10

Interacting with Alexa has improved constantly since the Amazon voice assistant was first launched back in 2014. But, while Alexa can be asked to do several things at once - like add multiple items to your shopping list with one utterance - getting the assistant to do something always involves saying its name, or wake word, first.

Amazon has recognized how this can become annoying if you need to involve Alexa in a conversation. For example, you and another member of your household want to use Alexa to pick a recipe or find a local restaurant. Normally, you would each need to say "Alexa," or your chosen wake word, each time you wanted to ask a question to your Amazon smart speaker or smart display, such as the distance to a restaurant, or what its opening hours are.

This changes with a new feature called Conversation Mode, which was announced by Amazon back in September 2020, and is now available on the third-generation Echo Show 10 smart display - albeit only in the US for now.

Fancy grabbing yourself one of Amazon's smart speakers or smart displays? Check out the best prices on offer right now. 

It's rare for Amazon to limit new Alexa features to a single Echo device, but this is the case here because Conversation Mode makes use of the Echo Show 10's camera and display, which rotates as you move around the room, as well as its microphone. The camera doesn't store footage, or upload it to the cloud, but is used to work out when someone is directly addressing Alexa. 

By determining this, the assistant knows when it is being spoken to, and therefore when to listen and come up with a response. When the camera isn't being looked at, Alexa will stay out of the conversation.

The thinking here is that Alexa can be invited to join a conversation between two or more users. Returning to the earlier example, these people could be discussing where to go for dinner. 

They could keep summoning Alexa with the wake word every time they want to ask a question about a local restaurant. Or, with Conversation Mode enabled, they can chat between themselves, then bring Alexa into the chat simply by looking at the Echo Show 10's camera when speaking.

Conversation Mode is not enabled by default. Instead, Echo Show 10 owners need to say: "Alexa, join the conversation," and after that, the assistant will respond to anyone in the room who looks at the camera when they speak. Alexa will then respond when spoken to directly, keep quiet when the two people are talking to each other, and stop speaking if it is interrupted.

With the feature enabled, Amazon says how Alexa "is able to use visual and acoustic cues to know when to respond to requests."

To show Conversation Mode is enabled, the display of the Echo Show 10 has a solid blue border. A light blue bar is added to the foot of the screen when Alexa is sending your requests to the cloud - in other words, when it is pulling information from the web, like a restaurant's opening times, to answer your question.

Amazon says that only requests determined to be directed to Alexa are recorded and sent to the cloud, just like the requests made after you say the wake word outside of Conversation Mode. These recordings can be viewed and deleted by opening the Alexa app and going to Settings, then selecting Alexa Privacy, followed by Review Voice History.

As for disabling Conversation Mode, this is done by looking at the Echo Show 10 display and saying: "Leave the conversation". Amazon also says the feature will switch off if the Echo Show 10 "doesn't detect a request within a short period of time."

If all of this sounds familiar, then you are likely thinking of Follow-up mode, which has been a part of Alexa since 2018 and when enabled, has Alexa listen for five seconds after answering your query. During this time you can ask another question or make another request without saying a wake word.

It lacks the intelligence of Conversation Mode, however, as in this instance Alexa doesn't stop speaking when interrupted, and doesn't know when it is being spoken to. With Follow-up mode, the assistant will try to respond with every statement it hears, which could include those directed to another person in the room, instead of the Echo device.

Looking to the future, Amazon is hoping to teach Alexa to understand 'anaphoric barge-ins', which would allow a user to interrupt a list read out by the assistant. This could mean being able to say "that one" when Alexa mentioned a recipe you'd like to try or a restaurant it thinks you might like. If reliable, this has the potential to feel more natural than saying, for example, "Alexa, pick option three".

Lastly, it is worth highlighting Google Assistant and its own ability to hold a flowing conversation, called Continued Conversation. This doesn't use a camera so cannot see when it is being directly addressed (and will therefore respond to every utterance, intended for it or not). But, like Alexa's Follow-up mode, keeps the mic open for a few seconds, giving users the opportunity to ask a follow-up question.

For example, you can ask the Google Assistant for the weather forecast, then say "what about tomorrow" and it will understand you are asking about the weather again, without explicitly saying so. For now - and for those US-based Echo Show 10 for whom Conversation Mode has rolled out - Amazon's AI intelligence has the upper hand.

samedi 25 décembre 2021 15:33:17 Categories: TechRadar

ShareButton
ShareButton
ShareButton
  • RSS
Suomi sisu kantaa
NorpaNet Beta 1.1.0.18818 - Firebird 5.0 Initial LI-T6.3.0.40

TetraSys Oy.

TetraSys Oy.