Learn how Alexa is making lives easier for older adults and explore the opportunity in creating voice first services for the ageing population. read more
What happens when you serve audio adverts to smart speaker listeners and ask them to make a charitable donation right there and then on their smart speaker, using their voice? read more
The cynic would think that the Amazon Echo and Alexa is just a trojan horse to allow Amazon to sell more products. read more
At Alexa Live 2020, Amazon dropped 31 new features on us, the most features dropped at one time ever. Alexa’s Head of Developer Strategy, Paul Cutsinger, joins us to break them down. read more
The microphone technology in your smart speaker, it’s pretty impressive.
When you talk, sound leaves your mouth, but it doesn’t travel in a straight line. It travels like a shotgun and it just blasts in all different directions and then bounces around the room and you get reflections.
It’s called reverb, and those reflections are timed differently, depending on the size of the room and the materials in the room and how much of the sound gets absorbed vs reflected.
That’s why your voice sounds different in the bathroom.
And so, to understand what’s being said, a microphone needs to pick up the direct signal that comes straight from the mouth, plus the reflections that are generated in the room.
Most smart speakers have more than one microphone. Some of them have about eight microphones. And so you’ll need to do all of that processing x 8,
then combine those 8 signals into one signal to then actually do the processing.
But that’s not all.
Imagine you’re not stood right next to the speaker. Imagine you’re stood at the back end of the room. Couple that with the sound of a Hoover and a screaming child and you’ve got some pretty complex work to do
So next time Alexa doesn’t quite hear what you’re saying or Google Assistant doesn’t understand you, give it a break. It’s trying to do some pretty complex stuff.
Now we have a couple weeks that have passed since Alexa Live: those few hours where Amazon just blew our minds and announced 31 new features. There are new features for designers, for businesses and, of course, for devs as well.
Now let’s look at a few features that I find most important and most exciting for developers and Alexa skill builders.
The first one is skill resumption. It’s what it sounds like. A user can resume a skill. But it’s more than that. A user could resume a skill already, right? So what’s new here?
What’s new is that a user doesn’t have to say “Alexa, launch (my skill)” to resume. Instead, the skill can sit in the background and either send notifications to a user or a user can ask for functionality without specifying the skill name.
Amazon gave the example of Uber; when you ask for a car on Uber, it takes a few minutes to find that car, and you can’t leave the skill open during that entire time. And so Uber can come back after a few minutes and say “hey, by the way, we found you a car, be on the lookout for a Ford Fiesta with this license plate”. All without you needing to ask for it.
Alternatively, a user can actually use other skills or Alexa features and then come back to the original skill without specifying the skill name.
If a user starts a yoga workout and then wants to play music or play a podcast, you can do that, then say “Alexa, what’s my next yoga move” and the skill will resume without you needing to ask Alexa to ask ’skill name’ for the next workout.
Users need to enable skill resumption permissions, which is a bit of a shame because that will add some friction to the user experience, but you understand that Amazon didn’t want this to be a runaway or risk surprising users. Imagine thinking you’re done with a skill, but the skill is right there in the background and pulls you back in.
Another thing that was announced was the controls framework.
The controls framework is a framework or a bit of code for the Alexa skills kit for node.js, and this helps developers build more complex skills and to better organize their skills as well.
The framework really manages the toughest parts of building a skill.
For example, managing what happens when a user exits the flow in the middle and then comes back to the flow later. Or managing knowing who has the initiative and if there’s a question that needs to be asked to move the user forward.
The controls framework also helps developers to organize their code a lot better.
Alexa conversations is a new way to build your skills.
Now, this isn’t something that Amazon announced at Alexa Live 2020, and in fact, I think they probably announced it about a year ago at ReMars, but they did open it up a little bit at Alexa Live.
It’s such a change because you’re moving away from code quite a bit and you’re using AI to build the skill and manage the conversation for you, instead.
You can integrate it with an existing skill and, depending on the kind of conversations you support, you could probably build a skill almost entirely or entirely with Alexa conversations.
So again, check it out. It’s something that I’ll be checking outas well and we’ll be talking about some more in the future
Something to keep in mind with Alexa conversations is that it’s only available in beta the US.
Multi value slots
Multi value slots are a smaller feature, but welcomed.
This is something that DialogFlow has had for a while, and other NLUs have had for a while too, but now Alexa has it as well.
In the past, let’s say you want to capture multiple t-shirt color choices from the user. For example:
“What color t-shirt are you looking for?”
“Blue and white”
To capture the above utterance and map it to your slots, you might have created multiple separate slots, like color_one, color_two, color_three.
Now, you can tell Amazon ‘Hey, this is a multi-value slot. You might get multiple values from this’ and Amazon will manage the rest.
VS Code toolkit
Another new new piece of tooling is the vs code toolkit. I believe vs code has had some tooling for quite a while – I’m actually am a WebStorm guy myself – but I know that vs code is had at something there, but this makes it more powerful.
Some of the things you can do now include previewing APL inside the VS code, whereas before you had to go to the dashboard in the Alexa skills kit.
Additionally, you can do local testing, which is incredibly exciting.
It’s something that certainly a lot of us have been using Bespoken for and I’m sure we will continue to use Bespoken because it is quite powerful, but with with VS code now, you can do a little more testing on your Alexa skills.
You can use that local testing outside of VS code, it’s just something that you have to set up yourself.
APL for Audio
I mentioned that you can preview APL which is great because there are new APL features as well.
There is APL audio. This is again something that Google has had for a bit of time, but it allows you to mix multiple audio tracks together in real time using APL.
If you want multiple audio files playing at once, like background music, sound effects and Alexa’s voice, you no longer have to string audio clips one after another or mix down 3 sounds together into one audio file.
This trims your audio editing time down immensely and offers new ways to be creative.
JSX for APL
There’s also JSX for APL.
Now, if you’ve ever written React code, you may have come across KSX. And if you’ve ever written APL in the past, you’ve certainly come across it, as well.
I don’t want to say that writing APL used to be a mess, but it was a bit of a struggle to write this nested Json and here using JSX, I think this is going to be a lot easier.
There’s also just so many more features as I mentioned. Here’s our top 7 new Alexa features for marketers and stay tuned for our top features for UX folks, too.
In the meantime, we hope you get to building some new interesting things with Alexa.