There’s nothing we love more than learning new things and then building something from what we’ve learned. Working with emerging tech is always exciting because of the discovery necessary, but blazing a new trail can also be a bit of a pain. The Alexa Echo Service by Amazon is no exception.
We recently launched our own Magnetry skill for Amazon’s smart speaker environment, hoping to learn all about the process of developing client solutions using far-field voice recognition and the Alexa Skills Kit. If you’re interested in having the benefit of our experience, keep reading.
Ten Things We Learned by Building Our Magnetry Alexa Skill.
1. First, it really is fun.
We enjoyed the adventure in understanding how to code a verbal command dialog. Since the beginning of the digital age we have always had to reach for a tablet, phone or other computer in order to have some sort of physical and visual dyad interaction. This was different enough that it provided the team with a lot of fulfilled learning. It also requires a totally different thought process.
2. There is more than one way to ask for an apple.
With the dawn of Amazon’s Echo and Google’s Home systems we have reached a new chapter in the computer revolution, where device interaction is intangible, your thoughts and voice communication are your only navigation tools. As programers we found ourselves in unknown territory. We are used to building visual experiences that mesmerize the eye and adding reasoned cold logic and math to make software do what our bidding. With the AlexaSkill it was necessary to understand the English language as spoken, not written and be able to predict what phrases a user may use. For example, there are several ways to ask for an apple and because of that, each verbal command needs to be accounted for in the programming.
3. Challenging documentation.
Challenges were abundant from a scattering of API documentation, to understanding the process of coding a fluid dialog with predictable verbal commands within the parameters set by Amazon. Documentation is always a challenge when you’re working with large companies like Amazon, of course. It is just one of those trial by error situations, where you figure out what you need to do—as you do it. And maybe learn something along the way.
4. Prepare for rejection.
One important piece of advice, we’d give anyone building an Alexa Skill is to expect your Skill to be rejected several times before acceptance. It does not matter how seasoned of a programmer you are, this is new territory and small specifications can and will be missed since no one is fluent yet in coding dialogue.
5. It’s still a little glitchy.
This is cutting edge tech that has been only out for a year and a half so we must realize the technology’s limits. There are API glitches that we encountered and work-arounds that were required to adjust for quirky guidelines. Not a surprise that the first few generations would see issues, but the benefits of being the first players to the platform really helps justify the risk.
6. Content writers must account for Alexa’s quirks.
The way one would approach writing a script, or a radio spot is completely different than how you’d approach writing content for Alexa to read aloud. She anticipates punctuation pretty well, and even all CAPS. But she’s not yet a great narrator, so we have to conform to her limitations as a reader.
7. Build off of what is already there.
Amazon offers pre-existing APIs that make getting started a little less daunting. We might recommend utilizing the Smart Home or Flash Briefing Skill APIs on a small scale before tackling something completely custom.
8. Think like a Jeopardy contestant.
The key interactions with Alexa are questions. The more specific and unique questions you decide to handle with your code, the more helpful your skill will be, and the less trouble Alexa will have serving answers. It’s also imperative to let the user know when and how to respond. Having Alexa ask a question like, “Which one do you prefer” after a list is better than simply listing a few options.
9. Watch your invocations.
Users say a skill’s invocation name to begin an interaction with a particular custom skill. For example, if the invocation name is “Daily Horoscopes”, users can say: Alexa, Ask Daily Horoscopes for the Aries horoscope. What we found out is that Amazon frowns upon single-word invocations, unless completely unique. Though Magnetry is a made-up word meant to convey attraction, Magnetry Agency met more with more approval as our own. This is a key consideration when planning your skill, since it will be used every single time.
10. Shorter is better.
It’s surprising how impatient you can get when listening to Alexa. Amazon recommends a maximum of about 1-2 minutes for text-to-speech (TTS); and we’d say you should keep it even shorter. She gets a little annoying after about 30-45 seconds.
We had a lot of fun building our skill, but we’re not done yet. We have plans to allow our clients to search open jobs, check on their status, and maybe even review invoices through Alexa using a unique passcode. We’re sure to learn a few more tips in that process, so we’ll report back once we’re finished.
And oh yeah, Alexa pronounces Magnetry a whole lot better when it’s spelled Mag Neh Tree.