Alexa Skills Kit Lets You Emphasize Select Words

Amazon today announced an update to the Alexa Skills Kit that will let developers control of how Alexa speaks. SSML Speechcons can spice up Alexa's pronunciation for the sake of clarity, emphasis, or just for fun.  

Listening to Alexa, Siri, Cortana, and the Google Assistant is a painful reminder that computers still don't know how to speak, at least not really. Nuances in pronunciation that humans find essential to meaning and comprehension are simply lost in the raw regurgitation of words we so often receive from Alexa and others. Amazon is trying to make Alexa's spoken word a bit more … human … via Speech Synthesis Markup Language or SSML.

If you're not familiar, SSML is a standardized language that allows words to be tagged for alternate pronunciations. Alexa already supports a handful of SSML tags that cover the basics of reading back text, including audio, break, p, phoneme, s, say-as, speak, and w. These denote to Alexa what it is supposed to do when encountering the text following each specific tag.

Amazon has added SSML Speechcons to the Alexa Skills Kit bag of tricks. Speechcons are "special words and phrases that are pronounced more expressively by Alexa," according to Amazon's David Isbitski. Developers need Speechcon tags, such as <say-as interpret-as="interjection">, around the words they want to emphasis. Alexa will encounter the tags and change its speech to reflect what the developer wants Alexa do to.

Amazon offers a great example (mp3 audio file here) with the word "bazinga." Listen to the way Alexa speaks the word. The emphasis makes the word sound more like it would if you or I said it to one another in the context of a conversation. (No one says "bazinga" monotone with no emphasis.) The SSML code can be wrapped around any word. Amazon suggests developers accompany Speechcon words or phrases with a pause to add dramatic effect.

More than 100 Speechcons are already recognized by Alexa, says Isbitski. Some standouts include: abracadabra, achoo, aha, ahem, ahoy, all righty, aloha, baa, bada bing bada boom, bah humbug, bam, bang, batter up, bazinga, checkmate, cheerio, cheers, and dozens more. A full list is available here.

Amazon has a ton of resources on the subject. You can find SSML reference materials here, and the Alexa Skills Kit here

Be sure to read the next Artificial Intelligence article: Amazon Alexa Has More Skills Than You Do


Comments (0)