Utterance --> Speech Streaming --> Your Service (back end) --> Response
Alexa (wake word) - open (launch) - google maps (invocation name)
Alexa (wake word) - ask (launch) - uber (invocation name) - for a ride (utterance)
Recommend a restaurant within four miles --> { distance: 4 }; "Four" is a slot the is extracted and processed by the back end.
Tell me about Italian food within four miles --> { cosine: 'italian' }, { distance: 4 }
AMAZON.CancelIntent
AMAZON.HELPIntent
AMAZON.STOPIntent
Required for certification
{ pet } -> dog, cat, fish
{ size } -> size, small, medium
{ color } -> white, black, grey
I want a { pet }, that's { size } and { color }.
In case the user will not say everything in one sentence.
I want to order a pizza. Pizza without tomato. Big size.
A good idea is to ask for confirmations in a Multi Turn Dialog.
User: I want to order a pizza.
Alexa: How do you want your pizza?
User: Pizza without tomato.
Alexa: A pizza without tomato, is that right?
User: Yes.
Alexa: What size?
User: Big.
Alexa: A big size pizza without tomato, should I place th order?
User: Yes.
- Voice First
- Cognitive Load
- Conversational Content
- Speech
Choose voice when it is...
Easier - Faster - More Natural
Choose { INVOCATION_NAME } very carefully, it cannot be changed once published.
The { INVOCATION_NAME } should also be at least two words, unless it's a specific trademark.
Handle Correction from the User and Under/Over Answering cases.
Present Definitive Choices
Alexa: We have pepperoni and cheese. Which one would you like?/ User: Pepperoni.
Make your answers different, increase variety.
Keep the dialog simple and clear, no many choices, brief and understandable answers and interactions.
Do not assume the user knows everything, guide the user through simple choices (even binary).
Up to 500 users can Beta Test your Alexa skill.