question

Itai mendelsohn asked Nov 12, '15

Support for multiple Utterances with same meaning but different context

When building short chats in Alexa (to invoke a command) you might need to have two different utterances that are similar but will have different meaning in different context. Yes and no are obvious examples. The forum has a great example on how to overcome it (by Freebusy.io). My question (or request) is on what are the plans to better support such a case looking forward. When there is a need for a tiny conversation to obtain from the user all the info to perform the request in few steps (in order to have a great user experience). Again freebusy defined it extremely clear. In the wiseguy example you show how we can use the session to keep track of the different steps. Following the terms in the example. Any plans to link the session stage to priorities or context to different intents but with similar utterances? Tnx

alexa skills kit

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

jjaquinta answered Nov 13, '15

I'm not sure what you expect the ASK to do. The feature I've asked for is to be able to turn on/off different intents depending on the context. However, working out what a given intent means for a given context is really the skill's responsibility. Any skill that maintains state is wise to implement it as a state machine. It's a very common pattern in software design: https://en.wikipedia.org/wiki/Finite-state_machine

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

Itai mendelsohn answered Nov 25, '15

I am not sure what question you answer. Yes familiar with state machines. But that wasn't the question. I am asking if there any plans in implement some kind of context within the same service. As mentioned in the original question. There are ways to overcome it. But those are external to Alexa and are done in the skill only. As it's a general issue I do expect for such functionality that BTW exist in similar platforms.

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

M. Mercer answered Feb 27, '16

Completely agree,I think this is major block on easily creating dynamic conversations. How about if the Response Object had an option to return a dictionary of common Utterances and Intents (if a session is being kept alive). For example, included within the Response Object is :- { { "Yes": IntentA, "No": IntentA, }} This would tell the Echo that if the user responds with either Yes, or No then the next REQUEST should map to IntentA - over-riding whatever is in the static utterances file. This would enable conversations to be much more flexible and easier to set up

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

jjaquinta answered Feb 27, '16

Paul Custinger was looking for some more advanced template skills to use during hack-a-thons. One of the ones I was thinking of doing for him was a skill that uses a finite state machine to maintain context over a conversation. My current though is to use the "you are in a maze of twisty passages, all different" from Colossal Cave as the example. But if you have a worked out sample conversation, I'd be happy to use that instead.

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.