Skill not working only on ADC - Choice Block issue

Help
#1

Hi everybody, I hope somebody can help me with my current issue!
Problem: the Skill works fine on the VoiceFlow prototype section, but in ADC there are some errors.
The Skill is supposed to be a simple riddle game for children. Single player, with points assignment (and cheerings for the kid playing).

The flow, summarized, is:
Welcome message -> Random Block (alexa chooses a riddle between 20 different options) -> Choice blocks (one for each different question). If choice is correct: +1 point (with set variable block); random block (alexa chooses a congratulation message between 3 different options, and communicates the total amount of points). If the choice is wrong: “Wrong answer, try again” message, no points assigned. After the congratulations or the try again message, Alexa will choose again another riddle from the list, and so on (without duplicates).

Everything works really well in the “Prototype section”. I checked the debug mode: everything works as it is supposed to.

But in ADC it happens something really weird.
The skill runs, but if the user gives a wrong answer, Alexa answers with one of the “congratulation” messages. The points are not assigned, though. Then, the same riddle is told once again by Alexa; if the answer is wrong again, the app suddenly stops working. If the correct answer is given, the game keeps running. Curious fact: yesterday the skill was running perfectly also on ADC.

More info about variables and intent/utterances: the only variable that I added is the “points” one, and works well. About the intent: i created a specific intent for each riddle, with utterances that are synonyms of the correct choice. Example: Riddle: “I fly but I don’t have feathers, i have wings but I don’t tweet, what am I?” Intent: riddlesolution1; utterance: plane, airplane (etc etc). If the answer is riddlesolution1, the path leads to “congrats message”; if it’s “else”, the path leads to the “oh no” message. But on ADC looks like that the “Else” path is bonded to the congratulation message. :thinking:

Thank you so much for reading all this, I tried to be as clear as possible. I hope somebody can help me!

#2

seems you have some problems, not only 1. you should break down.

the Skill works fine on the VoiceFlow prototype section, but in ADC there are some errors.

it’s common and not strange thing. Prototype testing uses different techniques from ADC because some features works only on ADC/Echo devices and 3rd party cannot use. Prototype simulates it but not same as ADC and Echo devices. So, I always recommend tests should be done on ADC because it is more similar to real Echo devices.

The skill runs, but if the user gives a wrong answer, Alexa answers with one of the “congratulation” messages. The points are not assigned, though.

maybe you put your answers as sample utterances of some intent in Choice Block? Using custom Intent, Alexa tries to match what user said with any of intents which she thinks the most similar one, even if what user said is not match for sample utterances. Alexa will “guess” and sometimes it goes wrong. Sample utternaces in the intent works like choosing a way which is the most possilble, not for match/not match, which is correct/wrong in quiz thing.

If you wanna get what user said as “value” and check if it matches with the other value, which is an answer, you should get it as a slot and check it using Condition Block.

should leave “point” things aside for now because it seems logic problem.

#3

Updated: Voiceflow now supports Skill Model Sensitivity Tuning.

About Skil Model Sensitivity Tuning, see here.

This means, by increase sensitivity, if what user said does not match with sample utterances, it will go to else path more than usual. Sounds nice?

In Voiceflow, open your project and click on gear icon on top.

s_1309
s_1309800×164 12 KB

Project Settings page will show up. Scroll down to “Channel Specific Feature” section and change “skill model sensitivity” to “Medium” or “High”.

1019×486 35.6 KB

This does not supports my locale(ja-JP), so I’m not sure how this change Alexa’s recognition. Also, this means the more incresing sensitivity, the less flexible and Alexa may match “ONLY” with utterances defined, I guess. But If your locale is en-**(All English locale) or de-DE(German), worth to try.

This might be help for your use-case.

#4

Hi Kun,
Thank you so much for your precious advices!
Yes, I put the answers as sample utterances… Unluckily the increasing sensitivity option is not supported in my locale too (italian), so I guess I will have to change every Choice Block with a Condition Block (sadly, long job :sweat_smile:)
Thank you again for your explanations, I had no idea that Alexa had this kind of “guessing” way of working with che Choice block. I will let you know if this strategy worked as soon as I finish the job :blush: