NL training without input

Question

I'm a beginner at developing using bixby, and I was trying to create and modify the sample Dice capsule that they have on their github. I wanted to add an extra feature of snake-eyes to it, which basically is an extra boolean concept which is true for 1-1 roll on a pair of dice.

Another difference is that I am not taking any input as the sample capsule. In the sample capsule, the user gives the input of number of dice to be roller and the number of sides each die has. I have restricted the number of dice to be 2 with 6 sides each, so there is no point of taking input, and it basically boils down to a simple dice rolling application, which spits out the rolls and the sum.

In the simulator, I am able to get the output perfectly fine by using the "intent { goal : .... }" syntax, but i want to train the model to be able to run on prompt like "roll die", although there is no concept to match the roll command to!

Is there any way to do this? Thanks in advance!

score 0 · Answer 1 · answered Jun 23 '20 at 21:30

0

https://bixbydevelopers.com/dev/docs/get-started/quick-start#add-training-examples

You can see here that even the roll 2 6-sided dice utterance isn't being trained to a goal like a "roll" command.

Instead, it is trained to the goal RollResultConcept.

Bixby will automatically call the RollDice action to produce the RollResultConcept - so you don't need to train on RollDice yourself.

answered Jun 23 '20 at 21:30

dogethis

519
6
15

Thankyou! I'll try this and get back to you! – Akshat Srivastava Jun 24 '20 at 06:02
another feature might help you is the default-init child-key in action, which you can default a value. Please read https://bixbydevelopers.com/dev/docs/reference/type/action.collect.input.default-init – BixbyDevSupport-Tokonyan Jun 25 '20 at 17:14

NL training without input

1 Answers1