试图让Alexa处理包含一些相同单词的不同插槽类型

Question

I'm having difficulty right now having Alexa address different slot types that contain some of the same words. I'm building this on the how-to tutorial skill https://github.com/alexa/skill-sample-nodejs-howto Let me explain what I'm trying to do: -At present you can say "Alexa, what are the ingredients for french toast"? This will return ingredients from a recipe.js file that is included in the above tutorial, which I have slightly modified. The words for these are built off a slot type, LIST_OF_ITEMS

-However, I want to be able to say "Alexa, how do I make french toast" and I want this to return the process of making french toast, which I also wrote up. These words are contain in a different slot, LIST_OF_STEPS

A sample of two intents I have are:

RecipeIntent what ingredients do I need for {Item}

StepIntent how to make a {Step}

In this example, {Item} and {Step} are the same thing(french toast) but are listed in different slots.

In my recipe.js file I have two languages being loaded at once. I've cut out most of this for clarity

"RECIPE_EN_GB" : {

//...

"french toast": "To make french toast, you will need 30g of unsalted butter, two eggs, milk 160g,

//..

},

"RECIPE_EN_GBTEST" : {

//...

"french toast": "Test test Lorem ipsum"

//...

},

When I say "What are the ingredients for french toast" it will give me back RECIPE_EN_GB french toast, as desired. When I ask how it's made, it will still give me RECIPE_EN_GB and not RECIPE_EN_GBTEST. Infact if I just use 'french toast' it will give me back RECIPE_EN_GB , I don't even need to use an utterance. I did at one point put in "testfood": "Testfood Lorem ipsum" in, and it called it fine when I said testfood. I'm aware this is unclear but it's a hard problem to try explain correctly, and I'm sorry. If anyone has experience extended the how_to tutorial please let me know

Answer 1

It is important to ensure that your intents are different enough that Alexa can tell the difference. Alexa uses a fuzzy comparison algorithm, not just exact matching to decide which intent the user has spoken.

So for example, two intents with these utterances: 1. you give me {slot_name} 2. he gave me {slot_name} would cause Alexa to have a very hard time decided which is correct when a user says "got me {slot_name}".

Try changing your utterances to something very unique in each intent, get those working, then add additional utterances. The design of your utterances is really the most important part of designing an Alexa skill. I recommend giving it a lot of time.