Use Dialog Management in Voice Interaction

We'll be using Dialog Management with an example "Video Match" skill. Before we add Dialog Management features, we need to create the basic skill: follow the walk-through below. If you need a refresher on how to create a skill, you should look at Codecademy's <a target="_blank" rel="noopener noreferrer" href="https://www.codecademy.com/courses/learn-alexa-skills-kit/lessons/intro-to-alexa/exercises/skill-builder-interaction-model?action=resume_content_item&course_redirect=learn-alexa">Build Your First Skill</a> course. 

The video to the right shows you how to create the Video Match skill and paste its JSON into the JSON editor.

1. Copy the JSON interaction model provided <a href="https://content.codecademy.com/courses/alexa/learn-alexa-dialog-management/0-start-interaction-model-with-synonyms-with-decade.json" target="_blank" rel="noopener noreferrer">here</a>.
2. In the <a target="_blank" rel="noopener noreferrer" href="https://developer.amazon.com/alexa/console/ask?&sc_category=Earned&sc_channel=RD&sc_campaign=DialogManagement&sc_publisher=Codecademy&sc_content=Content&sc_detail=Design_Course&sc_funnel=Visit&sc_country=WW&sc_medium=Earned_RD_DialogManagement_Codecademy_Content_Design_Course_Visit_WW_AdvancedDevs&sc_segment=AdvancedDevs">Alexa Developer Console</a>, click **Create Skill**.
3. Give it the name "Video Match" and click **Next**.
4. Select the **Custom model** and click **Create skill**.
5. Click **JSON Editor** on the left side bar. Delete the existing contents, and paste the JSON copied in Step 1.
6. Click **Save Model**, then **Build Model** (this may take a few minutes).

Once you save the model, you will see a new intent (`FindVideoByGenreIntent`) and slot types (`VIDEO_GENRE`, `VIDEO_TYPE`, `VIDEO_DECADE`) appear in the interaction model.



Build Your Skill

Well done! Let's test the intent confirmation.

#### Test intent confirmation

1. Go to the **Test** tab to open the Alexa Simulator.
2. You are testing that your Interaction Model confirms the intent after the `genre` and `videoType` slots have been elicited/confirmed. As you did before, enter a phrase that does not specify the genre, like:

```
ask video match to find me a video
-- or --
ask video match to give me a recommendation
```

3. Alexa should elicit the genre by saying: "what genre would you like. You can say action, comedy, horror, or drama"
4. Respond with a genre.
5. Alexa should confirm the genre by saying something like: "so you're looking for the action genre, correct?" or "You said action, right?" (This varies with the genre you provided in step 4).
6. Confirm the genre.
7. Alexa should elicit the video type by saying: "what type of video would you like. you can say movie or tv show"
8. Respond with a video type.
9. Alexa should confirm the intent by saying something like: "so you're looking for an action movie, right?" (This varies on the slot values provided earlier).

If Alexa confirmed the intent, then your Interaction Model passes the test! 

If Alexa did not respond as expected, the test failed...Don't despair! You can go back through the previous exercises to make sure that you followed the steps exactly. If that doesn't work, make sure that you don't have any other skills in your Developer Console with the same invocation ("video match").


User: ask video match to find me a video. Alexa: What genre would you like. You can say action, comedy, horror, or drama. User: action. Alexa: so you're looking for the action genre, correct? User: yes. Alexa: what type of video would you like. you can say movie or tv show. User: movie. Alexa: so you're looking for a action movie, right?

Intent Confirmation Testing


Congrats! You've added Dialog Management to your skill's interaction model and activated its three main features:
- _slot elicitation_: asking for input from the user with _prompts_ and expecting _utterances_ in response
- _slot confirmation_: confirming slot values with the user
- _intent confirmation_: confirming the entire intent request with the user

You used these features to handle both _single turn_ and _multi-turn_ conversations, and you did it all within the <a href="https://developer.amazon.com/alexa/console/ask?&sc_category=Earned&sc_channel=RD&sc_campaign=DialogManagement&sc_publisher=Codecademy&sc_content=Content&sc_detail=Design_Course&sc_funnel=Visit&sc_country=WW&sc_medium=Earned_RD_DialogManagement_Codecademy_Content_Design_Course_Visit_WW_AdvancedDevs&sc_segment=AdvancedDevs" target="_blank" rel="noopener noreferrer">Developer Console</a>.



Review


Some pieces of information deserve confirmation, such as:

- Amounts of money
- Dates and times
- Values with multiple meanings (e.g. Springfield, Illinois or Springfield, Massachusetts?)
- Words that are hard to interpret

Using the wrong value for these kinds of slots can be costly or frustrating to users. To make sure they are correct, your skill can prompt the user to _confirm_ the value provided. Alexa expects the user to answer with "yes" or "no". In Dialog Management, this is called _slot confirmation_.  

While useful when used appropriately, slot confirmations can make a dialog tedious for users:

```
...
ALEXA: What kind of video?
USER: Tv show.
ALEXA: You said tv show correct?
USER: Yes.
ALEXA: What genre?
USER: Action.
ALEXA: You said action correct?
USER: YES.
...
```

Remember to use slot confirmation sparingly, and use them to only confirm values of high consequence.

In our skill, we will use slot confirmation to confirm the value for the slot `genre`.


1. A slot must be marked as "required" (it must have slot elicitation) in order to have slot confirmation. Like we did earlier with `videoType`, mark the slot `genre` as "required" and add prompts and utterances under **Slot Filling**.

Prompt:
```
what genre would you like. You can say action, comedy, horror, or drama.
```
Utterances:
```
give me a {genre} {videoType}
{genre} {videoType}
{genre}
```
Tip: It's a good idea to add multiple utterances that capture single and multiple slots, e.g. some of these utterances capture both `genre` and `videoType` if the user says them.
2. Once a slot is marked as required, switch on "Does this slot require confirmation?" under **Slot Confirmation**. 
3. Add a few prompts like
```
You said {genre}, right?
So you're looking for the {genre} genre, correct?
```
4. Click **Save Model**, then **Build Model** (this may take a few minutes).



Slot Confirmation


Alexa can confirm intents as well as slots. Why? There can be serious consequences if Alexa misinterprets some requests, like:
- Actions that are publicly visible (e.g. posting to social media)
- Actions that affect another person (e.g. sending a message)
- Actions where money is involved (e.g. when the user is buying something)

With _intent confirmation_, Alexa will ask the user to confirm the intent once all required slot values are collected. A good intent confirmation includes the previously slot values, and prompts the user for a "yes" or "no" response. For example, "So you're looking for an action movie, right?" If the user responds positively, your skill can fulfill the request. If not, your skill should repeat the process of collecting information.


As with slot confirmation, use intent confirmation sparingly: dialogs with too many confirmations are tedious for users. 


Follow these steps to add intent confirmation to your intent.
1. In the Interaction Model, click on **FindVideoByGenreIntent**.
2. Scroll down to find **Intent Confirmation**.
3. Switch on "Does this intent require confirmation?" 
4. Add a few prompts like
```
so you're looking for a {genre} {videoType}, right?
so you're looking for a {genre} {videoType} from the {decade}, right?
```
Although `{decade}` is an optional slot, we include a prompt with it. If the user provided a value for the slot `decade`, your skill will confirm it using the second prompt. If it hasn’t been provided yet, it  will use the first prompt.
4. Click **Save Model**, then **Build Model** (this may take a few minutes).



Intent Confirmation

#### Human Conversations
Take a look at the script to the right. If you are recommending a movie to a friend, you may have a dialog that looks like this. Before you can make a good recommendation, you may ask for your friend's preferences (movie vs. tv? what decade? which genre?) and confirm them. 

As humans, we engage in this type of conversation with each other everyday &mdash; booking flights, ordering pizza, reserving a table &mdash; going back and forth with questions and answers. 

#### Dialog Management
While creating an Alexa skill, this is the kind of experience we should be aspiring to &mdash; engaging conversations with our users. In this lesson we will be using the Dialog Management feature of the Alexa Skills Kit to design those conversations. By the end of this course you will be able to prompt the user for input (slot elicitation), confirm input from the user (slot confirmation), and finally confirm the intent (intent confirmation).  You'll also learn the conversational design concepts behind these tools to make flexible and accessible skills.

--- 

This course assumes you are comfortable making Alexa Skills with the Alexa Developer Console and AWS Lambda. If you're not, start with our <a href="https://www.codecademy.com/learn/learn-alexa" target="_blank" rel="noopener noreferrer">Learn Alexa course</a> and come back when you're ready!

Human Conversation

Good work so far! Let's make sure that slot elicitation works before moving on.

#### Add an endpoint
1. You haven't made a Lambda function for the Video Match skill yet, but don't worry! Codecademy has provided one for testing. Copy this ARN (Amazon Resource Name).
```
arn:aws:lambda:us-east-1:776991280512:function:codecademyLearnAlexaDialogManagementL1
```
2. In the Developer Console, in the **Interaction Model**, click on **Endpoint** in the left sidebar.
3. Choose **AWS Lambda ARN**.
4. In the **Default Region**, paste the ARN you copied.
5. Click **Save Endpoints**.

Your interaction model is now linked to Codecademy's Lambda function, which will handle the actual video recommendation.

#### Test slot elicitation

1. Go to the **Test** tab to open the Alexa Simulator.
2. Enable testing by clicking the slider.
3. You are testing that your Interaction Model elicits the `videoType` slot when it is not provided by the user. Enter a phrase that does not specify the video type, like:

```
ask video match to find me a video
-- or --
ask video match to give me a recommendation
```

4. Alexa should elicit the video type by saying: "what type of video would you like. you can say movie or tv show"

If Alexa asked for the video type, then your Interaction Model passes the test! It elicited the `videoType` slot. Alexa may mention something about a "problem" after that point in the conversation. That's expected, and will be fixed later in this lesson.

If Alexa did not respond as expected, the test failed...It's good that we caught the error so early! Go back through the previous exercises to make sure that you followed the steps exactly. If that doesn't work, make sure that you don't have any other skills in your Developer Console with the same invocation ("video match").

Slot Elicitation Testing

In a typical voice conversation, a user might provide all parts of their request in one go (Single Turn Conversation), or our skill may need to go back and forth prompting the user for each piece of information (Multi Turn Conversation).


#### Single Turn
In our video match skill, in order to provide a recommendation, we need the user to provide the genre and the video type. They can provide this information all at once:
 
```
Alexa, ask video match to find me a thriller movie.
```

Alexa can fulfill the request without any further questions, making the conversation a _single turn_ conversation.


#### Multi Turn
Alternatively, the user can provide the information piece by piece. To fulfill the request, Alexa needs to _prompt_ the user for more information. The user and Alexa go back-and-forth in a _multi-turn_ conversation.
```
User: Alexa, ask video match to find me a video
Alexa: Movie or TV Show?
User: I'd like a movie.
Alexa: What genre?
...
```



Single and Multi-Turn Conversations

Alexa Skills Kit's Dialog Management feature provides a simple way to handle both single and multi-turn conversations. If a user doesn't provide all necessary slot values in one turn of the conversation, Alexa will need to ask for, or _elicit_, those values from the user.


#### User Utterances in Response to Prompts
You can use Alexa’s Dialog Management feature to specify the phrases  Alexa would use to prompt the user for answers, like "Would you like a movie or tv show?" You can also add a set of utterances that you expect the user to to say in response to those prompts. For example, the user might say "Give me a movie" or just "Movie". We will include these as user utterances. 


#### Enable Dialog Management
To enable Dialog Management, we must mark at least one slot as "required" in the Developer Console. The instructions below will show you how to mark a slot as required, add prompts for Alexa to use, and add utterances you expect users to use.


Let's create some prompts that Alexa can use to fill the `videoType` slot.

1. In the <a href="https://developer.amazon.com/alexa/console/ask?&sc_category=Earned&sc_channel=RD&sc_campaign=DialogManagement&sc_publisher=Codecademy&sc_content=Content&sc_detail=Design_Course&sc_funnel=Visit&sc_country=WW&sc_medium=Earned_RD_DialogManagement_Codecademy_Content_Design_Course_Visit_WW_AdvancedDevs&sc_segment=AdvancedDevs" target="_blank" rel="noopener noreferrer">Developer Console</a>, under the **FindVideoByGenreIntent**, click on the **videoType** slot.
2. Under **Slot Filling**, switch on "Is this slot required to fulfill the intent?"
3. In **Alexa speech prompts**, add a prompt like
```
what type of video would you like. you can say movie or tv show
```
4. Under **User utterances** write what you expect the user to respond with. Just like in sample utterances for your intents, you can include slot names in curly braces `{ }` to represent any slot value the user might say (like a variable!). For example,
```
I would like a {videoType}
{videoType}
give me a {videoType}
```
5. Click **Save Model** and **Build Model** (this may take a few minutes).







Slot Elicitation

Excellent! Let's test the slot confirmation for `genre`.

If your skill has multiple slots marked as required with prompts and utterances set, Alexa will elicit the slots in the order in which the slots are listed in the **Intent Slots** section within the intent. In this case, `genre` is listed first, so we expect it to be elicited and confirmed before `videoType`.


#### Test slot confirmation

1. Go to the **Test** tab to open the Alexa Simulator.
2. You are testing that your Interaction Model confirms the `genre` slot after it is provided by the user. Enter a phrase that does not specify the genre, like:

```
ask video match to find me a video
-- or --
ask video match to give me a recommendation
```

3. Alexa should elicit the genre by saying: "what genre would you like. You can say action, comedy, horror, or drama"
4. Respond with a genre.
5. Alexa should confirm the genre by saying something like: "so you're looking for the action genre, correct?" or "You said action, right?" (This varies with the genre you provided in step 4).

If Alexa confirmed the genre, then your Interaction Model passes the test! 

If Alexa did not respond as expected, the test failed: Go back through the previous exercises to make sure that you followed the steps exactly. If that doesn't work, make sure that you don't have any other skills in your Developer Console with the same invocation ("video match").


User: ask video match to find me a video. Alexa: What genre would you like. You can say action, comedy, horror, or drama. User: action. Alexa: so you're looking for the action genre, correct? 