Conversation canvas tutorial

Available in Classic and VPC

How to create a conversation flow using the conversation canvas

CLOVA Chatbot builder has added the conversation canvas feature, where you can design a conversation's flow and create an AiCall bot. This document guides you through how to create a basic conversation flow that starts with a welcome message, and create a simple AiCall bot out of it.

Basic Information

How to create conversation flow

Before diving into the exercise in earnest, the following describes how to create the most basic AiCall conversation flow.

1. Click Conversation canvas in the menu on the left.
2. Click the + button at the upper left on the canvas. You can see various types of conversation boxes in the dropdown menu.
3. Drag Welcome message to the canvas, and add conversation boxes you want that fit in the conversation flow.
4. To connect conversation boxes, click the dot on the lower right of the box you want to start the connection with, and drag the arrow to the box you want it to be connected to.

• If a system response is connected
  

• If a context is connected
  

5. After composing everything in a conversation flow, you can connect the "Finish call" box at the end to close the conversation flow.
   

Create AiCall with conversation canvas

Create AiCall with conversation canvas 1: Requesting parcel dispatch

You have created a basic conversation flow, now it's time to see how to create an AiCall bot which carries out a service in full. In the first step, we'll create a conversation where a courier company confirms a customer's intention for a parcel dispatch request.

Create welcome message
The first impression matters in any service.
To receive a parcel dispatch request, a Welcome message needs to be shown first, which gives the customer a simple introduction about the company.

1. Click the "+" button, and drag the Welcome message to the canvas.
2. Clicking on that welcome message will cause an editing panel to appear on the right. Fill the system response field with the message "Hello, this is OO courier at your service." Then, click the [Save] button.
   

Create system response
Since the welcome message has been created, now let’s create a system response message which guides the customer to the parcel dispatch request process.

1. Click the "+" button, and drag the system response box to the canvas.
2. Write "Do you want to send a package?" in the answer field of the system response box, and click the [Save] button. The saved answer will be displayed under the button.
   
3. Now let’s connect the two boxes and create a conversation flow. Click "System response" in the welcome message, and connect it to the system response box by dragging the arrow.
   

Create Yes and No options with general conversation
Now it's the customer's turn to answer. Let’s create two options, so they can answer "Yes" if they want to send a parcel, and "No" if not.

1. Click the "+" button, and drag a General conversation box to the canvas.

2. You can change the name of a conversation in a "General conversation" box according to the box's purpose. Double-click the title in the editing panel, and change the box's name to "Answer_yes".

3. Now let's enter some utterances which the user can respond to. From the "User utterance" tab in the editor window, enter "Yes" as a question and click the "Save" button. The saved question will be displayed under the "Save" button. The more user questions are input, the easier it is for the model to learn.

4. Let's enter a system response now. From the "System response" tab, enter "Items that may be damaged such as food or electric appliances can't be sent in a package. Please check if your package's content is general items such as clothes, books, or other miscellaneous goods." as the answer and click the Save button. The answer entered will be displayed under the "Save" button.
   

5. Create a flow for "No" in the same way as above. Drag a "General conversation" box to the canvas and change the box name to "Answer_no".
   From the "User input" tab, enter "No" as the question. From the "System response" tab, enter the answer "I see. Thank you for using OO courier." which will be returned if the customer doesn't want a parcel dispatch.
   

6. Let's connect the boxes now. From the previously created system response box's "output context," click the dot and drag the arrow to connect it to the "Yes" and "No" general conversation boxes we created.
   

Finish call
If the customer answers "No" to the question asking if they want a parcel dispatch, the AiCall can't continue the conversation any longer. Let's wrap up the conversation with a "Finish call" box.

1. Click the "+" button, and drag the "Finish call" box to the canvas.
2. In "System response" of the "Answer_No" box, click the dot and drag the arrow to connect it to the "Finish call" box.
   

Save
Content of a conversation is automatically saved, but you can also save it manually from the top menu, or click Ctrl/Command + S to save the conversation flow.

Finish
We have created a simple AiCall bot to receive parcel dispatch requests using the conversation canvas. The final conversation flow created is shown as below.
In the second step, let’s create a scenario where the conversation group connection feature is used to confirm a customer's information, such as a phone number, for customers who would like to request a parcel dispatch.

Conversation group

There are tons to think about while you create an AiCall bot.
For example, let's say you want to create a reservation bot for a restaurant. You have to have responses for FAQ questions such as opening hours and location of a parking space, and create a repair scenario to send the user to the original flow in case the reservation fails, as well as the restaurant reservation which would be the main flow. The conversation flow gets complicated trying to fit all these scenarios in one place, and it would take a long time to edit it as well. To prevent such issues, Conversation canvas provides the "conversation group" feature. You can draw a flow for each group by adding conversation groups, which prevents the entire conversation canvas from being tangled.

How to create a conversation group

1. To create a new group, click the "+" button next to the Home tab.
2. Double-click the group name to edit it as you'd like.
3. Create a conversation flow under each conversation group.
   

Connect conversation group
You may have created multiple conversation groups with the exercise above. Now the conversation groups have to be connected so they can work in one flow.

1. Drag a Conversation link box to the canvas.
2. Click the conversation group to connect and conversation name.
3. Let's connect the boxes now. The location to connect depends on the conversation type selected (e.g., general conversation or system response).
   • Situation 1: If the selected conversation's type is "General conversation" box, drag the arrow from the general conversation box's "output context" and connect it to "Conversation group 1."
     
   • Situation 2: If the selected conversation's type is "System response" box, drag the arrow from "System response" and connect it to "Conversation group 1."
     
4. If the connection was done right, the connected conversation group will show the link box to another conversation.
   

Create AiCall with conversation canvas 2: Requesting parcel dispatch

In the first step, we created a scenario where we asked a customer if they want to send a parcel.
In the second step, we’ll create a scenario where the conversation group connection feature is used to confirm information, such as a phone number, for customers who would like to request a parcel dispatch.
In this step, we'll divide conversation groups to manage the conversation flow better.

1. Add a "general conversation" box to the main conversation flow created in Create AiCall with conversation canvas 1.

2. Change the name of the "general conversation" to "Dispatchable items_check," and enter "Yes" as the question under the "User input" tab, and then save.
   Select No answer as the response type under the "System response" tab, and then click the [Save] button.
   

3. Drag the arrow from "Output context" from the Answer_Yes box, and connect to the "Dispatchable items_check" box created just now.
   

4. Next, click the "+" icon next to the Home group to create a new group, and change the group name to "Check phone number."
   

5. Drag a "System response" box to the Check phone number group canvas.
   

6. Change the box's name to System response_phone number, enter "Please tell us your phone number.", and then click the [Save] button. The successfully registered answer will be displayed under the button.
   

7. Now let's connect this flow to the conversation flow in Home. Go back to the "Home" tab where the main flow is contained, and drag a "Conversation link" box to the conversation canvas.
   

8. Select "Check phone number" as the conversation group to connect in the editing panel, and click "Check phone number" as the conversation box to connect.
   

9. Connect the boxes.
   

10. The complete flow looks like this.
    

11. To check if the connection has been made successfully, go to the Check phone number tab, which is the connected conversation group, and check if another conversation link box appears in front of the connected conversation box.
    

All steps have been completed. Conversations are automatically saved, but make sure to save manually from the top menu, or save by pressing Ctrl/Command + S so you won't lose precious data.

Export conversation list

Next, let's try "Export" and "Import" which are the features made so the conversation canvas and a chatbot's conversation lists can be used organically. You can create a conversation flow using the conversation canvas and export to conversation list, or "Import" conversations to the conversation canvas from a conversation list in reverse, so you can edit them while looking at the conversation flow.

1. Click the Export button at the upper right. If an error occurs during an export, a pop-up window will appear with the error content.
   

2. To see the exported conversations, select Conversation list from the menu on the left. You can check if the conversations have been exported correctly there.
   

FAQ

1. What is "Export" and "Import"?

• Conversations created using the conversation canvas are saved separately, and not synced to the chatbot until you "export" or "import" manually.
• When you click "Export," the conversations created using the conversation canvas will be exported to the chatbot. On the other hand, you can click "Import" to import conversations created using a chatbot to the conversation canvas.

2. What can I do when "Export" fails?

• If you select Export without entering number when you use "Forward to a number," the "Export" feature doesn't work.
• If an "Export" fails, the reason for the failure is shown on the alert window. Once you fix the error, you can "Export" to the chatbot again.

3. Can I use a chatbot and conversation canvas at the same time?

• You can sync a chatbot and conversation canvas using the "Export" and "Import" features, but caution is required in the following cases.

4. There are many different conversation boxes, what are the differences?