Enter questions

Available in Classic and VPC

When you enter conversation information, the field for entering questions becomes active. In the Question field, enter questions that users are expected to ask. The conversation model is trained using the entered questions, enabling it to understand not only the entered questions but also similar ones.
The more questions you enter, the better. We recommend registering at least 10 general questions with similar intent. It is more beneficial to enter diverse utterance types than to simply expand sentence endings to increase the number of questions. For example, if the user asks about the price of a specific product, consider how they might phrase the question and enter various types of utterances as shown below:

Example:

• How much is the product 00?
• How much does it cost to buy the product 00?
• Is the product 00 expensive?
• Tell me the price of the product 00.
• If there’s a price list, let me know.
• Can you give me a quote for the product 00?
• Tell me the employee discount price for the product 00.

Register questions

To register a question:

1. Enter the conversation information by referring to Register conversations.
2. Select the question type.
   • General question: Questions entered in natural language.
   • Regular expression question: Questions using regular expression syntax. For how to enter regular expressions, see How to enter regular expressions.
   • Intent: Enter the intent name to be passed from the intent classifier integrated to the chatbot. Instead of the chatbot directly classifying user intent by pre-entering and training from expected utterances, it receives intents classified by an intent classifier integrated to the chatbot and use them for conversation analysis.
   • Call event: A feature that enables conversations to proceed by triggering a call event. It matches conversations by triggering an event delivered at the timing when voice recognition begins in the Aicall service. This menu is enabled only for Aicall domains. For more information, see Call event
3. Enter the question and click [Add].
4. Enter at least 10 questions.
   • You can omit exclamation and question marks because they do not affect the training of the conversation model.
   • Don't just extend the same conversation with different endings; enter diverse conversations.

Batch registration of questions

You can upload a large number of questions by entering them into an Excel file. Using the Upload feature makes it easy to register questions all at once without having to enter them individually. When uploading questions as an Excel file, they must be formatted according to the specified file format. If the format is not appropriate, the upload may fail.

To register questions in bulk:

1. In the NAVER Cloud Platform console, navigate to > Services > AI Services > CLOVA Chatbot > Domain.
2. Click [Run builder] f the domain you want to run the Chatbot Builder.
3. In the Chatbot Builder, click Conversation List > [Upload].
4. When the "Upload Conversation" window appears, click [Download Template] for each upload type.
5. In the downloaded template's Excel file, enter the question type (select either "general question" or "regular expression question") in Column A, and enter the question content to add in Column B, then save the file.
6. In the "Upload Conversation" window of the Chatbot Builder, select the upload type, then upload the Excel file containing the questions.
7. Click [Start Transfer].

Move questions

You can move a registered question to a different conversation.
To move a question:

1. Select the question to move from the list of questions, and click [Move].
   

2. When the window for moving conversation appears, select the conversation to move and click [Move].

Call event

A feature that enables conversations to proceed by triggering a call event. This feature matches conversations by triggering an event delivered when voice recognition begins in the AiCall service. This feature is only supported for AiCall domains, and currently only supports the voice recognition resume event during the call event.

How it works

The call event works as follows:

• When a voice recognition resume event occurs via the call event, the Chatbot Builder searches for a conversation where a voice recognition resume event is registered.
• If there is a response that matches the context or conditions, the response from that conversation is provided to the user. If there is no response that matches the context or conditions, the call event is not processed.
• The voice recognition resume event occurs whenever voice recognition is performed in the AiCall service. Therefore, if no response exists, the event is not processed.
• Be careful that for general/regular expression questions and intent, the behavior differs from the standard failure message response triggered when no context or condition matches.

Usage example

The call event feature can be used when the chatbot first responds with a prompt to encourage the user to speak again when voice recognition resumes. For example, if it is more appropriate to receive information via the app interface than via voice during an AiCall quickstart, or if such a quickstart must be performed for security reasons, it can be designed as follows:

The following example explains the quickstart where you want to receive information via the app or web, rather than through voice prompts, when changing your delivery address.

1. Create a conversation about changing the parcel pickup location.
   • Insert the action method "$2{App Screen to Change Parcel Pickup Location}" into the response.
   • The output context adds the "App Screen to Enter Address."
   • Set the hold time to "30 minutes." This is to maintain context even when users remain on an app or web interface for extended periods.
     
2. In the API for the action method "$2{App Screen to Change Parcel Pickup Location}", perform the following tasks:
   • Process the action of displaying app or web interfaces on the user's mobile phone.
   • While the user is entering information, request the AiCall server to pause voice recognition.
3. When a user completes information entry in an app or web interface, perform the following actions:
   • The entered information is processed on the internal server.
   • To return to the call, request the AiCall server to resume voice recognition.
4. After entering information in the app or web, register the conversation to execute the subsequent quickstart in the call environment.
   • Register the voice recognition resume event as a trigger.
   • Add "App Screen to Enter Address" to the input context to ensure continuity in the "l Pickup Loscation" conversation.
   • Enter a response appropriate for the quickstart, such as "Address is entered."
     

Request to stop voice recognition

To send a request to stop voice recognition to the Aicall server, see the following:

Request form

• Http Method : PUT
• Path : /recognition/{sessionId}/control/pause
• parameter

Request examples

curl --location --request PUT 'https://ac86e44bd82542c39b6bfbc2fb6f488a.test-apigw.ntruss.com/write/v1/recogni- tion/8a9c75f8-1ed1-4925-8bed-7d741baf3abe/control/pause?contactCenterId=1' \
--header 'X-CLOVA-AICALL-API-KEY: 90d97f24-6f0b-4d9a-bbf9-ed391b44c6ea' \
--header 'X-CLOVA-AICALL-API-VERSION: 2' \
--header 'X-CLOVA-AICALL-TRACE-ID: caab126fedf14575b66b554f21b84a57'

Response form

• Http Status: 204
• There is no response body.

Request to resume voice recognition

To request resuming voice recognition with the Aicall server, see the following:

Request form

• Http Method : PUT
• Path : /recognition/{sessionId}/control/resume
• parameter

Request examples

curl --location --request PUT 'https://ac86e44bd82542c39b6bfbc2fb6f488a.test-apigw.ntruss.com/write/v1/recogni- tion/8a9c75f8-1ed1-4925-8bed-7d741baf3abe/control/resume?contactCenterId=1' \
--header 'X-CLOVA-AICALL-API-KEY: 90d97f24-6f0b-4d9a-bbf9-ed391b44c6ea' \
--header 'X-CLOVA-AICALL-API-VERSION: 2' \
--header 'X-CLOVA-AICALL-TRACE-ID: caab126fedf14575b66b554f21b84a57'

Response form

• Http Status: 204
• There is no response body.