Custom question answering
Element Name: Custom Question Answering
Category: Cognitive
Description: Allows handling the knowledge base (questions and answers) created in the Microsoft Azure portal.
Functionality: Displays answers from the knowledge base in a unified manner.
Prerequisites:
-
Build resource. For further guidance: Access the Microsoft Azure portal (portal.azure.com) in the Create Resource section. For more information, refer to Azure AD B2C information https://learn.microsoft.com/en-us/azure/active-directory-b2c/.
-
Build items, each with their respective questions or possible answers.
- For some of the following fields, locate in the Azure AI / Language Studio / Custom Question Answering environment in the "Deploy Knowledge Base" section. There, you should find the option "Get Prediction URL," which opens a popup with the necessary data:
QNAMakerV2 Configuration (Required)
[Expression to Evaluate]: String field where an expression, in the form of an entity or a quoted expression, must be entered for evaluation.
[QNA MAKER EndpointKey (Optional if no AUTHORING_KEY)]: String field where authorization keys for connection configuration should be entered. Obtain these from the Azure environment by locating the created resource in the Keys and Endpoint option:
[OCP-APIM-SUBSCRIPTION-KEY]: String field where subscription keys for connection configuration should be entered. Obtain this from the "Get Prediction URL" popup.
[Minimum Evaluation Score. Valid values between 1 and 100]: Dropdown list allowing the definition of the percentage match for answers.
[PROJECTNAME]: String field where the project name is defined. Obtain this from the "Get Prediction URL" popup.
[DEPLOYMENTNAME]: String field expression.
[Maximum Menu Iterations. Valid values between 1 and 5]: When the response is a menu, this dropdown list allows defining the maximum number of accesses to the menu.
[Entity to Save the Menu]: When the response is a menu, this dropdown list allows selecting the entity where the menu will be saved.
[Intent to Execute in Case of Error. The action flow will be interrupted, and the error description will be passed to the selected intent with the name: 'QNA_MAKERV2_ERROR.]: Dropdown list where the entity name for error handling must be selected.
Example of Custom Question Answering Implementation
--Prerequisites:
Resource Created: xxxxxx
Item: Menu
Response:
Other available plans 3 play
Go Home
--Input
[Expression to Evaluate]: ‘Hello’
[QNA MAKER EndpointKey (Optional if no AUTHORING_KEY)]: ‘https://xxxxxx.cognitiveservices.azure.com’
[OCP-APIM-SUBSCRIPTION-KEY]: x56797vv98vc6cv789ooo
[Minimum Evaluation Score. Valid values between 1 and 100]: 70
[PROJECTNAME]: ‘Project_1’
[DEPLOYMENTNAME]: ‘Production’
[Maximum Menu Iterations. Valid values between 1 and 5]: 4
[Entity to Save the Menu]: MENU_2
[Intent to Execute in Case of Error. The action flow will be interrupted, and the error description will be passed to the selected intent with the name: 'QNA_MAKERV2_ERROR.]: ERROR_Project_1
--Result (Lynn Simulator Window)