Semantic Recognition

Diagram Element

Parameter Description

Node Name: name of a node.
Description: details of a node.
Context: used to filter intentions to be matched to improve the matching speed.
Semantic identification content: content to be matched for semantic identification. The value can be a character string or an existing variable of the current node. If it is left empty, the system directly identifies the customer voice.
Event: whether to clear the slot data in the matched intention if the node is not entered for the first time in the flow. Generally, the value is Clear Current Context Cache if you need to allow customers to modify information in the halfway.
- Clear Session Cache: used to clear the values of all variables in the current flow instance.
- Clear Current Context Cache: used to clear the variable value of the current intention. Generally, it is used to execute the same semantic recognition for multiple times, and to clear the intention data retained previously.
Intention Quantity: number of intentions. This applies to the scenario where multiple intentions can be returned. The value needs to be a positive integer. If this parameter is left blank, the value is the same as that of Max. Intentions(1,5) set on the Configuration Center > Chatbot Management > Semantic Understanding > System Management > Intelligent Engine Parameter > Intention Parameters page by default.
Confidence Interval: maximum confidence difference between each intention and the first intention. This applies to the scenario where multiple intentions are returned.
Cache Variable and Response Attribute: flow variables and their values. The response parameters or slot parameters of the intention can only be valid on the current node. If you want to use the value in subsequent flow nodes, define a cache variable and assign the value of the response parameter or slot parameter in the intention to the cache variable.
Condition and Condition Expression: branches of the exit of the current diagram element. For example, the intention matching success and failure can be configured as two branches. Different subsequent flows can be specified for each branch.

Condition Branch Description

**Table 1** Condition branch description
Condition Branch	Description	Usage
timeout1	First timeout	Triggered when a third-party system (for example, the IVR system) determines that a timeout occurs in receiving the customer feedback and sends a message whose input is timeout to the OIAP. This branch is generally used on the connection line between the Semantic Recognition diagram element and subsequent node, indicating that semantic recognition times out before a response is received from the customer.
timeout2	Second timeout	Triggered when the trigger condition of timeout1 occurs for the second time during a call. During script flow design, a flow retry is required when the first timeout occurs. If the second timeout occurs, the current step is skipped, and the subsequent steps are performed.
nomatch1	First recognition failure	Triggered in either of the following scenarios: The Semantic Recognition diagram element does not match a result (that is, the diagram element returns the command word indicating an unknown intention). A third-party system (for example, the IVR system) returns a message whose input is nomatch. In this scenario, the ASR system fails to recognize the intention and sends a failure response to the IVR system.
nomatch2	Second recognition failure	Triggered when the trigger condition of nomatch1 occurs for the second time during a call.
error3	Third error	Triggered when the trigger condition of timeout1 or nomatch1 occurs for the third time during a call. If the number of error times does not need to be accumulated, use the Error Count Clearance diagram element to clear the number of error times before using this branch.
SYSERROR_INNER	ODFS internal error	Triggered when an unknown error occurs in the ODFS. If no option is selected, the flow goes to the exception handling flow.
SYSERROR_INTERFACE	Business interface invocation failure	Triggered if an error occurs when the Business Interface Invocation diagram element is used to invoke a third-party interface. This branch is generally used on the connection line between the Business Interface Invocation diagram element and subsequent node.
slientInsertReply	Direct reply by a silent agent	Used only when the silent agent system is configured for the Semantic Recognition diagram element. After an agent directly replies to a customer, the flow can continue to go to other expected business flows.

Using the Diagram Element

The Semantic Recognition diagram element is used when intention matching needs to be performed for customer voices to identify customer requirements.

Click the diagram element or drag it to the canvas. Before using this diagram element, add an intention or Q&A to TUC management.

Typical Application Scenario

The following describes how to use the Semantic Recognition diagram element to interact with a robot.

Sign in to the AICC and choose Configuration Center > Chatbot Management > Semantic Understanding.
Configure a domain and an intention.
1. Choose Knowledge Management > Domain Management, click Create in the upper right corner, set Name, and click Save.
  Figure 1 Creating a domain
1. Click the name of the new domain. The Intention Management page is displayed.
2. Click Create, choose General intent, enter an intention name, configure Corpus and Response, and click Save.
  Figure 2 Creating an intention
3. Select the new intention and click Active to activate the new intention.
Configure an intelligent IVR flow. The flow orchestration scenario is as follows: After a customer is connected to the robot, the customer hears that the robot says "Hello". After the customer says "Hello", the robot recognizes the customer voice and says "Hello" again.
1. Choose Flow > Orchestration and click New to add a simple flow.
2. Click + in the Flow Variable area. In the dialog box that is displayed, set the variable name and data type. You need to add the welcome variable to play the default value (for example, Hello) of the variable to a customer after the customer is connected, and add the default variable to send the default reply (for example, Intention not identified) when semantics are not identified.
  Figure 3 Adding a flow variable
  
  Figure 4 Flow orchestration example
  - Use the reply variable FLOW.ChatBotRespContent for the Robot Reply diagram element below the Semantic Recognition diagram element to reply "Hello" to the customer after the customer says "Hello".
  - Use the reply variable default for the Robot Reply diagram element on the right of Semantic Recognition diagram element to reply "Intention not identified" when no "Hello" said by the customer is matched.
3. Save and publish the flow.
Choose Intelligent Chatbot and bind the flow to a robot.
Choose Intelligent Chatbot. In the last column corresponding to the robot, click More and choose Bind Domain to bind an active domain.
Choose Intelligent Chatbot. In the last column corresponding to the robot, click Test Call.

In the test dialog box that is displayed, click Start Call to test the robot. Enter "Hello". If the robot replies "Hello", semantic recognition is successful.