How to determine that agent speech recognition is working
Troubleshooting Scenario
The Agent Answer Recognition toggle switch is set to Voice for a Call Target, and you need to determine whether the Speech Recognition feature is working. You may need to verify the functionality of the feature for one of several reasons:
- To verify the feature is working as expected during integration or after enabling it for the first time
- To investigate reports of the feature not recognizing agent voice when enabled
- To report on the effectiveness of the feature after deployment
Log Statement
To monitor the results of the agent speech recognition, you will need to review logs from the Adhearsion framework that is used to implement the feature.
The following excerpt from an Adhearsion log shows a failure to match the configured keywords. In this example, the word "piano" was detected, and it was not in the list. This excerpt also notes that Allow any Speech Recognition was not enabled, so this would not be considered an acceptable match.
[2021-05-27 15:59:35.668] INFO Adhearsion::OutboundCall: aa84cbe8-adc1-45d5-a707-2b444077edb1@127.0.0.1: Speech Recognition result: Allow any speech recognition is Disabled. Utterance: piano doesn't match with keywords confidence 99 for Call Request: 23c0117e5d7f201a CallTarget: 1 client: 3
Consult the table below for more information on the parameters in the log statement.
| Parameter | Definition |
|---|---|
| OutboundCall | An identifier for the outbound leg of the call being logged |
| Speech Recognition result | Indicates whether Allow any Speech Recognition was enabled or keyword matching was required |
| Utterance |
|
| CallTarget | A numeric value representing the Call Target to which the log statement applies |
| client | A numeric value representing the client (Organization) to which the log statement applies |
For complete documentation on the Agent Answer Recognition feature, see How to configure speech recognition.
