How to capture and convert customer voice to text with the 'ASR' component in wolkvox Studio
Table of Contents
Introduction
The “ASR” (Automatic Speech Recognition) component allows your voice flow in wolkvox Studio to listen to the customer, transcribe what they say, and save that result into a variable to be used later (for example, for validations, branching the flow, or sending information to another system).
Additionally, you can adjust the language, listening time, and apply security measures to protect sensitive data.
This component is only available for voice routing points and is located in the “Cognitive” component group.
Configuration
- Double-click on the “ASR” component to open its configuration panel.
- Configure the following fields:
- In the “ASR analysis time” field, enter the number of seconds during which the system will listen to the customer's response.
- Example: if the question is short (“What is your ID number?”), you could use between 3 and 6 seconds, depending on the context.
- If you expect a longer answer, increase this time to prevent the system from cutting off the response.
- In the “ASR Language” field, select the appropriate language and region for the call (for example, es-CO for Spanish Colombia, es-MX for Spanish Mexico, etc.).
- Verify that the language matches your customers' way of speaking, as this improves recognition accuracy.
- In the “ASR result variable” field, type the name of the variable where the transcribed text will be stored.
- Example: customer_response, id_number, call_reason, etc.
- Also, keep in mind that the ASR component also generates an internal variable called $precision_asr, where the recognition accuracy score is stored. You can use it to validate how reliable the transcription was (for example, only continuing if the precision exceeds a certain threshold).
- Check or uncheck the “Enable beeps at the start and end of recognition” checkbox according to your needs:
- Checked: the customer will hear a beep when the system starts listening and another when it finishes. This helps clearly indicate when they should speak.
- Unchecked: recognition will be silent, without guidance tones.
- If sensitive data (for example, payment cards, passwords, or other critical information) might appear in the customer's response, enable the “Mask text in log (PCI)” checkbox.
- When enabled, the value recognized by the ASR will be masked with asterisks in logs and reports, preventing it from being read directly, in line with PCI-type security practices.
- Note: the flow will continue to use the variable normally, but it will be protected in reports and records.
- Click the “Save ASR configuration” button to apply all changes to the component.
- In the “ASR analysis time” field, enter the number of seconds during which the system will listen to the customer's response.
