A Test Case may fail for many reasons. One reason is when the Expect To Hear prompt is not written properly, and the recognition result is sometimes very low due to a prompt mismatch. The Prompt Analyzer allows you to build and fine-tune your Expect To Hear prompt text compared to the actual audio instantly, without having to make calls. In previous versions, fine-tuning the Expect to Hear prompt field could take time, since you had to make validation calls to obtain confidence levels for the steps. With the Prompt Analyzer, offline recognition of audio is possible. The analysis also shows detailed analysis of the overall confidence level by exposing the confidence levels of the various segments. You are able to modify the expected prompt and quickly run the analysis again to increase confidence levels.
Using the Prompt Analyzer from the Navigation Menu
You can access the Prompt Analyzer from the main menu.
- To access Prompt Analyzer, from the main menu,
- Enter the Expect to Hear field.
- Specify the Major/Minor Confidence Threshold values or use the slider to set these values.
- Browse and select the audio file to compare the prompt with the .wav file. By default, the maximum allowed file size is 50 MB.
- Click the Analysis button.
The Cyara Web Portal analyzes the prompt entered with the audio file selected
and the screen shown below appears:
Using the Prompt Analyzer from the Test Case Results Screen
After a Test Case has been executed, the following screen appears in the Test Call Results
- Click to open
the Prompt Analyzer for a particular Step.
If the confidence level is lower than expected, review the analysis results to check what part of the prompt has possible errors. Listen to the audio file, tweak the prompt, and analyze again.
- You can also modify the prompt text of the test Step directly from the Prompt Analyzer page. After making your changes to the Expect to Hear prompt, click Update Test Case to save the changes.