The top-level self property in the response body is the evaluation's URI. "description": "My Evaluation Description " Here's an example Speech CLI command that creates a test: The Speech CLI name parameter corresponds to the displayName property in the JSON request and response. This is the name that will be displayed in the Speech Studio. The Speech CLI language parameter corresponds to the locale property in the JSON request and response. This should be the locale of the dataset contents. Set the language parameter, otherwise the Speech CLI will set "en-US" by default.Set the required dataset parameter to the ID of a dataset that you want to use for the test.If you don't want to compare two models, use the same model for both model1 and model2. Set the required model2 parameter to the ID of another model that you want to test.Set the required model1 parameter to the ID of a model that you want to test.You can run the spx csr project list command to get available projects. This is recommended so that you can also view the test in Speech Studio. Set the project parameter to the ID of an existing project.Construct the request parameters according to the following instructions: To create a test, use the spx csr evaluation create command. Review the test details, and then select Save and close. Select up to two models to evaluate, and then select Next.Įnter the test name and description, and then select Next. This approach can provide a more realistic sense of the model's performance. It's important to select an acoustic dataset that's different from the one you used with your model. If there aren't any datasets available, cancel the setup, and then go to the Speech datasets menu to upload datasets. Select one audio + human-labeled transcription dataset, and then select Next. Select Custom Speech > Your project name > Test models. After you get the test results, evaluate the word error rate (WER) compared to speech recognition results. You can compare a custom model's accuracy a Microsoft speech-to-text base model or another custom model. A test requires a collection of audio files and their corresponding transcriptions. You can test the accuracy of your custom model by creating a test.
You should provide from 30 minutes to 5 hours of representative audio. Audio + human-labeled transcript data is required to test accuracy.
#AZURE SPEECH TO TEXT CUSTOM MODEL HOW TO#
In this article, you learn how to quantitatively measure and improve the accuracy of the Microsoft speech-to-text model or your own custom models.
0 kommentar(er)
