Audio data labeling using Azure Machine Learning – Hands-On Exploring Data Labeling Tools

In this section, let’s see how to label audio data using Azure Machine Learning.

In Azure Machine Learning, we can annotate pieces of audio using text labels. We are going to use the cat_1.wav audio dataset for this. We will play the audio in Azure Machine Learning and use the cat label for that audio. This file is located in the Datasets/Ch12 folder at the GitHub path specified in the Technical requirements section. The same process can be followed to label the required number of audio files.

After labeling, we are going to export the labeled audio files to Azure storage, and from there, we can consume them in the Azure Machine Learning pipeline as a dataset.

Similar to image data and text data, first, we will create a labeling project for audio data.

All the steps all similar to the previous section, except the labeling task type, for creating a new audio project.

Let’s create the project first for audio data labeling in Azure Machine Learning.

Step 1 – Create the project

Let’s create an audio project by following the same steps that we have seen in the previous section for creating image and text data labeling projects.

Enter the project name, select Audio for Media type, and select Audio Transcription (Preview) for Labeling task type:

Figure 12.27 – Creating a project for audio transcription

The sample audio dataset that we are going to label is cat_1.wav. This is available in the GitHub repository.

Once the project is created, go to the Data Labeling page and click on the project name. On the project page, click on Label data in the same manner that we saw for the text labeling project.

Step 2 – Label the audio data

You will now be on the audio page where you can play the audio and enter the tag name in text format in the Transcription area under the audio.

Now, you need to add the transcription label for your audio data. As shown in the following screenshot, there is a Play tab to play the audio:

Figure 12.28 – Playing the audio

Select the cat label for this audio, as shown here:

Figure 12.29 – Labeling the text for a piece of audio

You have seen how to create a project for audio data and label the audio data in Azure Machine Learning. Let’s now see how to integrate this labeled data for training ML models in the Azure Machine Learning pipeline.

Leave a Reply

Your email address will not be published. Required fields are marked *