Merge pull request #54430 from Orin-Thomas/orthomas-28APR26a

Freshness & Technical Review

Author: ttorble

learn-pr/paths/tensorflow-fundamentals/index.yml

@@ -18,7 +18,6 @@ prerequisites: |
   - Basic knowledge about how to use Jupyter Notebooks
   - Basic understanding of machine learning
 iconUrl: /training/achievements/tensor-intro-trophy.svg
-hidden: true
 levels:
 - beginner
 - intermediate
@@ -32,7 +31,7 @@ modules:
 - learn.tensorflow.intro-machine-learning-keras
 - learn.tensorflow.intro-computer-vision
 - learn.tensorflow.intro-natural-language-processing
-- learn.tensorflow.intro-audio-classification-tensorflow
+- learn.tensorflow.intro-audio-classification
 - learn.tensorflow.intro-machine-learning-tensorflow
 trophy:
   uid: learn.tensorflow.tensorflow-fundamentals.trophy

learn-pr/tensorflow/intro-audio-classification-tensorflow/1-introduction.yml

@@ -1,13 +1,15 @@
 ### YamlMime:ModuleUnit
-uid: learn.tensorflow.intro-audio-classification-tensorflow.introduction
+uid: learn.tensorflow.intro-audio-classification.introduction
 title: Introduction
 metadata:
   title: Introduction
   description: Introduction
   author: Orin-Thomas
   ms.author: orthomas
-  ms.date: 08/03/2021
+  ms.date: 04/20/2026
+  ms.update-cycle: 180-days
   ms.topic: unit
+  ms.collection: ce-advocates-ai-copilot
   ms.custom:
   - team=nextgen
   - team=cloud_advocates

learn-pr/tensorflow/intro-audio-classification-tensorflow/2-understand-audio-data.yml

@@ -1,44 +1,46 @@
 ### YamlMime:ModuleUnit
-uid: learn.tensorflow.intro-audio-classification-tensorflow.understand-audio-data
+uid: learn.tensorflow.intro-audio-classification.understand-audio-data
 title: Understanding audio data
 metadata:
   title: Understanding audio data
   description: Understanding audio data
   author: Orin-Thomas
   ms.author: orthomas
-  ms.date: 08/03/2021
+  ms.date: 04/20/2026
+  ms.update-cycle: 180-days
   ms.topic: unit
+  ms.collection: ce-advocates-ai-copilot
   ms.custom:
   - team=nextgen
   - team=cloud_advocates
   ms.product: learning-tensorflow
   ms.contributors:
   - cassieb-08202021
 durationInMinutes: 10
-sandbox: true
-notebook: notebooks/2-understand-audio-data.ipynb
+content: |
+  [!include[](includes/2-understand-audio-data.md)]
 quiz:
   title: Check your knowledge
   questions:
     - content: "What is the sample rate?"
       choices:
-        - content: "Frequency mapped to time."
+        - content: "The number of audio samples captured per second."
+          isCorrect: true
+          explanation: "Correct. A 16 kHz sample rate means 16,000 samples are captured each second."
+        - content: "Frequency mapped over time."
           isCorrect: false
-          explanation: "Incorrect, frequency mapped to time is a Spectrogram."
-        - content: "The audio channels."
+          explanation: "Incorrect. Frequency content over time is represented by a spectrogram."
+        - content: "The number of audio channels."
           isCorrect: false
-          explanation: "Although audio channels can be used in sampling, this is not what sample rate is"
-        - content: "Sampling analog sound at consistent intervals of time to create a digital sound representation."
-          isCorrect: true
-          explanation: "Correct!"
+          explanation: "Incorrect. Channels describe how many separate audio signals are stored, such as mono or stereo."
     - content: "What is the waveform?"
       choices:
-        - content: "Frequency mapped to time."
-          isCorrect: false
-          explanation: "Incorrect, frequency mapped to time is a Spectrogram."
-        - content: "Sample rate and frequency visualized."
+        - content: "The amplitude of an audio signal over time."
           isCorrect: true
-          explanation: "Correct! We can visualize our data using a waveform to map sample rate and frequency"
-        - content: "The audio channels."
+          explanation: "Correct. A waveform shows how the signal amplitude changes across samples or time."
+        - content: "Frequency mapped over time."
+          isCorrect: false
+          explanation: "Incorrect. Frequency content over time is represented by a spectrogram."
+        - content: "The number of audio channels."
           isCorrect: false
-          explanation: "Incorrect."
+          explanation: "Incorrect. Channels describe separate audio signals, not the waveform itself."

learn-pr/tensorflow/intro-audio-classification-tensorflow/3-visualizations-transforms.yml

@@ -1,49 +1,57 @@
 ### YamlMime:ModuleUnit
-uid: learn.tensorflow.intro-audio-classification-tensorflow.visualizations-transforms
+uid: learn.tensorflow.intro-audio-classification.visualizations-transforms
 title: Visualizing and transforming data
 metadata:
   title: Visualizing and transforming data
   description: Visualizing and transforming data
   author: Orin-Thomas
   ms.author: orthomas
-  ms.date: 08/03/2021
+  ms.date: 04/20/2026
+  ms.update-cycle: 180-days
   ms.topic: unit
+  ms.collection: ce-advocates-ai-copilot
   ms.custom:
   - team=nextgen
   - team=cloud_advocates
   ms.product: learning-tensorflow
   ms.contributors:
   - cassieb-08202021
 durationInMinutes: 15
-sandbox: true
-notebook: notebooks/3-visualizations-transforms.ipynb
+content: |
+  [!include[](includes/3-visualizations-transforms.md)]
 quiz:
   title: Check your knowledge
   questions:
     - content: "When you resample the audio, you are..."
       choices:
-        - content: "Increasing the size."
-          isCorrect: false
-          explanation: "Incorrect."
-        - content: "Reducing the size."
+        - content: "Changing the number of samples used to represent each second of audio."
           isCorrect: true
-          explanation: "Correct! We can reduce the size of the file by reducing the sample rate for the audio track."
+          explanation: "Correct. Resampling changes the sample rate; it can downsample or upsample depending on the target rate."
+        - content: "Always increasing the size."
+          isCorrect: false
+          explanation: "Incorrect. Resampling can increase or decrease the number of samples."
+        - content: "Always reducing the size."
+          isCorrect: false
+          explanation: "Incorrect. Downsampling can reduce size, but resampling also includes upsampling."
     - content: "What is a spectrogram?"
       choices:
-        - content: "Maps the frequency to time of an audio file."
+        - content: "A visualization of frequency content over time, usually with intensity or color showing magnitude."
           isCorrect: true
-          explanation: "Correct!"
-        - content: "The audio channels."
+          explanation: "Correct. A spectrogram shows how the strength of different frequencies changes over time."
+        - content: "The number of audio channels."
           isCorrect: false
-          explanation: "Incorrect."
-        - content: "Sample rate and frequency visualized."
+          explanation: "Incorrect. Channels describe separate audio signals, such as left and right stereo channels."
+        - content: "The amplitude of the audio signal over time."
           isCorrect: false
-          explanation: "Incorrect this is a waveform."
-    - content: "Audio classification can only be done with computer vision on spectrograms."
+          explanation: "Incorrect. That describes a waveform."
+    - content: "Which input representation can be used for audio classification?"
       choices:
-        - content: "True"
-          isCorrect: False
-          explanation: "Incorrect. There's more than one way to build audio classification models."
-        - content: "False"
-          isCorrect: True
-          explanation: "Correct! There's more than one way to build audio classification models."
+        - content: "Waveforms, engineered audio features, or spectrogram tensors, depending on the model design."
+          isCorrect: true
+          explanation: "Correct. This module uses spectrograms, but audio classifiers can also learn from raw waveforms or other audio features."
+        - content: "Only PNG images created from spectrograms."
+          isCorrect: false
+          explanation: "Incorrect. Saving spectrograms as images is optional and can add unnecessary file I/O or resizing artifacts."
+        - content: "Only the number of audio channels."
+          isCorrect: false
+          explanation: "Incorrect. Channel count is useful metadata, but it doesn't represent the audio pattern to classify."

learn-pr/tensorflow/intro-audio-classification-tensorflow/4-speech-model.yml

@@ -1,19 +1,21 @@
 ### YamlMime:ModuleUnit
-uid: learn.tensorflow.intro-audio-classification-tensorflow.speech-model
+uid: learn.tensorflow.intro-audio-classification.speech-model
 title: Build the model
 metadata:
   title: Build the model
   description: Build the model
   author: Orin-Thomas
   ms.author: orthomas
-  ms.date: 08/03/2021
+  ms.date: 04/20/2026
+  ms.update-cycle: 180-days
   ms.topic: unit
+  ms.collection: ce-advocates-ai-copilot
   ms.custom:
   - team=nextgen
   - team=cloud_advocates
   ms.product: learning-tensorflow
   ms.contributors:
   - cassieb-08202021
 durationInMinutes: 15
-sandbox: true
-notebook: notebooks/4-speech-model.ipynb
+content: |
+  [!include[](includes/4-speech-model.md)]

learn-pr/tensorflow/intro-audio-classification-tensorflow/5-summary.yml

@@ -1,13 +1,15 @@
 ### YamlMime:ModuleUnit
-uid: learn.tensorflow.intro-audio-classification-tensorflow.summary
+uid: learn.tensorflow.intro-audio-classification.summary
 title: Summary
 metadata:
   title: Summary
   description: Summary
   author: Orin-Thomas
   ms.author: orthomas
-  ms.date: 08/03/2021
+  ms.date: 04/20/2026
+  ms.update-cycle: 180-days
   ms.topic: unit
+  ms.collection: ce-advocates-ai-copilot
   ms.custom:
   - team=nextgen
   - team=cloud_advocates

learn-pr/tensorflow/intro-audio-classification-tensorflow/includes/1-introduction.md

@@ -1,12 +1,17 @@
-Ever wonder how the voice assistants actually work? How do they understand the words that we say? When you think about voice assistants you have the first step, which is speech to text, then the Natural Language Processing (NLP) step, which is the word embedding (turning words into numbers), then you have a classification of the utterance (what people say) to the intent (what they want the voice assistant to do). If you are following this learning path, you will have learned how the NLP part works already. Now we want to look at how we get the text from the spoken audio. Audio classification can be used for many things, not just speech assistants. For example, in music you can classify genres, or detect illness by the tone in someone's voice, and even more applications that we haven't even thought of yet.
+Ever wonder how voice assistants recognize short commands such as "yes," "no," or "stop"? Full speech assistants usually combine many systems, including audio capture, speech recognition, natural language processing, and intent classification. This module focuses on one smaller but important task: keyword classification from short audio clips.
 
-In this learn module we will be learning how to do audio classification with TensorFlow. There are multiple ways to build an audio classification model. You can use the waveform, tag sections of a wave file, or even use computer vision on the spectrogram image. In this tutorial, we will first break down how to understand audio data, from analog to digital representations, then we will build the model using computer vision on the spectrogram images. That's right, you can turn audio into an image representation and then use computer vision to classify the word spoken! We will be building a simple model that can understand `yes` and `no`. The dataset we will be using is the open dataset Speech Commands which are built into TensorFlow datasets. This dataset has 36 total different words/sounds to be used for classification. Each utterance is stored as a one-second (or less) WAVE format file. We will only be using `yes` and `no` for a binary classification.
+There are multiple ways to build an audio classification model. A model can learn directly from waveforms, from engineered audio features, or from spectrograms that represent frequency content over time. In this module, you use TensorFlow to transform audio waveforms into spectrogram tensors and train a simple convolutional neural network to classify the words `yes` and `no`.
+
+The examples use the smaller mini Speech Commands dataset that TensorFlow provides for tutorials. The original [Speech Commands dataset](https://www.tensorflow.org/datasets/catalog/speech_commands) ([Warden, 2018](https://arxiv.org/abs/1804.03209)) contains more than 105,000 one-second or shorter WAV files across 35 spoken words. The mini Speech Commands dataset contains eight commands, and this module uses only the `yes` and `no` folders for binary classification.
 
 ## Learning objectives
-- Understand some key features of audio data.
-- Introduction to how to build audio machine learning models.
-- Learn how to build a binary classification model from wave files.
+
+- Understand key features of audio data, including sample rate, amplitude, channels, and waveforms.
+- Convert audio waveforms into spectrogram tensors.
+- Build and evaluate a binary keyword classification model from WAV files.
 
 ## Prerequisites
-- Knowledge of Python
-- Basic understand of machine learning
+
+- Basic Python knowledge
+- Basic understanding of machine learning
+- A Python environment that supports TensorFlow 2.10 or later, with TensorFlow and Matplotlib installed. Use a Python version supported by the TensorFlow release you install. For setup guidance, see [Install TensorFlow with pip](https://www.tensorflow.org/install/pip) and [Install Matplotlib](https://matplotlib.org/stable/users/installing/index.html).