You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: articles/ai-services/speech-service/how-to-custom-speech-evaluate-data.md
+2-6Lines changed: 2 additions & 6 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -26,12 +26,11 @@ You can test the accuracy of your custom model by creating a test. A test requir
26
26
27
27
::: zone pivot="ai-foundry-portal"
28
28
29
-
30
29
1. After the model is trained, select **Test models** from the left menu. Then select **+ Create test**.
31
30
32
31
:::image type="content" source="./media/ai-foundry/custom-speech/new-fine-tune-test-model.png" alt-text="Screenshot of the page with an option to test your custom speech model." lightbox="./media/ai-foundry/custom-speech/new-fine-tune-test-model.png":::
33
32
34
-
1. In the **Create a new test** wizard, select the test type. In this example, we select **Evaluate accuracy (Audio + transcript data)**. Then select **Next**.
33
+
1. In the **Create a new test** wizard, select the test type. For an accuracy (quantitative) test, select **Evaluate accuracy (Audio + transcript data)**. Then select **Next**.
35
34
36
35
:::image type="content" source="./media/ai-foundry/custom-speech/new-fine-tune-test-model-select-type.png" alt-text="Screenshot of the page with an option to select the test type." lightbox="./media/ai-foundry/custom-speech/new-fine-tune-test-model-select-type.png":::
37
36
@@ -45,13 +44,11 @@ You can test the accuracy of your custom model by creating a test. A test requir
45
44
46
45
:::image type="content" source="./media/ai-foundry/custom-speech/new-fine-tune-test-model-status-processing.png" alt-text="Screenshot of the page that shows the status of the test as processing." lightbox="./media/ai-foundry/custom-speech/new-fine-tune-test-model-status-processing.png":::
47
46
48
-
49
-
50
47
::: zone-end
51
48
52
49
::: zone pivot="speech-studio"
53
50
54
-
Follow these steps to create a test:
51
+
Follow these steps to create an accuracy test:
55
52
56
53
1. Sign in to the [Speech Studio](https://aka.ms/speechstudio/customspeech).
57
54
1. Select **Custom speech** > Your project name > **Test models**.
@@ -256,7 +253,6 @@ You should get the test results and [evaluate](#evaluate-word-error-rate-wer) th
256
253
257
254
When the test status is **Succeeded**, you can view the results. Select the test to view the results.
Copy file name to clipboardExpand all lines: articles/ai-services/speech-service/how-to-custom-speech-inspect-data.md
+13-3Lines changed: 13 additions & 3 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -25,6 +25,18 @@ Side-by-side model testing is useful to validate which speech recognition model
25
25
::: zone pivot="ai-foundry-portal"
26
26
27
27
28
+
1. After the model is trained, select **Test models** from the left menu. Then select **+ Create test**.
29
+
30
+
:::image type="content" source="./media/ai-foundry/custom-speech/new-fine-tune-test-model.png" alt-text="Screenshot of the page with an option to test your custom speech model." lightbox="./media/ai-foundry/custom-speech/new-fine-tune-test-model.png":::
31
+
32
+
1. In the **Create a new test** wizard, select the test type. For a quality test, select **Inspect quality (Audio-only data)**. Then select **Next**.
33
+
1. Select the data that you want to use for testing. Then select **Next**.
34
+
1. Select up to two models to evaluate and compare accuracy. In this example, we select the model that we trained and the base model. Then select **Next**.
35
+
36
+
:::image type="content" source="./media/ai-foundry/custom-speech/new-fine-tune-test-model-select-models.png" alt-text="Screenshot of the page with an option to select up to two models to evaluate and compare accuracy." lightbox="./media/ai-foundry/custom-speech/new-fine-tune-test-model-select-models.png":::
37
+
38
+
1. Enter a name and description for the test. Then select **Next**.
39
+
1. Review the settings and select **Create test**. You're taken back to the **Test models** page. The status of the data is **Processing**.
28
40
29
41
::: zone-end
30
42
@@ -223,7 +235,7 @@ You should get the test results and [inspect](#compare-transcription-with-audio)
223
235
224
236
::: zone pivot="ai-foundry-portal"
225
237
226
-
238
+
When the test status is **Succeeded**, you can view the results. Select the test to view the results.
227
239
228
240
::: zone-end
229
241
@@ -500,8 +512,6 @@ You can inspect the transcription output by each model tested, against the audio
0 commit comments