Testing a tag
You can test a tag when creating or editing it. The test shows what dialogs you can add this tag to.
Warning
The semantic Pro tag testing is available in a limited mode. The semantic Pro tag's applicability is rated for ten dialogs at a time.
To test a tag:
-
Open the SpeechSense home page
. -
Go to the space you need and select a project.
-
Navigate to the Tags tab.
-
If you want to create or test a new tag, click
Create tag, set the semantic, semantic Pro or dictionary tag parameters, and enable the Test tag option for a dictionary tag or Tag testing, for a semantic or semantic Pro one. -
If you want to test an existing tag, open the tag, click Edit, and enable the Test tag option for a dictionary tag or Tag testing, for a semantic or semantic Pro one.
-
Specify the test data:
- Select the dialog selection period.
- Optionally, use
next to the Period field to select the filters. For more information about filters, see Dialog filtering.
This will display the dialogs matching the search phrase criteria specified for the tag and the filters set for the test. For dictionary tags, the search for key phrases is conducted through the whole dialog. For semantic tags, SpeechSense analyzes the dialogs again to see how close it is semantically to the search query in the context of the selected semantic attribute.
The Summarization option makes it possible to display the summarization data YandexGPT generates for each dialog.
Semantic tags additionally show the probability value for how closely the dialog matches the search criteria. A search selection includes a maximum of 1,000 random dialogs with different probabilities (high, average, low). The higher the probability, the better the dialog matches the search query. Select the Confidence field value: the tag will trigger for the dialog if its probability is at or above this value.
Semantic Pro tags display an applicability rating which indicates whether the tag is applicable to the found dialog.
At the first stage, the search selection comprises a maximum of 1,000 random dialogs of different probabilities (high, average, low). The higher the probability, the better the dialog matches the search query. The found dialogs are ranked based on their probability. At the second stage, the first ten dialogs with high probability are selected. For these dialogs, the Pro model is started. Once the testing is over, you get the applicability rating for the tag. To test the next ten dialogs, click Show more.
To test a tag against the new criteria, make the required changes and click Apply in the test section.
Do one of the following to navigate to a dialog:
- In the dialog row on the right, click
. - In the dialog row on the right, click
→ Open dialog.