Skip to content

Check categorization quality

To check the quality of your project's categorization, you must first set the target results, that is, you must "explain" to Studio what you expect from the categorization of test document. This is accomplished by annotating target categories for test documents.

Once the targets have been set, they are automatically checked by Studio after every analysis. Studio compares actual results with the targets then reports concordances and discrepancies through various outputs.

Single document analysis

After the analysis of a single test document, quality information is displayed:

  • At the bottom of the editor.

    The values shown are the precision and recall. They are computed with respect to all the document's categorization targets.

  • In the Categories tab of the Annotation tool window.

    Here quality information is visual and by target.

  • In the Categorization tool window.

    Here quality information is visual and by result.

All-document analysis report

After an all-document analysis, a report is produced showing overall and by-document quality indicators.
It is shown as a table in the Report tool window. By default both categorization and extraction indicators are shown.
To display analysis reports choose ANALYSIS in the Filter list on the first panel toolbar.

Note

The report will be produced only if there is more than one test document in the project.

The report is an XML file located in the analysis sub-folder of the reports folder that you can easily inspect in the Project tool window.

Overall indicators

Extraction quality indicators for all reports are displayed by default.

To turn them on and off select Categorization Quality Data on the toolbar. The button acts as a toggle switch.

Report indicators for documents

Extraction quality indicators for documents are displayed by default in the report's detail window.

To display extraction indicators for report's documents:

  1. Double-click a report.

Or:

  1. Select a report then select View Report on the toolbar.

To turn indicators on and off, in the Analysis details window, Documents tab, select Categorization Quality Data . The button acts as a toggle switch.

  • To sort report entries, select the column header.
  • to display or hide the error column, select Error Column .
  • To export a report, select Export CSV . Exported files are stored alongside the original XML report files in the analysis sub-folder of the reports folder.

    Tip

    If you export the reports in CSV format you can use an external tool, such as a spreadsheet software, to analyze the data.

Report indicators for categories

To display categorization quality indicators for the categories:

  1. Double-click a report.

Or:

  1. Select a report then select View Report on the toolbar.
  2. Select the Templates tab.

By default, the view is collapsed and shows first-level categoris indicators. Expand rows to display lower level categories.

Report properties

To display report properties:

  1. Double-click a report.

Or:

  1. Select a report then select View Report on the toolbar.
  2. Select the Properties tab.

Progress or regression

To check if the quality of your project is improving—a progress—, remaining stable or getting worse—a regression—, perform an all-document analysis every time the development reaches a milestone, possibly giving a meaningful name to the analysis report to indicate which milestone it refers to.

Info

Studio automatically keeps track of the time when all-document analysis is performed and assigns an ID to the report.
These info are displayed in the Date and in the ID column of the Report tool window, so that you can always distinguish between different reports even if they have the same name.

Then, to assess the trend between two all-document analysis, use their reports to create a comparison report.

Delete reports

To delete a report using the Report tool window.

  1. Select the report.
  2. Choose Delete Reports on the toolbar of the first panel.

Or:

  1. Press Del.

You can also remove a report, both the XML file and possible exports in CSV format, by deleting the files form the Project tool window.