Optimize Classification Results using the Selection Document Set
If your classification benchmark contains unclassified or incorrectly classified documents, open those documents by clicking a column in the Result Matrix. The documents for the selected row are loaded into the Selection document set so that you can view them and troubleshoot why they were not classified as expected. You can open and view documents from the Selection document set while the Classification Benchmark window is open. This means that you can add documents from your Selection document set to the Classification Set, train your project, and then re-run the benchmark.
Generating a new benchmark removes documents loaded in the Selection document set. When you select a column in the new benchmark Result Matrix, documents are reloaded.
You can optimize your classification results by following these steps:
- Open the Documents window if it is not already open.
-
Run a
classification benchmark.
The Result Matrix is populated with the classification benchmark results.
-
Click an
Unclassified (yellow) column or an
Incorrectly Classified column (red).
Documents from the selected column are displayed in the Selection document set in the Documents window.
- Review the documents and add one or more of them to the Classification Set for the relevant classes.
-
On the
Process tab, in the
Train group, select
Classification or
Separation &
Classification
When you enable separation, Separation & Classification becomes available.
Your project is trained for the newly added training documents.
-
Run another
classification benchmark.
You should see fewer unclassified and incorrectly classified documents.
The Result Matrix is populated with the classification benchmark results.