You can move five percent of the text objects, randomly selected, from one training object to another.
This can be of use when you have a training object that is so large that it is not practical to incorporate cross-validation into the training. As a substitute for cross-validation, you can do as follows:
Suppose the large training object is called TO1. Make a copy of it, calling it CopyofTO1.
Create an new empty training object, called TO2, using the same category tree as TO1.
On the Training Tab, left hand pane, select the Root Training Objects node to give the two-pane view.
On the right-hand pane, right-click TO2. Select Move part of Training Object.
In the resulting Add Training Object dialog box, select CopyofTO1.
Knowledge Manager randomly selects five percent of the text objects in CopyofTO1, copies them to TO2, and deletes them from CopyofTO1.
Train a model on CopyofTO1, then test it on TO2.