Welcome > Use Project Builder > Manage recognition > Paragraph detection > Train paragraphs

Train paragraphs

When you open a document in the Document Viewer, you can train one or more paragraphs on a document if the following criteria are met.

  1. There are paragraphs detected on the page.

  2. A class is selected in the Project Tree.

Select a single paragraph by clicking your mouse or lasso to select multiple paragraphs. Once you one or more paragraphs are selected, the Save as training text for selected class context menu option is available.

After clicking Save as training text for selected class, a modified version of the original document is added to the training set for the selected class. The filename of the new training sample is in *.xdc format and the name is based on the original document name. This *.xdc file contains the OCR results for selected paragraphs only. All other content from the original document is eliminated.

Tip If you open a training sample that contains paragraphs for training, you can edit that document in the Document Viewer. This is useful if the extracted paragraph text contains recognition errors.