In training mode, a user pattern is created that can be used when performing OCR on the entire text. A user pattern is used when there are parts of the text that are unclear, fonts different from ABBYY's defaults, or special characters.
Note: "Pattern training is not supported for Asian languages" as per the ABBY website.
To access options for training, navigate from the topmost main menu.
Here is a screenshot of the OCR Options Tab:
NOTE: If you select the Also use built-in patterns option under the Use training to recognize new characters and ligatures, ABBYY will use its built-in patterns along with the user pattern you created which will lessen the amount of time you will have to spend on training.
Here is a screen shot of the Pattern Training pop up dialog box:
Adjust the bounding area as needed, and select effects if you wish to introduce such text features into your output. Once the boundary is set and you've inputted the correct corresponding character or letter, select train and proceed to the next level
Note: You don't need to train on the entire document. But you will need to keep going until you've provided sufficient exemplars for each character or letter in your document, often cited by OCR makers at 15 to 25 instances per character.