Manual training

Manual training allows user-selected characters to be trained.

Most characters need not be trained. As a rule, you should train character shapes which are repeatedly misrecognized or unrecognized. You can also train uncommon characters and symbols. Typically you train one character, the maximum you can train is about 5 characters. You can train only one word at a time. Use only a few pages to train characters. You can apply training made on the current page to the already existing pages of a document. You may train in order to make a training file to give greater accuracy and shorten proofing time on future documents.

Do not train individual errors caused by accidental spots on the image. You should not train character shapes unlikely to occur again.

  • The purpose of training is not to teach the program to read characters from non-supported languages or alphabets. Japanese, Chinese and Korean characters should not be trained.

To train characters manually:

  1. Right-click on the character in the Text Editor and choose Train Character… in the shortcut menu or select Train Character… from the Tools menu.

The Train Character dialog box appears.

  1. Enter the correct character in the edit box. Click Omnipage button dots Manual training to display the Character Map to insert characters from the map.

  2. Unwanted or missing fragments of desired character shapes are editable by clicking on them in the zoom window. Blue colored shape indicates the active shape to be trained, yellow shapes show the neighboring parts. If you click on a yellow-colored image it will become blue and vice versa. The colors indicate which part of a character shape will be involved in the training and which will be excluded.

  1. Click Train.

The program examines the rest of the page. If characters with similar shapes are found on the same page, the Check Training dialog box appears. (If there are no similar shapes found on the page, the dialog box will not appear.)

  1. Check if all proposed changes are correct. Some might be incorrect due to the similar shapes of some letters e.g. ‘b’ and ‘h’ , ‘q’ and ‘g’. You have the following choices:

    • If all words are correct, click on OK.
      The proposed changes will be made.

    • If only a few proposals are incorrect, select an incorrect word and click on Re-Train…
      The Train Character dialog box appears where you can re-train that single occurrence.
      Repeat the step as required, click OK when finished.

    • If many proposals are incorrect, click Cancel.
      That training will be abandoned.

To apply training to existing pages in a document:

The Apply Training dialog box appears when you leave a page on which training data was generated.

  • Select the appropriate option: from current page on, or to all existing pages.

Manual training