Figure 2-1. Form Definition layout

Fast Track to Form Recognition

ILINX Advanced Capture Studio is an integrated environment for creating form definitions and testing the recognition process. ILINX Advanced Capture Studio is specifically designed to seamlessly integrate with the ILINX Advance Capture recognition engine, which copes with virtually any text regardless of its type and quality and allows for unprecedented accuracy and reliability of recognition.

ILINX Advanced Capture Studio has a convenient and clear interface for easy setting of form registration zones, defining of form fields, specifying preprocessing features, setting context information critical for text recognition quality, specifying recognition and validation rules, and defining output format. The business rules mechanism built upon C# and VB Net are embedded in ILINX Advanced Capture Studio to provide for coding complex application-specific processing tasks. To maximize recognition efficiency, ILINX Advanced Capture Studio allows you to test the recognition of individual forms and batches “on the fly,” collect statistics and fine-tune recognition settings.

Figure 2-2 shows the steps that the user follows to create a form definition file for a form recognition application.

Create Form Definition File

Start by creating a Form Definition file with a unique name. Then outline the configuration of the Form Definition: specify documents described by this form definition and set the number of pages for each document. These are the documents (forms) that will be recognized during the recognition session. This initial version of the form definition file may be subsequently detailed and modified as required.

Use Test Library

When working with any page, you can use the so-called Test Library. Test images library helps to organize user’s images and drastically accelerates the process of adding test images to form definition. The purpose of Test Images Library section is to organize your image library.

Form Registration

Further work with the definition file will require providing a training set of forms. For each page, select a sample form image (or images, if page clones exist) associated with that page.

First, define the registration zone on a sample image. The registration zone includes preprinted text or elements of form design that remain constant on all forms. This zone is used as a clue to help the recognizer identify form images as matching the form sample. You can fulfill form registration and check its correctness “on the fly” using the test deck of images. Form registration will reject any form images that are not images of the correct form as well as images that are so skewed or so noisy that they cannot be recognized as the correct form.

Note: If there are variations of page layout (page clones exist), the registration zone has to be set individually for each of them.

Specify Fields for Recognition

After registering form images, you can define locations of fields that have to be recognized on a sample form image. This step also involves specifying the field type and context for recognition and setting field clean-up parameters. The correctness and accuracy of every step here may significantly influence the quality of recognition. The set of fields as well as field clean-up parameters and recognition context are specified for a form page and are common for all existing page clones.

Test & Tune Recognition Parameters

At the next stage, evaluate form recognition and establish optimal settings for snippet cutting, clean-up and recognition parameters.

Testing and tuning a form definition requires a truth data file to be prepared for the set of form images used for evaluation. Truth data can be added by loading a text file that contains correct values for each field in the batch of forms. ILINX Advanced Capture Studio offers a convenient Truth Editor for rapid keying and editing of truth data.

Once the field recognition parameters are defined and the truth file is loaded, you can test the recognition results achieved for this field. After obtaining the recognition results, you can tune the parameters of recognition more precisely to achieve the best possible results. ILINX Advanced Capture Studio displays the field recognition result received for an individual form and also displays a field recognition statistics graph for a batch of forms. This graph shows the acceptance rate for the recognized fields, the character error rate, and the field error rate, as well as the current threshold setting, which can be easily adjusted in accordance with the accuracy requirements for a given field.

Initial threshold tuning can be performed on small batches for speed and efficiency, yet fine-tuning of this parameter should be performed on a large batch or as many batches as possible to obtain the most accurate value.

Specify Business Rules

Using the business rules mechanism, you can specify application-specific processing conditions or restrictions, which are applied to forms and form data. You can define rules that specify the format of answers, correct recognition answer, define relations across several fields on a form, or set other guidelines for field processing.

The business rules subsystem uses VBScript to code additional processing tasks. All of the standard VBScript functions are available for scripting in form definitions.

Configure Output Data

ILINX Advanced Capture Studio allows you to save the Form Definition to be uploaded to ILINX Capture workflow. Once the form definition tuning is finished, the test batch of forms can be processed to obtain the overall recognition results.

Main Parameters

> Input Image Type – ILINX Advance Capture accepts one- or multi-page images of forms, with no restrictions on the form design.

> Resolution – ILINX Advance Capture accepts images with resolution ranging from 200 to 300 dots per inch (DPI).

> Input Image Formats – ILINX Advance Capture supports the following image formats: black-and-white, color and grayscale. The following file types are supported: BMP, TIFF and JPEG. Black-and-white and grayscale images are supported with 8 bits per pixel. Grayscale images are automatically binarized before the recognition.

Starting ILINX Advanced Capture Studio

Important! If working under Window 7, right-click on ILINX Advanced Capture Studio icon and choose Run as Administrator from the context menu.. Then click Yes on User Account Control window.

Note: Windows 7 interface does not have ‘Run As Administrator’ option by default. The extended context menu that has that option appears on Shift+Right Mouse Click.

Figure 2-3 shows the main screen that appears. At first, this screen does not contain any definitions.

ILINX Advanced Capture Studio design follows the well-established Windows User Interface.

ILINX Advanced Capture Studio

ILINX Advanced Capture Studio Window contains the following panes:

Navigation Pane

Contains the list of windows that let the user create a form definition application. You can use it to create and tune your form definitions in order or jump to different stages of the form definition development. To jump to a specific stage, click the appropriate window heading.

These windows correspond to stages of Form Definition development cycle:

> Form Definition: This window allows you to create a form definition.

> Classifiers Summary: Visual status of all Classifiers. Classification helps you to determine the type of each page in your document flow.

> Content Classifier: Content classification uses OCR data to determine a page class.

> Layout Classifier: Layout classification uses the layout of document’s structural elements such as lines, boxes, images, etc.

> Combined Classifier: Combined classification that uses both layout and content of a page.

> Rules Classifier: Rules classification uses a set of special rules that describe one class of pages. You could use field location or recognition data as well as any automatic page classifier to write rules.

> Registration: Registration allows you to identify form images as matching the sample image(s) used for the form definition.

> Fields: The Specify Fields window is designed to let you specify a new field, define its location on the form, and set field.

> Recognition: This window enables recognition and fine-tuning of fields in your form definition.

> Scripts: In any form definition VB.Net and C# code may be added to enhance form processing.

> Release: You can design the format of the output file independently from recognition and then map recognized fields into output columns.

> Shared Objects: This window presents such additional features as Alias and Vocabulary editors.

> Truth: In this release, you are only allowed to open and save truth-files, as well as to edit individual cells.

> Image Libraries: Image Library window helps to organize user’s images.

> Configuration: Here you can define the Postal Database path and License Server.

Tool Bar Pane

Tool Bar presents icons that are shortcuts for various ILINX Advanced Capture Studio commands. The appearance of the Tool Bar depends on the window selected on the Navigation window. For information about Tool Bar for Definitions and Image tabs of the Form Definitions window see the Form Definitions Window.

The Form Definition Layout Pane

The Form Definitions Layout pane is reserved to display the current form definitions (see figure below).

The Image Preview Pane

The “Image Preview” pane is reserved for image previewing (see figure below). For further information refer to the section Working with Images.

Creating Form Definition

This section will walk you through the basic process for creating a form definition. The instructions are brief and do not include all the facets and features, which can be part of a form definition. By following the steps outlined here, you will be able to create a functional form definition.

You can begin creating a form definition using one of the following methods:

> By creating a new form definition.

> By opening an existing form definition into ILINX Advanced Capture Studio.

> By copying an existing form definition from Repository into ILINX Advanced Capture Studio.

To create a new form definition:

Step 1: Select Form Definition item in the Navigation Bar.

Step 2: Click the New icon on Tool A definition with the name New Definition appears in the Form Definition Preview window.

A new form definition will be created in the Repository.

Note: Several definitions in the system may have identical names. A form definition name can contain up to 32 characters. Valid characters are: a-z, 0-9, _.

You can rename created new form definition.

To rename a form definition:

Step 1: To rename a form definition, first select it in the Form Definition Preview pane, click the Save As icon on the Tool Bar. The Save Form Definition window appears. 

Step 2: Enter in the File Name field a new name for form definition and click the Save button.

A form definition will be renamed in the Repository. You can open an existing form definition from another location (the hard drive or network).

To open an existing form definition:

Step 1: Select Form Definition item in the Navigation Bar.

Step 2: Click the Open icon on the Tool bar. Open Form definition window appears in the Form Definition Preview pane.

Step 3: From the file list, choose the definition to be Form definition files have an .xpbd extension.

The system opens the existing form definition and adds it to the Repository. You can compare any form definitions. However, it makes sense to compare such definitions that differ from one another only slightly.

To compare two form definitions:

You can compare any form definitions. However, there is some sense to compare such definitions that differed one from another slightly. Do the following:

Step 1: Select two form definitions you want to compare.

Step 2: Click the Compare button on the Toolbar. The Compare Definitions tab opens on your screen.  Here we can see form definitions for Short and Full Passwords.

The system compares two form definitions on all their parameters: Release Columns, Scripting files, used Fields, Registration zones, and so on.

Using Form Definition Repository

ILINX Advanced Capture Studio allows you to use only one form definition at a time. But all the form definitions you have worked with are stored in the Form Definition Repository pane. Every time you open a form definition, it is automatically directed into this pane.

To view contents of the Form Definition Repository pane, shift the border between Navigation pane and Form Definition Preview pane to the right with the mouse.

Using context menu in the Form Definition Repository pane, you can do the following:

> New – Create a new form definition by selecting the New item.

> Open – Open an existing form definition by selecting the Open item.

> Save As – Save a form definition under a new name by selecting the Save As item.

> Copy – Copy selected form definition to Clipboard by selecting the Copy item.

> Paste – Paste a form definition from the Clipboard into the Form Definition Repository window by selecting the Paste item.

> Delete – Delete a form definition from the Form Definition Repository pane by selecting the Delete item.

> Select All – Select all the form definition by choosing the Select All item.

---

Form Definition

The basic steps in designing a form definition include defining the structure of the form definition, specifying test images, and defining a truth table for image registration and fields recognition. To begin designing a form definition, select Registration on the Navigation pane.

Defining the Structure

Once a new form definition is created you can proceed with defining its structure.

To define a form definition structure:

Step 1: Select the Registration window on the Navigation pane. The Form Definition Layout pane will appear on the screen (see figure below). At first, this pane is prepared for the simplest case, when a document has just one page. However, ILINX Advanced Capture Studio allows you to create form definitions for multi-paged documents as well. Refer to the section Adding Pages to the Form Definition, below, for the details of creating definitions for multi-page documents.

Step 2: In the Form Definition Layout pane, right-click the Document item. From the context menu, select Rename. Type a new document name.

Step 3: In the Form Definition Layout pane, right-click the Page item and type a new page name to rename the Figure presented below shows the definition layout after renaming.

Adding Pages

Once a new form definition is created you can proceed with adding pages to it.

To add pages to the form definition:

Step 1: Right-click on the document you would like to add a page to. The following context menu appears.

Step 2: From the context menu, select New Page and then one of the following options:

> Ordinary: Ordinary form allows you to use predefined field types and define the location of fields on it. For more information about field, types refer to the section Appendix A: Field Types.

> Invoice: Defines a page with dynamic fields, the page contains an example of Invoice document at that.

– For more information refer to the section Appendix B: Field Types for Invoice Forms.

> HCFA (CMS-1500): Includes both the HCFA0212 & HCFA0805 template pages. Defines specific HCFA 1500 forms for medical applications, which do not require any zones to be defined by the customer (defining the registration zones are disabled in this mode).

Note: Only one HCFA(CMS-1500) page is allowed per Document.

> HCFA (CMS-1500) 08/05: Includes only the HCFA0805 template pages. Defines specific HCFA 1500 forms for medical applications, which do not require any zones to be defined by the customer (defining the registration zones are disabled in this mode). Includes the HCFA0212 & HCFA0805 template pages.

Note: Only one HCFA(CMS-1500) page is allowed per Document.

> UB-04 (CMS-1450): Defines specific HCFA 1450 forms for medical applications.

Note: Only one UB-04(CMS-1450) page is allowed per Document.

– If you select any option except of HCFA(CMS-1500) or UB-04(CMS-1450), type a new name for the page added to the form

– If you select HCFA(CMS-1500) or UB-04(CMS-1450), you are prompted to set Normalize X/Y Resolution parameter to 300

Step 3: If you select any option except of HCFA(CMS-1500) or UB-04(CMS-1450), type a new name for the page added to the form definition layout (see figure below).

 

Adding Test Images

The next step in creating your form definition is to select test images. You have to select form images that will be used to fine-tune and test your form definition. You can select form images from:

> Your disk – For details see corresponding procedure below.

> Test Images Library – For further information refer to the section Image Libraries.

To add test images from a disk:

Step 1: Right-click in the Deck pane.

Step 2: From the context menu, select Add Images. The Open Images window appears (see figure below).

Step 3: Locate and open the folder containing the images you are going to process. The list of files appears in the window. You can preview image files by selecting file names.

Step 4: Select the files to be used as test images:

> To select a set of consecutive files, click the first item, press and hold down Shift, then click the last item.

> To select a set of random (non-consecutive) files, press and hold down Ctrl, and then click each item.

Step 5: Click the Open button to add the selected files. The list of Test images will display in the Deck pane of the ILINX Advanced Capture Studio workspace (see figure below).

Notes:

1. You can copy test images to a new test deck. Select these images, right click on the selection and then choose the command Copy Images to New Deck from the context menu.

2. To create a new deck, click the + icon on the deck list.

To add test images from a library:

Step 1: Click the New Deck from Library button on Tool Bar and then select the desired library from the ‘Select Library’ window.

Step 2: The list of Test images will display in the Test Images pane.

Working with Images

While working with Field and Recognition window, you can do the following:

> Center an image in the image window by clicking the button Center Image.

> Display the rulers around an image by clicking the button Rulers. If clicked once more the rulers disappear.

> Increase the image size displayed on the screen by selecting Zoom in.

> Decrease the image size displayed on the screen by selecting Zoom out.

> Adjust the zoom ratio so that the image becomes fully visible by selecting Fit to Window.

> Adjust the zoom ratio so that the entire window width is used by selecting Fit to Width.

> Adjust the zoom ratio so that the entire window height is used by selecting Fit to Height.

Transforming Images

With the images you can do the following:

> Rotate an image 90 degrees right by clicking the Rotate Right 90 button on Tool Bar.

> Rotate an image 90 degrees left by clicking the Rotate Left 90 button on Tool Bar.

> Rotate an image 180 degrees (Flip) by clicking the Rotate 180 button on Tool Bar.

> Invert an image. Click down arrow on Other icon of Tool Bar and then select Invert.

> Convert an image to grayscale format. Click down arrow on Other icon of Tool Bar and then select Convert to Grayscale.

> Convert an image to black and white format. Click down arrow on Other icon of Tool Bar and then select Convert to BW.

> Despeckle an image. Click down arrow on Other icon of Tool Bar and then select Despeckle. Isolated small chunks of black pixels will be deleted from the image. This procedure can reduce the size of the compressed form image by up to 30%, thus improve system throughput and reduce disk storage requirements.

> Deskew an image. Click down arrow on Other icon of Tool Bar and then select Deskew. Deskewing images can significantly improve the recognition process, especially for table fields.

> Remove white margin on an image. Click down arrow on Other icon of Tool Bar and then select Remove White Margin.

> Remove black margin on an image. Click down arrow on Other icon of Tool Bar and then select Remove Black Margin.

Editing Images

With the images you can do the following:

> Paint over by a black selected part of an image clicking the Fill with black label on Tool Bar.

> Paint over by a white selected part of an image clicking the Fill with white label on Tool Bar.

> Draw a straight line of selected width by selecting Line from the Edit tab.

> Draw a free-hand line using a circular brush of selected diameter by choosing Pencil from the Edit tab. You can change the line width to draw with Nx down arrow.

> Erase the background using a circular brush of selected diameter by choosing Eraser from the Edit tab. You can change the line width to erase with Nx down arrow.

> Select the width for the selected tool by choosing Nx down arrow from the Edit tab (here N may have a value from 1 to 10).

Saving Test Images

To save test images in a separate directory:

Step 1: In the Deck pane, select the files you want to save:

> To select a set of consecutive files, click the first item, press and hold down Shift, and then click the last item.

> To select a set of random (non-consecutive) files, press and hold down Ctrl, and then click each item.

Step 2: Right-click in the Deck and select from the context menu, select Save Images. The Browse for Folder dialog is displayed. Select the directory you want to save your files to and click Save.

Deleting Test Images

To delete test images from the Deck pane:

Step 1: In the Deck pane, select the files you want to delete:

> To select a set of consecutive files, click the first item, press and hold down Shift, and then click the last item.

> To select a set of random (non-consecutive) files, press and hold down Ctrl, and then click each item.

Step 2: Right-click in the Deck and select from the context menu, select Delete.

Viewing Image Properties

Image Properties contain the following parameters:

> Width: Displays the number of rows (sometimes described as scanlines) in the image.

> Height: Displays the number of columns in the image, i.e., the number of pixels per scanline.

> Horizontal resolution: Displays the number of pixels per Resolution unit in the Width (typically, horizontal) direction.

> Vertical resolution: Displays the number of pixels per Resolution unit in the Height (typically, vertical) direction.

Important Note: The captured images/documents may fail during Registration and/or Extraction if the resolution (DPI) and Height/Width do not match the settings in the Image Properties.

> Compression: Data can be stored either compressed or uncompressed. This tag displays one of the resolution type, for example CCITT T.6.

> Photometric interpretation: Displays one of the following values:

– White Is For bilevel and grayscale images: 0 is imaged as white. The maximum value is imaged as black.

– Black Is For bilevel and grayscale images: 0 is imaged as black. The maximum value is imaged as white.

> PlanarConfiguration: Defines how the components of each pixel are stored. Displays one of the following values:

– 1 = Chunky format (default). The component values for each pixel are stored The order of the components within the pixel is specified by Photometric interpretation.

– 2 = Planar The components are stored in separate “component planes.” It is not recommended for general interchange.

> Resolution unit: Displays one of the following values:

– 1 = No – absolute unit of measurement

– 2 = Inch – default value

– 3 = Centimeter

Setting Sample Images (Template)

An important component of the form definition is the sample image. A Sample Image is a typical image from a batch of form images you are going to process, which will serve as the template for all the form images. Usually, an image of a blank form is used as a sample image, but you can also create one from an image that contains data.

Notes:

1. The image designated as the sample image should be cleanly scanned and as free from background noise and distortion as possible.

2. If the Sample Image is scanned with the same settings as the images that will be processed from this form definition, the recognition quality will be higher.

To Add Sample Image (Template):

Step 1: Select the page of the document in the definition layout for which you are setting the sample

Step 2: From the context menu select Add Template. The Open Images window

Step 3: Using the Open Images dialog box, browse to the image file you wish to set as a Template and click Open. This image will be your Sample Image with the name Template for this form definition. You can rename it.

As soon as the sample image is set, it is added to the form definition structure and appears in the Definition Layout pane. You can save the edited Sample Image as a disk file for future use.

To save the sample image:

Step 1: Right-click on the sample image you want to save in the Definition Layout pane.

Step 2: From the context menu, select Save Template.

Step 3: Select the proper location for the file and type the new name in the Save Image dialog.

Note: This operation does not change the name of the sample image displayed in the Definition Layout window. To change this name in the definition layout, right-click the sample image you wish to rename in the Definition Layout window, select Rename from the context menu, and type the new name.

Multiple Sample Images

Some of your applications may require processing forms that have the same set of fields with similar contents but different versions of page layouts. This may happen, for example, if two forms look alike and contain exactly the same data, and yet are preprinted in different languages or the company logo in the upper left corner of the form is different for different companies.

To deal with such cases, ILINX Advanced Capture Studio allows you to create a form definition where two or more sample images represent a certain page of a document. These are known as clones.

Creating a form definition layout with several clones requires nothing more than selecting another typical form image from a batch you are going to process and adding it to the form definition layout.

Follow the steps described earlier in the section Setting the Sample Image to add a new sample image to the form definition layout.

Form Definitions Window

Appearance of the Tool Bar depends on the window selected on Navigation pane. For the Form Definitions window, it depends on the selected tab – Definitions or Image.

The Definitions Tab

The icons on Tool Bar have the following meaning:

The Image Tab

The icons on Tool Bar have the following meaning:

---

Document Classification

Introduction

Document classification allows you to define the classes into which your documents should be grouped, train the automatic classifiers, write rules for the Rules Classifier, add scripts to deal with complicated conditions for classification, set image preprocessing procedures, and solve other tasks necessary to classify your documents.

There are four main components – classifiers — that can be employed to achieve the final classification output. Three classifiers: Content, Layout, and Combined are automatic, i.e. they can be trained on a sample classified set of documents without direct user access to the principles of their operation. The fourth is Rules Classifier where the user explicitly formulates rules for classifying pages based on automatic classifiers’ output and fields identification and recognition.

You can also employ the Scripts tool to handle more complicated conditions. Scripts can be initiated at certain stages of classifier operation and alter it.

Note: In the current version only single pages can be classified.

From the navigation menu, you can access different classifiers as well as some auxiliary tools that help enhance the classification process as well as the Classifiers Summary to review the configurations of all four.

If you choose one of the four classifiers on the left pane, right below the upper ribbon and above the view panes you will see some sub-tabs to help you switch between different stages of work within a chosen classifier.

Note: The set of sub-tabs changes depending on which classifier you choose on the left navigation pane.

You will also have functional buttons on the toolbar at the top of the screen, these buttons are grouped thematically for different tasks (e.g. creating a Classification Set, Training Classifier, tuning thresholds for best results, etc.), and you can access different groups by switching the toolbar tabs.

The appearance of the toolbar and the tabs available there change when you switch between menu items and sub-tabs located below the upper ribbon.

In the middle and right part of your screen, there are view panes where different kinds of current project info will be shown. The appearance and contents of the view panes depend on which sub-tab is active. Meanwhile, some identically named sub-tabs in different menu items access and represent the same information (e.g. under Classification Sets sub-tab you will see Image Pool and Classification Set panes, which are common for all four classifiers).

Adding Images

Before you can start training classifiers, you need to collect document examples (images) on which you plan to train and test classifiers. The images for testing should contain examples of all document types that you expect to encounter in your workflow.

Supported formats include: .bmp, .jpeg, .pdf, and .tiff

There are two main places where you can store your images: Image Pool and Classification Sets. In the process of training and testing the classifiers you can access only the images from Classification Sets, but you may choose to first store your images in the Image pool and then form your classification sets from the Image Pool.

Note: Also, you will need to fill the image pool if you want to use clustering.

Image Pool

You will see the Image Pool window in the upper middle part of your window as soon as you choose one of the classifiers on the left navigation pane and Classification Sets sub-tab.

You can use the Image pool to store the images you plan to use for training the automatic classifiers (Content and Layout), for creating fields and for testing.

Only one Image pool is associated with every Document Definition; it will be common for all four classifiers.

For every new Definition created, the Image pool will be reset, and you will need to fill it again with images.

All images in the Image pool have labels (a label may be blank). These labels are used to group the images and are meant to be used as prototypes for future classes in Classification sets.

If you choose the Managing Classes tab, functional buttons will appear on the toolbar to manage image sets. The toolbar is divided into areas with buttons for managing images in the Image Pool and their labels in the left part of the ribbon and buttons for managing Classification Sets and classes – to the right, (followed by External Data buttons – to save classification sets to and retrieve them from your file system).

To fill the Image Pool you can use the Add Images button in Image pool section of the toolbar, it allows you to select files from your file system and copy them to the Image Pool. Folder names from which the files were taken automatically become labels, which you can later change.

After you click the Add Images to Pool button, a dialog window appears to browse through your folders. You need to select the images you want to copy into the Image Pool and click Open. Or, you can drag-and-drop the folders or images into the Image Pool from the browsing window opened.

Note: You need to choose the folders that contain the images immediately, and not the higher level folders.

Changing Labels

To change a label on an individual image use Assign label button under Managing Classes tab on the toolbar. To rename a label for all images with this label, use Rename label button.

> Delete label button will delete all images with a certain label (specified in the dialog box after you press the button).

> Select label button allows you to select all images with a certain label (specified in the dialog box after you press the button).

# Column

If you try to add multi-page tiff files to the Image pool (or Classification set), the pages get separated into individual frames (pages). In this case, several files are created within the  Image pool (one for every page), they all get the same name as the original file, and a frame number corresponding to the page number in the original file is added to the description of the file — to distinguish the images of different pages.

While file names themselves can be repeated, every image in the Image Pool or Classification set must have a unique combination of file/image name and frame number.

Frame number is reflected in the # column; you cannot change it, but you can use it for sorting. This field will be blank for all originally single page files, and it will contain the page number for all multi-page files.

Sorting Images

You can sort images in the Image Pool by name, label or frame number. To do so, click on the corresponding column names in the Image Pool window: to sort images by name click at the top of the “Image” column; to sort images by label, click at the top of the “Label” column; to sort by frame number, click on the “#” sign.

Viewing Images

If you click on the image line/name, it will become highlighted, and you will see the picture of your image in the right-hand side of the screen. You can zoom in and out using the slider in the bottom right corner of the window.

The Image tab on the toolbar contains additional buttons you may wish to use for convenient viewing of the images:

Note: Don’t worry if the images you add to the image pool before clustering automatically receive labels. The clustering procedure will neglect those labels and assign new ones at the end of its work.

Once the images in the Image Pool have labels which correspond to classes into which you want to group your documents, you can transfer this information to classification set.

Clustering Option

If you are unsure about how to group your documents and which classes you want to use for your classification, you can simply add all the example documents to the Image Pool and then start the automatic Clustering procedure. (See the section on Clustering for a detailed description)

Classification Set

To train Content and/or Layout classifiers it is necessary to create a Classification set, i.e. a classification model which serves as a knowledge base for learning. You need to determine the classes into which you want to group your documents, select sample images for each type (class), and associate them with the corresponding class.

Once you choose one of the Classifiers on the left navigation pane and when Classification Sets sub-tab is active, a window labeled “Classification Set” will be seen in the lower middle part of your screen, under Image Pool.

In this window, you can see existing Classification Sets, edit them, and create new Classification sets within the current Document definition.

Filling Classification Sets

There are two ways to fill and edit Classification sets.  You can get files (document images) directly from your file system or you can use images from your Image Pool:

> Get images directly from your file system – To fill a Classification Set you can use the function buttons in the Managing Classes tab on the toolbar. These buttons allow you to browse through your file system, copy files and folders into Classification Set; folder names automatically become class names in the Classification Set.

– Start by pressing Add folder as class button:

A dialog window will appear to browse through your files. You can drag-and-drop some previously prepared folders with images from your file system into the Classification Set window. The folder names will automatically become Class names.

Note: You cannot add a higher level folder to create multiple classes (corresponding to all subfolders with images) at once.

Instead of drag-and-drop, you can also stand on the folder with images you want to add and press Select Folder button in the dialog window.

You can also use the Add Images button in the Classification Set section of the toolbar. In this case, you will have to select images, not folders, in the dialog window and press Open.

Again, the images will be put in the class with the same name as the folder from which the images were taken.

> Take images from the Image Pool – Convenient function buttons for transferring images from the Image Pool to Classification Sets are located on the upper ribbon.

– Select Label – In the Labels section of the Managing Classes toolbar, use this button to select all images with a certain label.

– Move or copy to set – In the Image Pool section of the Managing Classes toolbar, use this button to select from the following options:

» Move selected images to class – You will be prompted to specify class name in the dialog box to which selected images will be moved. It can be an existing class or a new one created during the operation.

» Copy selected images to class – You will be prompted to specify class name in the dialog box to which selected images will be copied. It can be an existing class or a new one created during the operation.

» Move selected images to set – Selected images will be moved to classification set, labels will become class names.

» Copy selected images to set – Selected images will be copied to classification set, labels will become class names.

» Move all images to set – All images will be moved to classification set.

» Copy all images to set – All images will be copied to classification set.

You can create a number of classification sets to use for training different classifiers. All classification sets you create will be equally accessible by all classifiers.

Note: If you plan to use Combined classifier, it is important that you train Content and Layout classifiers on classification sets with coinciding class names. Otherwise, Combined classifier would yield inadequate results on the classes that exist in only one of the two classifiers.

Also, you can save additional Classification Sets within your project for testing purposes. Again, to get adequate testing results you should see that the class nomenclature in the Classification Set for testing coincides with the one used for learning.

To create multiple classification sets use the “+” button at the bottom right corner of the Classification Set window or Create New Set button on the toolbar. Once you have more than one Classification Set, tabs at the bottom of the window will appear to switch between the sets, but only one set can be used for current training and classification.

Truth Value

All images from the Classification Set automatically receive a truth value – information about which class the image belongs to, and what would be correct classifier answer. This information is used both for learning and for testing – to evaluate the classifier performance.

Note: Classification truth value shouldn’t be confused with Truth for field recognition.

Images without Class Folder

All documents without a truth value (for which there is no preliminary knowledge about their truth value — i.e. class assignment) should be put into Images without class folder which is automatically added to each new classification set.

The Images without class folder will appear in the Classification Set window when you create a new set. You can fill it with any number of images that don’t have a predefined class – these images will be classified if you run a test on the classification set. Images from this folder will not be used for training and their results will not be reflected in the results diagram, since there are no truth values for these images.

Classification

Classification is achieved by using classifiers which have to be previously trained, or otherwise “taught” — in case of Rules Classifier.

All your work on classifier training will be saved within a Definition. Each new training overwrites the results of previous training.

If you want to try different training strategies, you should do so in different Definitions. Later, you will be able to choose the one that works best for you.

Choose Classifier

At this point, you should decide which classifier you want to build up first. There are four different classifiers: Content, Layout, Combined, and Rules Classifier. You can access each of them using navigation buttons on the left pane.

Also, within Classifiers Summary section you can see a summary for the current definition. It gives information on whether a classifier is trained and, if yes, some parameters of the last training.

You can get redirected to individual classifiers by pressing “Modify classifier” button next to the classifier you want in classifiers’ summary pane.

When you start the training process, you can choose to train either Content or Layout classifier; Combined Classifier doesn’t have an independent training procedure, and it will be ready for use once Content and Layout classifiers have been trained. Rules Classifier doesn’t have automatic training as well, its work is based on the rules you write. Rules Classifier can be used as soon as you add rules to the list. However, if you want to use in your rules results from Content, Layout or Combined classifier for the rule conditions, the Content and Layout classifiers need to be trained, correspondingly.

If you want to classify documents by content, documents related to the same category should have similar contents.

If you choose to classify documents by layout, documents related to the same category should look alike or have common geometrical features. (see more detailed descriptions of the classifiers under corresponding chapters).

Combined classifier can only be used after Content and Layout classifiers have been trained. Combined gets the results from the latter two and uses a certain algorithm to choose the best answer between them.

Content Classifier

If you want to classify documents by content, documents related to the same category should have similar contents.

Content classification is based on a detailed analysis of the text content of a page. This classifier employs OCR output of documents. It performs a full-page recognition and the document is classified according to the text it contains. This method is most efficient if used for unstructured documents that typically have no organized graphical layout or are mostly composed of text. Examples of unstructured documents are contracts, legal documents, and reports.

For training Content classifier you will need to provide a classification set to learn on and specify the OCR and language (of the documents) parameters.

When the classification set you wish to use for training is ready, press the Train classifier button located on the toolbar under the Classifier Training and Testing tab.

Remember to make sure that the appropriate Classification Set tab is active when you start training.

A dialog box will appear, prompting you to specify the OCR and language parameters.

You can choose the OCR engine from FineReader, Tesseract, and Cuneiform in the drop-down menu.

Note: To use FineReader you will need a corresponding license to be installed on your computer.

Available languages include English, German, French, Russian, Portugese (Brazil), and Spanish (Latin America).

Note: To use these languages you need to have a corresponding license.

After you specify all the parameters in the dialog box, press the Run button, and training will begin. Green progress bar will appear at the bottom of the window, and the red Cancel button will light up on the toolbar. You can press this button to interrupt training. Once training is complete, you will receive a message “Training success”, and the status of the Classifier on the upper pane (and in classifiers’ summary) will be changed accordingly. Date and time of the last training session will appear in the Description box on the toolbar. You can edit this description for your convenience.

Every training overwrites the results of previous training.

The Clear button deletes current training results, and returns the classifier to non-trained state.

After pressing this button, you will need to train again before you can use the classifier. After you have trained your classifier, you can either start testing or train another classifier.

Testing

After the classifier has been trained, you can start testing it to see how it performs.

You can run tests either on the same classification set on which the classifier has been trained, or you can use a different classification set from your selection.

Start testing by pressing the Test button on the toolbar under Classifier Training and Testing tab. The red Cancel button will light up on the toolbar, and the green progress bar will appear at the bottom of the window. You can press the Cancel button if you want to stop testing.

Testing will run on the classification set from the Classification Set tab that was active at the moment the Test button was pressed. You can add all documents for which you don’t want to specify the truth value into the Images without class folder in the classification set.

After you run a test, you can see the classifier output in the classification set window. It shows class assignments for each image in the classification set and confidence values for these assignments:

A confidence value is a number between 0 and 100 indicating how confident the classifier is about classification answer. The higher the confidence value, the higher the probability that the answer is correct.

You can try adjusting certain parameters and run multiple tests to see how it affects the performance of the classifier. To facilitate this task, some parameters can be adjusted in the appropriate boxes on the toolbar in the Settings section.

> Preprocessing – You can select a preprocessing profile accompanied by Description which can be created in the Image Preprocessing menu item.

> OCR type – You can test how the choice of OCR engine affects the performance of the classifier. This parameter can be set independently for test mode and work mode. Test mode changes do not affect what happens while processing the workflow.

Note: The OCR engine you choose here is not the engine you chose for training the classifier, it is used for processing the test classification set. If you want to change the engine used for training, you will have to run the training again with a different OCR engine parameter.

Other parameters are shown for your information and cannot be changed.

To be able to better assess the results, you may go to the Results sub-tab, where you can see a summary and graphical representation of the Classifier performance, as well as adjust the Threshold.

Results Evaluation

After the testing is complete, you can go to the Results sub-tab and take a look at the statistics and graphic representation of the classifier performance.

A summary will be shown in the upper-middle section of your window, it will give you statistics on the number of pages processed, numbers of accepted and rejected pages, numbers of correctly classified and error pages, and percentages.

Only images that have truth values are accounted for in this summary. Accepted are the pages for which the classifier has an answer, rejected are those for which the classifier did not give a satisfactory answer, these pages were not classified.

In the right-hand side of your window you will see a 3D picture with a table showing Truth values and Classifier answers (results). Cells in this table represent combinations of Truth values and Classifier answers. The statistics is graphically represented by pillars of different color and size. Answers that coincide with truth values (located on the diagonal of the table) are represented by green pillars, where the height of the pillar corresponds to the number of such answers. Wrong answers are represented by red pillars, rejected answers – by yellow pillars.

In the menu on the toolbar, you can choose to represent absolute or relative value of the number of answers.

If you choose absolute, the height of the pillar will get greater when the number of answers increases. For relative the height of the pillar will correspond to the ratio of the answers versus the total number of images in the class.

You can also zoom in, zoom out, and rotate the graphics. You can always return to the original view by pressing the Default view button.

Images from all the classes in the Classification Set will have truth values, except Images without class folder, which lacks truth value and, therefore, automatic results evaluation is not possible for this class. To evaluate how the classifier performed on the images from the Images without class folder, you will have to manually sort the images and class assignments for them.

You will also see the list of images (by names) with class assignments and confidence values – below the statistics section. These can be saved to a comma delimited text file if you press Save results to file button on the toolbar.

Threshold

There will also be a section for tuning thresholds on the toolbar. A Threshold is a numeric criterion that Classification Confidence must meet or exceed to accept the class assignment.

By adjusting the threshold, you can change the number of rejected answers. All answers with confidence lower than the threshold will be rejected, the rest will be accepted.

If you are looking for an accurate classification output, you may choose to reject the answers with low confidence values, as the decision on their class assignment was more ambivalent.

Or, you may choose to accept all class assignments if your goal is the volume of processed documents.

The Threshold has two values – “test threshold” and “work threshold”. Changing the “test” value will take effect only for the current testing. These changes do not automatically transfer to the work mode of the classifier. You can change the test threshold using the slider above or by entering the desired number into the box.

After you have found the optimal threshold value by adjusting the test threshold, you can push the Set Threshold button to transfer the “test value” to work threshold or you can manually enter the value for work threshold.

Layout Classifier

If you choose to classify documents by layout, documents related to the same category should look alike or have common geometrical features Layout classification mainly uses geometrical features of the documents to detect the graphical similarity between document pages. It is based on the analysis of the overall layout and structure of a page, including lines, boxes, logos, and placement of text. This classifier is best used for documents with little text and visible differences between document types. Typical examples of such documents are structured forms; they have an explicit structure and organized layout.

The only parameter you need to specify when training the Layout Classifier is Classification Set. After the classification set is ready, and the correct tab in the Classification Set window is active, you can press the Train classifier button, and training will start immediately.

Green progress bar will appear at the bottom of the window, and the red Cancel button will light up on the toolbar. You can press it if you want to interrupt training. Once training is complete, you will receive a message “Training success”, and the status of the Classifier on the toolbar (and in classifiers’ summary) will be changed accordingly. Date and time of the last training session will appear in the Description box on the upper pane. You can edit this description for your convenience.

Now you can proceed to testing and viewing results.

Testing

After the classifier has been trained, you can start testing it to see how it performs.

You can run tests either on the same classification set on which the classifier has been trained, or you can use a different classification set from your selection.

Start testing by pressing the Test button on the toolbar under Classifier Training and Testing tab. The red Cancel button will light up on the toolbar, and the green progress bar will appear at the bottom of the window. You can press the Cancel button if you want to stop testing.

Testing will run on the classification set from the Classification Set tab that was active at the moment the Test button was pressed. You can add all documents for which you don’t want to specify the truth value into the Images without class folder in the classification set.

After you run a test, you can see the classifier output in the classification set window. It shows class assignments for each image in the classification set and confidence values for these assignments:

A confidence value is a number between 0 and 100 indicating how confident the classifier is about classification answer. The higher the confidence value, the higher the probability that the answer is correct.

You can try adjusting certain parameters and run multiple tests to see how it affects the performance of the classifier. To facilitate this task, some parameters can be adjusted in the appropriate boxes on the toolbar in the Settings section.

> Preprocessing – You can select a preprocessing profile accompanied by Description which can be created in the Image Preprocessing menu item.

> OCR type – You can test how the choice of OCR engine affects the performance of the classifier. This parameter can be set independently for test mode and work mode. Test mode changes do not affect what happens while processing the workflow.

Note: The OCR engine you choose here is not the engine you chose for training the classifier, it is used for processing the test classification set. If you want to change the engine used for training, you will have to run the training again with a different OCR engine parameter.

Other parameters are shown for your information and cannot be changed.

To be able to better assess the results, you may go to the Results sub-tab, where you can see a summary and graphical representation of the Classifier performance, as well as adjust the Threshold.

Results Evaluation

After the testing is complete, you can go to the Results sub-tab and take a look at the statistics and graphic representation of the classifier performance.

A summary will be shown in the upper-middle section of your window, it will give you statistics on the number of pages processed, numbers of accepted and rejected pages, numbers of correctly classified and error pages, and percentages.

Only images that have truth values are accounted for in this summary. Accepted are the pages for which the classifier has an answer, rejected are those for which the classifier did not give a satisfactory answer, these pages were not classified.

In the right-hand side of your window, you will see a 3D picture with a table showing Truth values and Classifier answers (results). Cells in this table represent combinations of Truth values and Classifier answers. The statistics is graphically represented by pillars of different color and size. Answers that coincide with truth values (located on the diagonal of the table) are represented by green pillars, where the height of the pillar corresponds to the number of such answers. Wrong answers are represented by red pillars, rejected answers – by yellow pillars.

In the menu on the toolbar, you can choose to represent absolute or relative value of the number of answers.

If you choose absolute, the height of the pillar will get greater when the number of answers increases. For relative the height of the pillar will correspond to the ratio of the answers versus the total number of images in the class.

You can also zoom in, zoom out, and rotate the graphics. You can always return to the original view by pressing the Default view button.

Images from all the classes in the Classification Set will have truth values, except Images without class folder, which lacks truth value and, therefore, automatic results evaluation is not possible for this class. To evaluate how the classifier performed on the images from the Images without class folder, you will have to manually sort the images and class assignments for them.

You will also see the list of images (by names) with class assignments and confidence values – below the statistics section. These can be saved to a comma delimited text file if you press Save results to file button on the toolbar.

Combined Classifier

Combined classification exploits both the methods of text reading and geometrical analysis of a document page. Combined classifier does not require separate training; however, it is based on a combination of Content and Layout classifiers that have been trained. It uses the output produced by both Content and Layout classifiers and selects the best result. Combined classifier is especially efficient if the workflow consists of a mixture of structured pages having distinctive graphic structure and pages of text that have no organized graphical layout.

Combined classifier will be ready to use as soon as Content and Layout classifiers have been trained. There is no separate training procedure, and the main functionality of this menu item is for running tests and studying results.

Testing and Results Evaluation

After Combined and Layout classifiers have been trained, you can select Combined Classifier on the Navigation pane and start testing it, to see how it performs, and evaluating results. (Refer to the Testing and Results evaluation sections).

Rules Classifier

Sometimes documents look alike or have similar content but still have specific characteristics based on which they should be referred to different classes. These may be: the presence of a company logo, signature, initials, or certain data fields. Automatic classification cannot provide accurate results, or is even inefficient in such cases. However, these elements can serve as “hints” that can be efficiently used for rules. That is why rule-based classifier is a necessary mechanism that, being used alone, or in combination with automatic classifiers, may ensure accurate classification of this type of documents.

Rule-based classifier operates with the following components: fields and automatic classifiers.

A rule comprises two parts: a Matching Condition and a Classification Action.

> Matching Condition is a condition that evaluates to true or false for a certain event. There are three different possible Matching Conditions: field is present, field has <value>, and classifier produced <result>.

> Classification Action is an action the classifier should execute if the Matching Condition is satisfied, i.e. assign a document to a particular class.

You can use these basic conditions to write simple rules such as “If field Name has the value ‘John’ then the document relates to Class B.” Some rules can be very exact such as “if field ‘Name’ is present in the upper-right hand quadrent of an image, then the document relates to Class B,” or more flexible such as “if field ‘Name’ is present anywhere on the document page then the document relates to Class B.”

It is also possible to combine these basic conditions using Boolean operations (And and Negation) to build more sophisticated rules, like “If field Name has ‘John’ and content classifier answered Class B then the document relates to Class B.” A rule may use any of the automatic classifiers (Layout, Content, or Combined). Each rule can use only one classifier of each type: one Layout Classifier, one Content Classifier, and one Combined Classifier. Herewith, Combined Classifier used in the rule has to be built on those layout and content classifiers that are already used in the rule.

Note: If you want to bind two conditions by the operator OR, you will need to write two separate rules – one for each condition.

Every rule is associated with a distinct image that has an integrated system of related fields located on it. Field is an area on the image intended to contain a certain type of data. The system of fields is specified on the image when creating a classification rule and it serves as a classification feature for a page. Actually, the image is a container for fields related to a particular rule. Fields may be of different types (formats), they can have a specific location on a page, field value — they can have a vocabulary of possible values that may be encountered in this field, and have many other characteristics. Relations between fields are defined explicitly: field A is located above field B or implicitly: two different fields cannot be located in the same area on the image.

Each rule has a unique individual set of fields and is not related or connected to another rule with its system of fields. Different rules can define the same class. The classifier evaluates rules in the set of rules, chooses the best matching rule, and assigns class to the document based on this rule.

Once an image satisfies a rule condition, it is assigned to the specified class, and other rules are not computed on that image. So, to increase processing speed, you are recommended to place rules with simpler conditions before more complicated rules.

Edit sub-tab

To write or edit simple rules that don’t employ fields you can go to the Edit sub-tab of the middle window. Rules editor is conveniently designed to facilitate the process of creating a rule.

To create a new rule click on the “edit” icon (with a pencil drawn on it), which you will see in the right-hand pane.

Then, a template will appear with a few boxes that you will have to fill. A name will be automatically assigned to each new rule, you can later change it.

Under the “Name” box you will see “Class name” box. This is to determine to which class the page will be assigned to as a result of rule application.

The following boxes are used to specify rule conditions.

If you want to use answers from automatic classifier click on the “Select type” box (under Type column), then click on the down arrow to see the options, and choose “Classifier” in the drop-down menu. Two more boxes will unfold where you will specify which classifier you want to use and the class answer that you seek. Also, to the left of the “Type” column, there is “Not” checkbox. You may check it if you want to apply negation to the condition.

The “Enabled” checkbox is checked by default; you may uncheck it, and the rule will not be calculated during classifier operation.

Important! If you plan to use Rules Classifier for every class that you want in the resulting classification you should create at least one rule that returns this class. The results of the  Content, Layout, and Combined classifiers will NOT be automatically translated when you use Rules Classifier. Only the classes present in rule definitions of enabled rules will appear in the final classification.

Create Rules for All Classes

This button is accessible in the Rules Classifier under Classification Sets sub-tab and Managing Classes tab on the toolbar.

If you press Create rules for all classes button, rule templates will be created for each class in the current classification set that assign these classes. Also, the first image corresponding to the appropriate class in the Classification Set will be added to the rule to use for fields.

After pressing Create rules for all classes button you will be automatically transferred to the Edit sub-tab in the Rules Classifier menu. However, no conditions for the rules are supplied automatically. You will have to specify the conditions if you want to use the templates and create rules using them. Such conditions may be very straightforward, like: if Combined Classifier answered “Class A”, assign “Class A”.

It is convenient to use this functionality to ensure that you don’t forget any classes in the Rules Classifier.

Fields

Besides using the results of automatic classifiers, Rules Classifier can also access and employ data obtained by text recognition of certain page fragments – fields.

If you plan to use more complicated conditions in your rules that employ fields, you may start creating your rule from the Fields sub-tab directly. Or, if you have previously created a rule, you may continue editing it in the Fields sub-tab.

Fields sub-tab is located to the right of the Edit sub-tab in the Rules Classifier. To create fields and rules with fields activate the Fields Layout tab on the toolbar.

You will see all previously created rules in the Classification Rules view pane.

You can choose one of these rules, or create a new rule by pressing the Add rule button.

A new rule will appear in the Classification Rules section of your window. It has the following sections (columns):

> Enabled – Whether the rule should be used by classifier.

> Image – Shows whether an image is attached to the rule. If you click on the image icon, the image will appear in the right view pane.

> Rule name – Double click on this section to edit rule name.

> Class name – Name of class that the rule will return. When you click on this section you will see a drop-down menu of classes that are present in your classification sets.

If you right-click on the rule line a context window will appear with the list of actions that repeat the actions in the Rules section on the toolbar (under Fields Layout tab).

After adding a rule you will need to attach an image to it. Click to select the rule line, press Add Image button from the toolbar. Choose an image from your files.

After you add an image, it will appear in the view pane to the right. After you attach an image to the rule, you can start creating fields.

You can also use Create Rule by Selected Image button, which is found when you choose Classification Sets sub-tab and Managing Classes toolbar tab in Rules Classifier. Before pressing the button, you can select an image either from the Image pool or Classification Sets. A rule with the image attached will be created; you will see it if you go to the Edit or Fields sub-tab.

To create fields on the image, press Field button from the Fields Layout toolbar.

Select the desired field type from the menu list of field types. There are over 20 commonly encountered field-types available so that you can easily select a proper field type for any field that is encountered on your documents. When creating Address Fields you also have to select the proper type of field that matches the format of Address field on your documents.

Using your mouse, mark out a rectangular area on the document image to indicate the field location.

Note: After you select the field type, Cancel button will light up on the ribbon. If you change your mind, you can press the button before marking out field location and the field will be deleted.

After you mark out the field area, you can delete, move or copy field to a different rule and image by using the clipboard buttons on the upper ribbon.

Properties window will appear to the right of the image view pane. (If you don’t see it, try expanding the pane to the left).

Fields Properties

Every field you create will have to be recognized on your pages during classification. Recognition process is a complex process which involves many parameters that you will need to specify as Properties when creating a field.

Field properties are grouped into the following thematic units (in the order they appear):

> Classification – Specify how the field is used in the rule condition.

– Include in Rule condition – Specifies whether the field is used for the rule condition. (A field may be needed to determine relative location of another field. In this case it may not be used in rule conditions).

– Negation – Select “Yes” if you want to apply the NOT operator to the rule condition. Otherwise, select “No”

– Field value – Specify the value for the field.

> General Properties – You may specify the following general properties:

– Name – Specifies the field name. The property length must be at least 1 character. The property length must not exceed 32 characters.

– Description – Specifies the field description. This property is for your reference only. The property length must not exceed 255 characters.

– Enabled – Specifies whether the field should be processed (Yes) or not (No).

– Keying only – Specifies whether the field should not get recognized but keyed only.

> Location Properties – You may specify the following location properties:

– Position – Defines the type of field position (may be Fixed, Dynamic, Recipient Address, Sender Address, Boxed or Payor Block).

» Fixed – the position of field has fixed coordinates on every page.

» Dynamic – the coordinates of the field on different pages may vary.

» Recipient address – Locate recipient address block.

» Sender address – Locate sender address block.

» Boxed – The text is printed within the confines of special boxes or combs.

» Payor Block – Locate payor block.

If you select Dynamic from the drop-down menu for Position property, the system automatically expands Location Properties folder. The following additional properties appear:

– Location Type – Specifies the type of dynamic field location:

» Keyword and value – No and not empty Keyword vocabulary.

» Keyword only – Yes and not empty Keyword If this type is set, than some properties (value position, value vertical proximity, value horizontal proximity) disappear from the list of properties.

» Value only – No and empty Keyword vocabulary.

» Anchor –

» Optional keyword –

– Location Area – A position of the field on the page. Specifies one of the following regions: Whole page, Right half, Left half, Top half, Bottom half, Top-left quadrant, Top-left quadrant, Bottom-left, Bottom-right, Inside field zone. Meaning of the regions:

» Whole Page – The field can be located anywhere on the page.

» Right Half – The field is located in right half of the page.

» Left Half – The field is located in left half of the page.

» Top Half – The field is located in top half of the page.

» Bottom Half – The field is located in bottom half of the page.

» Top-Left quadrant – The field is located in top-left quadrant of the page.

» Top-Right quadrant – The field is located in top-right quadrant of the page.

» Bottom-Left quadrant – The field is located in bottom-left quadrant of the page.

» Bottom-Right quadrant – The field is located in bottom-right quadrant of the page.

» Inside field zone – The field is located inside the field zone defined by user on template image. The zone rectangle coordinates are scaled to the page image’s coordinates.

– Answer priority – Specifies which answer to select as the best from competing candidates. Specifies one of the following priorities: Highest confidence, Rightmost, Leftmost, Topmost, Bottommost, Longest. The Longest value is used if the dynamic field contains more than one word or defines the longest answer.

– Required – Specifies whether the field and its value must be present on the page.

– Multiline keyword – Specifies whether keyword can take several text lines.

– Keyword vocabulary – A vocabulary that contains field keywords. Here you can create a new Vocabulary or DB Vocabulary, or select one of existing vocabularies which can be edited. Using vocabularies allow you to increase the recognition quality.

– Keyword writing style – Defines one of the following writing styles for the keyword – Machine-print, Hand-print.

– Field relations – This field defines relationship rules between this and other fields on the page. Each rule contains name of related field, type of reference element (Keyword or Value) for this and related fields, distance between reference elements, and their relative position.

– Keyword reject vocabulary – A vocabulary that contains words that cannot be field keywords.

– Keyword language – It defines the language for dynamic field’s keywords. All dynamic fields on the page should have the same language for keywords. You can specify one of the following languages: English, German, French, Russian, Portuguese (Brazil), and Spanish (Latin America).

Note: To use these languages you should have corresponding license.

If you set Keyword and value in the Location type, the following additional properties appear in Location properties:

– Value position – A position of the field value relative to its keyword. May be one of the following: N (top position from keyword), E (right position from keyword), S (bottom position from keyword), W (left position from keyword) SW (bottom-left position from keyword), SE (bottom-right position from keyword), NW (top-left position from keyword), NE (top-right position from keyword), I (intersects with keyword), and SM (in the same cell with keyword). You can select several values.

– Value Vertical Proximity – Expected distance between field value and its keyword in vertical direction. Minimum and Maximum distances are measured by numbers of keyword rectangle height. If distance between closest edges of located keyword and value will be outside of specified range then the dynamic field answer will have low confidence. To define Minimum and Maximum distance click this field and set the corresponding values. You can leave this property blank.

– Value Horizontal Proximity – Expected distance between field value and its keyword in horizontal direction. Minimum and maximum distances are measured by numbers of keyword rectangle height. If distance between closest edges of located keyword and value will be outside of specified range then the dynamic field answer will have low confidence. To define Minimum and Maximum distance click this field and set the corresponding values. You can leave this property blank.

Note: You can specify several rules per each part of fields. Rules are combined using disjunction (OR) rule.

Click Field relations to display the table in the top part of a window. The table with the relation rules has the following columns:

» Field Element – Defines the type of reference element (Keyword or Value) of this field. To define type of reference element click this column and select Keyword or Value from the drop-down list.

» Related Field – Defines related field. To choose related field, click this column and select the wanted field from the drop-down list.

» Related Field Element – Defines type of reference element (Keyword or Value) of related field. To define type of reference element click this column and select Keyword or Value from the drop-down list.

» Horizontal Distance – Defines Minimum and Maximum distance between reference elements of this and neighbor fields in horizontal direction. The distance is measured by the number of keyword rectangle widths. If distance will be outside of specified range, this element will not be available. To define Minimum and Maximum distances click this column and insert corresponding values. You can leave this column blank.

» Vertical Distance – Defines Minimum and Maximum distance between reference elements of this and neighboring fields in vertical direction. The distance is measured by the number of keyword rectangle heights. If distance will be outside of specified range, this element will not be available. To define Minimum and Maximum distance click the column and insert corresponding values. You can leave this column blank.

» Position – A position of the Related Field Element to Field Element. May be one of the following: N (top position from Field element), NW (top-left position from Field element), NE (top-right position from Field element), E (right position from Field element), S (bottom position from Field element), W (left position from Field element), SW (bottom-left position from Field element), SE (bottom-right position from Field element).

> Image Cleanup and Preprocessing – Specify operations that have to be performed on an image before recognition. These include image rotation, filtering to remove extraneous elements like lines, combs, boxes, or intrusions from field images, and also the aligning of field images. These properties are available for modification in the Fields and Recognition windows.

– Rotate image: Specifies whether the field must be pre-rotated before recognition and the rotation angle desired. Select from the following options:

» No rotation and no flipping: Fields are recognized as is.

» 90° rotation without flipping: Field image will be rotated 90 degrees clockwise before recognition.

» 180° rotation without flipping: Field image will be rotated 180 degrees before recognition.

» 270° rotation without flipping: Field image will be rotated 270 degrees clockwise or 90 degrees counter-clockwise before recognition.

» No rotation followed by a horizontal flip: Fields are flipped horizontally.

» 90° rotation followed by a horizontal flip: Field image will be rotated 90 degrees clockwise and flipped horizontally before recognition.

» 180° rotation followed by a horizontal flip: Field image will be rotated 180 degrees and flipped horizontally before recognition.

» 270° rotation followed by a horizontal flip: Field image will be rotated 270 degrees clockwise and flipped horizontally before recognition.

» No rotation followed by a vertical flip: Fields are flipped vertically.

» 90° rotation followed by a vertical flip: Field image will be rotated 90 degrees clockwise and flipped vertically before recognition.

» 180° rotation followed by a vertical flip: Field image will be rotated 180 degrees clockwise and flipped vertically before recognition.

» 270° rotation followed by a vertical flip: Field image will be rotated 270 degrees clockwise and flipped vertically before recognition.

» Default – No rotation and no flipping.

– Remove boxes and/or combs: Specifies whether the boxes or combs are deleted from field image. Select from the following options:

» No: Removal of boxes and combs is turned off.

» Remove box: Removes boxes from fields.

» Remove box and clean around the box: This option removes boxes and other lines, words, or intrusions that occur outside of the boxes.

» Remove combs: Removes combs from around individual characters.

» Default – No.

– Remove lines: Specifies whether the horizontal and/or vertical lines are deleted from field image. Select from the following options:

» No: Line removal is turned off.

» Remove vertical lines: All vertical lines will be removed from the field image.

» Remove horizontal lines: All horizontal lines will be removed from the field image.

» Remove vertical and horizontal lines: Both vertical and horizontal lines will be removed from field image.

» Remove vertical and horizontal dash lines: Both vertical and horizontal dash lines will be removed from field image.

» Default – (No).

– Remove template: Specifies whether the sample image template will be removed from the field image. Turns on (Yes) and off (No) the removal image template. Default – (No).

– Remove pre-printed text: Turns on (Remove from top or Remove from bottom) and off (No) the removal of pre-printed text and other pre-printed information from the field image. You must have a blank sample image to be used as a pattern for image cleaning when removing preprinted text. This method of cleaning is highly effective for field snippets that include pre-printed material. If a blank sample image is unavailable, you can use the built-in image editor to clean all written information from the sample image. Default – (No).

– Remove noise: Specifies whether the small isolated noise and intrusion lines are deleted from field image. Turns on (Yes) and off (No) the removal of small noise objects from the field image. Noise objects are stray lines or curves that extend into the snippet area from above, below or the sides. Default – (No).

– Align text: Specifies whether the field text is horizontally aligned before recognition. Turns on (Yes) and off (No) the horizontal alignment of text in a field image. If the mode is on, places the image on a straight, horizontal plane. Corrects writing that slants upward or downward. Default – (No).

Note: Setting Horizontally align field text option to Yes may lead to misinterpretation for some barcode types if those barcodes are rotated to some specific angles.

– Auto: Automatically detects flaws and enhances images. Turns on (Yes) and off (No).

> Recognition – Specify for the recognizer what information to expect in the field (style of writing, date format, value limit, etc.) and how to recognize the field (with or without vocabulary, whether to use an alias table, database, etc.). The set of properties available for a field depends on the field type. Only properties available for a particular field type are displayed in the Properties of window for the corresponding field. Recognition properties are available for modification in Fields and Field Recognition sub-tabs. The list of Recognition Properties is provided below.

– Field type – The field type is a primary description of field content. A field type specifies the format of the field, the set of characters that may occur in the field, and therefore the way it will be recognized. Thus, it is important that you choose an appropriate field type for each field that is encountered on your documents. Select from these values:

» Address: The field contains a USA address in various formats.

» Age: The field contains a person’s age.

» Alpha: The field contains a generic alpha text.

» Alpha-numeric: The field contains a generic alpha-numeric text.

» Amount: The field contains a dollar amount (including cents).

» Barcode: The field contains a barcode.

» Checkbox: The field contains a checkbox.

» Credit card number: The field contains a number of credit card.

» Date: The field contains a date in various formats.

» Empty: The field is recognized as blank or non-blank.

» First Name: The field contains a person’s first name.

» Last Name: The field contains a person’s last name.

» Full Name: The field contains a person’s full name.

» Middle Name: The field contains a person’s middle name initial.

» Length: The field contains a length measurement expressed in feet, inches, or both.

» Measure: The field contains a numeric value and an abbreviation (such as “115 lbs.”). The measurement unit (lbs., , kg., mm., and so on) should not exceed four characters.

» Numeric: The field contains a generic numeric text.

» Phone number: The field contains a USA phone number.

» Postnet Barcode: The field contains a USA PostNet barcode.

» Social Security Number: The field contains a USA social security number.

» Default – Alpha. The default value set for the page is used for the field.

Notes:

> Russian language is supported for the following field types: Alpha, Alpha-numeric, First Name, Last Name, Middle Name, and Numeric.

> French language is supported for the following field types: Alpha, Alpha-numeric, and Numeric for Cursive writing style.

– Language: The following languages are supported: English, German, French, Russian, Portuguese (Brazil), and Spanish (Latin America). To use these languages you should have corresponding license. When working with the Date or Numeric field types you are allowed to select one of the available languages – English, German, Portuguese (Brazil), or Spanish (Latin America). The last two languages are also supported by Alpha and Credit card number fields.

Because recognition of German numbers is only realized for hand-written text, an error of recognition arises, if you simultaneously select German for the Language property and Machine-Print for the Writing Style property: The current combination of language and recognition settings is not supported.

The same error arises, if you simultaneously select German for the Date field type and define one of the following date formats: DD MMMM YYYY or MMMM DD, YY. Default – English.

– Writing style: Specifies the field text writing style. Select from these values:

» Cursive: The text is handwritten; the letters in the words are generally linked together.

» Hand-print: The text is hand-printed without any constraints.

» Boxed hand-print: The text is hand-printed within the confines of special boxes or combs.

» Machine-print: The text is machine-printed.

» Auto-detect: The style of writing is unknown. The application will automatically detect the style of writing by selecting between machine-printed and unconstrained hand-printed styles.

» OCR-A: This style is one of machine-printed styles, which have the good recognition results. It is available for the following field types: Numeric, Alpha, Alpha-numeric, Phone, Date, Social Security Number, First Name, Last Name, and Full Name.

Notes: Machine-print and OCR-A writing styles are not supported for Russian language recognition. French language is supported for the following field types: Alpha, Alpha-numeric, and Numeric for Cursive writing style.

» Default – Hand-print.

– Speed/Accuracy: Selects whether the speed or accuracy is a priority during recognition.

» Accuracy: The recognition is more accurate.

» Speed: The recognition takes less time.

With the speed setting, recognition is faster but might not be as accurate, especially for fields containing names written with cursive or unconstrained handprint. With the accuracy setting, the recognition process includes an additional step, which makes it slower but more accurate. Consider using the accuracy mode for first name, last name, and full name field types and the speed mode for other field types. Default – Accuracy.

– Vocabulary (1): Specifies the recognition vocabulary. Using a vocabulary – even when it’s optional – can significantly improve recognition accuracy. Default – Not Set.

– Vocabulary (n): Specifies additional vocabulary files to be concatenated with the main vocabulary. This mechanism simplifies the use of large vocabulary files. It eliminates the need to physically create a large vocabulary. Instead it allows you to work with smaller vocabulary files, which contain the components of the words from the desired vocabulary.

Additional vocabularies are used to recognize fields, which could contain more than one word. If, for example, you have a field “City, Country”, you could set vocabulary (1) to the list of Cities, and vocabulary (2) to the list of Countries.

Select one of the vocabulary files defined in the object table. You can specify up to 4 additional vocabulary files. Default – Not Set.

– Vocabulary coverage: Specifies the percentage of possible field values, which should match those in the vocabulary. Valid property values range from 0 to 100. The higher the percentage, the higher the probability that a recognition answer will be a word from the vocabulary. A value of 100 means that only vocabulary answers are allowed. By shifting the property value towards 0, you allow more non-vocabulary answers. Default – Not Set.

– Alias table: Specifies the recognition alias table. After recognition, all recognition answers are converted from aliases to their standard form. Select one of the alias tables defined in the object table. Refer to the section Using the Alias Editor to learn how to select an existing alias table or create a new one. Default – Not Set.

– Spell-checker Vocabulary: Checks data extracted for spelling.

» Create new Vocabulary: Enter data directly into studio for spell checking.

» Create new DB Vocabulary: Enter data source, table, and words into configurations.

– Special charset: Specifies additional characters to be added to the field default character set (alpha, numeric, or alpha-numeric). Type here any characters you want to recognize within this field (comma, period, etc). For example, add a hyphen to the special character set when recognizing account numbers, which are otherwise a numeric field type. Default – Not Set.

– Template: ILINX Advanced Capture allows you to define field format through a template. A template is a string of characters and symbols that represents both an input and (partially) an output string. Templates may include any of the following special characters:

# matches numerals

@ matches alpha characters

$ matches special symbols

* matches any symbol

The set of symbols depends on field type. For example, if you are recognizing insurance application numbers that always consist of three alpha characters followed by four numeric characters, you would specify a template of @@@####. Default – Not Set.

– Regular expression: Specifies whether the field template is the regular expression. Exclude charset: Specifies additional characters which cannot be used in the current field. This character set is enabled for Alpha-numeric fields only. For example, it can contain the letter “O” to avoid confusion with digit “0”.

– Valid length: Specifies the valid range of lengths of recognition answers. Enter the minimum and maximum answer length in the corresponding boxes that will unfold.

Notes: If a certain length is set, and the recognizer finds that the length of the current field is less than this value, the result is considered confidently empty. Property ‘Valid length’ affects the acceptance of recognition results.

» alid values: Specifies the valid range of values of numeral recognition answers. Enter the minimum and maximum value in the corresponding boxes that will unfold.

– Multi-line zone: Can be set for fields that contain a few lines of machine-printed text, for example when the name and address are typed on separate lines. Specifies what line (or lines) within the field must be recognized. It works best for machine-printed text, but could also be used for any multi-line field. Select from the following possible options:

» Top: Only the top line of the multi-line field will be recognized.

» Second: Only the second line of the multi-line field will be recognized.

» Third: Only the third line of the multi-line field will be recognized.

» Fourth: Only the fourth line of the multi-line field will be recognized.

» Bottom: Only the bottom line of the multi-line field will be recognized.

» SecondFromBottom: Only the second line from the bottom line of the multi-line field will be recognized.

» ThirdFromBottom: Only the third line from the bottom line of the multi-line field will be recognized.

» FourthFromBottom: Only the fourth line from the bottom line of the multi-line field will be recognized.

» AllBeneathTop: All the lines located below the top line in the multi-line field will be recognized.

» AllAboveBottom: All the lines located above the bottom line in the multi-line field will be recognized.

» DefaultOrAllAboveBottom: Default option.

Note: The options AllBeneathTop, and AllAboveBottom make sense for address fields only.

– Threshold: Specifies the recognition threshold. Recognition answers whose confidence levels are equal or higher than the threshold are accepted, other answers are rejected. The value must be within 0 and 101 range. Default – 101.

> OCR properties – Specify how third party OCR results should be merged with the recognition answer.

– Merge mode: Specifies the merge mode. May be one of the following:

» Ignore: For recognition of field is only used internal engine. This is default value.

» Replace: For recognition of field is only used page OCR.

» Merge: For recognition of field is used internal engine as well as page OCR and then to merge results.

> Validation properties – Specify parameters of validation for the field recognition results. The parameters apply to recognition answers in the order they appear in the Validation section in the Properties window. Validation properties are available for modification in Fields and Recognition windows. The list of Validation Properties is provided below.

– Remove charset: Specifies characters to be removed from the field answer after recognition. Default – Not Set.

– Alias table: Specifies an alias table to be used to convert recognition results. Select one of the alias tables defined in the object table, edit it or create a new one. Default – Not Set.

– Allow blank fields: Specifies whether the blank recognition answers are allowed or not. If set to No and the field was recognized as empty, the field will be rejected. Default – Yes.

– Valid values: Specifies a vocabulary that contains valid field values. If a recognition answer is not in this vocabulary, the field will be rejected. Select one of the vocabulary files defined in the object table, edit it or create a new one. Default – Not Set.

– Invalid values: Specifies a vocabulary that contains invalid field values. If a recognition answer is in this vocabulary the field will be rejected. Select one of the vocabulary files defined in the object table, edit it or create a new one. Default – Not Set.

> Verification properties – Verification properties specify parameters which tune process of keying.

– Document field: Specifies whether the field data may be entered from any page in the document.

– Keying type: Defines the type of keying that applies to all fields on the page.

– Double keying mode: A number of operators that have to process fields on this page. The value may be Single, Double or Triple.

– Enlarge field image: Specifies whether the field visible area should be extended by 50% in each direction for the field keying.

– Removed white margin: Specifies whether the white margin should be removed from the field image during field keying.

– Keyer tip (field keying):  Defines a help string to display to the operator during page keying.

– Prompt, if cannot validate: Specifies whether the dialog box prompting invalid field value should be displayed to keyer when his/her answer cannot be validated.

– Template: A field keying template.

– Valid length: A minimum and maximum length of the keying answer.

– Valid values: A minimum and maximum value for the numeric keying answer.

– Vocabulary: A vocabulary to use during keying.

– Threshold: A keying threshold. Shows recognition answer to keyer if recognition answer equals current field answer and confidence level is greater than or equal to the threshold.

Field Recognition

Under this tab you can see the recognition results and statistics for a selected field on all images from your classification set. You will see a table with file names, snippets of fields, confidence value and recognition answer.

Field properties, except for Classification properties, are also accessible from this tab. You can edit them and see how it affects recognition right away.

Test bag

Every time you switch to Field Recognition sub-tab, the recognition process is triggered. It may take some time. If you don’t want to wait, and a small number of images is enough for you to tune the recognition of the field, you can take a small subset of images to the Test bag. To do so, go to the Field Recognition sub-tab (this time you will have to wait while the whole classification set is getting recognized), once the recognition is complete and recognition information appears in the middle pane, select the images you want to take to the test bag; right click on them – a context menu will appear, choose Send to Test Bag option.

Copies of selected images will be placed into the test bag. Test Bag can be accessed by switching the tabs (Set and Test Bag) at the bottom of the view pane.

Now, after you go to the Field Recognition sub-tab, you can interrupt the recognition process by switching to the Test Bag tab, then, only the images from the Test Bag will be recognized.

Scripts

In any document definition VB.Net and C# code may be added to enhance classification. The process of creating document definition includes writing, testing and running the script.

To start writing scripts go to the Scripts sub-tab. The system provides the skeleton of your future script that will help process recognition and classification of your documents. The skeleton of your script contains the Main class (you are not recommended to change it) and several methods for processing events which can be used in your script.

For more information regarding Scripts within the Rules Classification refer to the ILINX Advanced Capture Technical Guide.

Clustering

In real life it is often difficult, if possible at all, to come up with predefined categories and create a predefined classification set. For example it is quite typical that organizations accumulate vast numbers of unsorted documents, so that it is very time-consuming or infeasible to specify, even roughly, the types of documents and, therefore, to create a classification set to train classifiers.

Clustering is a purely automated process that does not require training or specifying any criteria for sorting. You only need to provide a set that contains a mixture of unorganized documents that may be encountered in the real life stream or in an archive and have to be classified.  The system will analyze the document features and produce an organized set where documents will be grouped into clusters based on geometrical or text similarity.

As a result of clustering a number of document clusters are created, where each cluster contains a certain type of documents. These clusters can serve as rough prototypes for future classes and may be used to train the classifier.

At this point you may want to edit the results of clustering by giving appropriate names to the resulting clusters, combining two clusters into one class, etc.

When the structure of clusters has been refined, you can use it for the Classifier training by transferring clusters into Classification Set.

How to use clustering

If you decide to define your Classification Set with the help of clustering, fill the Image Pool with all the relevant page examples (images). It doesn’t matter if the images receive different labels (based on the names of the folders from which the images were taken), these labels will be ignored by the Clusterizer and new labels will be assigned as a result of clustering.

Once you have collected all the pages within the Image Pool, activate the Classifier Training and Testing upper tab and press the Clusterize button.

Image pool can be accessed if you choose one of the four classifiers on the Navigation pane, and Classification Sets sub-tab. Clustering is available only for Content and Layout classifiers, as it is the procedure intended to precede automatic training.

After you press the button a dialog box will appear prompting you to specify clustering parameters. For Content classifier you will have to specify the language of the documents and OCR engine.

Note: Language and OCR engine parameters are needed only if you use clustering for Content  classifier.

You can also choose the number of classes into which you want to divide your documents. The default setting for the number of classes is “Auto”, it means that Clusterizer will make a decision on the appropriate number of classes. If you want to specify the number of classes, you can uncheck the “Auto” setting and enter the number of classes into the box above the checkbox (this number should be greater than one)

Note: Even if you enter a specific number, you are not guaranteed to get that exact number of classes, as the algorithm takes into account many other parameters, too.

After you have specified all the parameters press Run, and clustering will start working. Red Cancel button will light up on the upper ribbon and green progress bar will appear at the bottom. You can press Cancel to interrupt clustering.

When clustering is complete you will see a message “Clustering success”, and labels will be assigned to every image in the Image pool.

Now you can evaluate clusterizer output, edit class (label) assignments, if necessary, and create your Classification Set by translating Image pool labels to Classification Set classes. (Use the buttons on the Managing Classes tab of the upper ribbon to move or copy images to Classification Set according to the label).

---

Registration & Classification

Registration is the process by which form images are identified as matching the sample image(s) used for the form definition. This process will reject any form images that are not images of the correct form type as well as images that are so skewed or so noisy that they cannot be recognized as the correct form. A registration zone is set on a sample image and is used as a landmark for defining all other elements on the form. The registration zone should contain information that remains constant on all forms in a batch.

Important! From registration point of view there is no difference between structured and semi-structured pages. The same registration technique is applicable to semi-structured pages as they are applicable now to structured pages.

Notes: If no registration zone is defined, the entire image will be used for registering images in the batch. For larger or more complicated forms, this may significantly slow form processing and in some cases may even cause registration to fail.

Each Sample Image can have multiple registration zones. Using at least one registration zone is highly recommended, while using two may increase accuracy, especially if placed in different locations on the form. For a long or complicated form, consider creating registration zones for each section of the form.

Creating and verifying a registration zone involves the following:

> Setting up the registration zone.

> Processing the images.

> Checking to ensure that most or all of the images were registered successfully.

> As necessary, modifying the registration zone or creating additional or different zones until most or all of the images in a batch are registered successfully.

All operations related to setting registration zones have to be done and are displayed in the Image Preview window. Form Registration window allows you to do the following:

— Designing the Form Definition

— Setting Up and Modifying the Registration Zone

— Setting Form Definition Registration Properties

— Setting Page Registration Properties

— Working with Test Images

— Setting Sample Images

— Using Recognition for Registration

— Working with Drop-out Template

Registration Screen

To begin registration process select Registration from Navigation pane. After registration the ILINX Advanced Capture Studio will look as follows (the Registration tab is selected).

Except of Navigation pane this screen consists of the following panes:

> Tool Bar – Appearance of Tool Bar depends on the window selected on Navigation pane. For information about Tool Bar for Registration and Image tabs of the Registration window see the Registration Window Tools.

> Form Definition Layout – Contains the structure of your form definition.

> Test Images – This window shows test images prepared and registered for recognition.

> Properties – Contains properties of form definition, document or page selected in the Form Definition layout pane (form definition on the picture).

> Image Preview – Displays the image selected in the Test Images pane.

Generally speaking ILINX Advanced Capture Studio allows you to use multiple test image decks. Their purpose is to facilitate definition tuning in Recognition window. One of the possible scenarios is to have two decks with test images, one small and one large. Small deck can be used for quick (course) tuning of the definition and large deck for more fine tuning, thus saving overall tuning time.

Multiple decks are available in Test images section of Registration window and Snippets section of Recognition window. Three decks are supported now. Images can be added to any of them as needed. You can register images in three different decks.

Setting The Registration Zone

ILINX Advanced Capture Studio offers primary (binary) and additional registration zones to be set for still more efficient form registration. A binary registration zone is always used for form registration. Registration zone may include any type of text, company logo or other graphic element that is present and remains constant on all forms in a batch.

To set a binary registration zone:

Step 1: Select the Registration window on the Navigation Click the Anchor zone button on the Tool Bar. The cursor changes to crosshairs.

Step 2: Using the image displayed in the Image Preview window put the crosshairs in the upper left corner of the desired

Step 3: Hold down the left mouse button, drag a marquee around the registration zone area, and release the mouse button.

Step 4: A registration zone, highlighted in blue, is created on the image.

Besides the regular registration zone, you can add more registration zones. To do this, repeat the current operation.

Additional registration zones are used in combination with the primary registration zone to speed up the process of registration and make it more reliable. Using advanced zones is especially efficient when dealing with a big batch of forms including different types of layout (clones).

Notes:

1. A couple of binary zones located in different corners of the Sample Image will provide better accuracy.

2. Usually, some preprinted text works better as the binary registration zone.

Running the Form Registration

After setting the registration zone(s), you can run the process of registration.

To run the process of registration:

Step 1: To process multiple images, click the Register button on the Tool Bar. All selected images are processed.

Step 2: Registration results appear at the bottom of the screen, just above the Status Bar.

Checking the Registration Results

To check the process of registration:

Step 1: Make sure that most or all images were registered successfully. A registered image is identified by a green check mark to the left of its name. Images that did not register have a red cross mark to the left of their names. The Registered column displays the results of form registration for every test image in the following format: Name | # | Document | Page | Template | Class.

Step 2: If you are not satisfied with the results, go to the next section of the document to modify the registration zone(s), create an additional zone, or create a new zone.

You can also use truth information to test registration performance and error rate. Truth information is the presumably correct data about form registration and field recognition. These data are stored in an ASCII file. Using truth data for registration allows you to verify that registration was completed correctly. During registration the information stored in the truth file is compared to the actual registration data obtained for every test image and the results of the comparison are displayed on the screen.

Modifying the Registration Zone

If for some reason registration failed for all or most of the images, it is recommended that you modify the registration zone.

To modify the registration zone:

Step 1: To modify the existing registration zone, select the registration zone on the sample image and change its location or size and shape.

Step 2: To add another registration zone, follow the instructions in the section Setting Up the Registration Zone above.

Step 3: To delete the existing registration zone: right-click on the registration zone and select Delete on the Tool Bar.

Step 4: To create a new zone, follow the instructions in the section Setting Up the Registration Zone above.

Step 5: To clear the previous registration results, click appropriate icon on the Tool Bar.

Step 6: After modifying or adding a registration zone, process the images To do this, follow the instructions in the section Running the Form Registration above.

Step 7: Once you have created registration zones and processed the test images successfully, you can proceed with the next steps of designing the form definition.

Registration Properties

You can preview and specify registration properties for the whole form definition and for a specific form page.

Form Definition Properties

To preview and specify form definition registration properties for a certain form definition, click on the form definition name in the Form Definition Layout window. Registration properties will be displayed on the Form Definition tab of the Properties window. Figure 4-4 shows the Form Definition registration properties:

You can fill out the fields directly in the window. To do this, double click the corresponding field. You may specify the following form definition properties:

General Properties

> Name: Specifies the form definition name. The property length must be at least 1 character. The property length must not exceed 32 characters.

> Description: Specifies the form definition description. This property is for your reference only. The property length must not exceed 255 characters.

> Author: Specifies the name of the person who created the form definition. This property is for your reference only. The property length must not exceed 32 characters.

> Company: Specifies the name of the company this form definition was created for. This property is for your reference only.

> Version: Specifies the version of a form definition.

Image Pre-Processing Properties

> Rotate and Flip: Specifies whether the field must be pre-rotated before recognition and the rotation angle desired. Select from the following options:

– No rotation and no flipping: Fields are recognized as is.

– 90° rotation without flipping: Field image will be rotated 90 degrees clockwise before recognition.

– 180° rotation without flipping: Field image will be rotated 180 degrees before recognition.

– 270° rotation without flipping: Field image will be rotated 270 degrees clockwise or 90 degrees counter-clockwise before recognition.

– No rotation following by a horizontal flip: Fields are flipped horizontally.

– 90° rotation following by a horizontal flip: Field image will be rotated 90 degrees clockwise and flipped horizontally before recognition.

– 180° rotation following by a horizontal flip: Field image will be rotated 180 degrees clockwise and flipped horizontally before recognition.

– 270° rotation following by a horizontal flip: Field image will be rotated 270 degrees clockwise and flipped horizontally before recognition.

– No rotation following by a vertical flip: Fields are flipped horizontally.

– 90° rotation following by a vertical flip: Field image will be rotated 90 degrees clockwise and flipped vertically before recognition.

– 180° rotation following by a vertical flip: Field image will be rotated 180 degrees clockwise and flipped vertically before recognition.

– 270° rotation following by a vertical flip: Field image will be rotated 270 degrees clockwise and flipped vertically before recognition.

– Default – No rotation and no flipping.

> Despeckle: Specifies whether an isolated small chunks of black pixels should be deleted from the form image. This procedure can reduce the size of compressed form image by up to 30%, thus improve system throughput and reduce disk storage requirements. The allowed values are Yes (default) or No.

> Margin removal: Specifies whether the margin will be removed from the image. The allowed values are Remove white margin, Remove black margin or Off (default).

> Deskew: Specifies whether the form image should be deskewed before processing. Deskewing form images can significantly improve recognition process, especially for table fields. The allowed values are Yes or No. The default value is Yes. On images with a black border, using this option may produce unpredictable results.

> Normalize X-resolution: Specifies the new horizontal resolution an image should be scaled to.

> Normalize Y-resolution: Specifies the new vertical resolution an image should be scaled to.

Classification

> Classifier type:

Document Separation Properties

> Mode: A mode to use for document separation. There are several configuration modes from which to choose:

– Use registration: This is a default mode that uses registration results to detect document boundaries. All versions of the software prior to this one behaved in this fashion.

– Each multi-page tiff is a document: Each multi-page tiff file is considered as a separate document. Pages in such documents are always kept together and are not split regardless of registration results.

> Enable batch level separation: Indicates whether a batch level document separation is enabled, if Yes selected.

> Number of documents in batch: Defines minimum and maximum number of documents in batch. To set these values click the small angle item in right bottom corner of the field.

> Documents are separate: Defines whether documents are separate from each other or pages of one document can appear between pages of another document.

Registration Properties

> Detect upside-down pages: Specifies whether a registration process should detect upside-down images. If this property is set to Yes, all images will be pre-rotated properly before the registration, if necessary. If the image is not registered, it tries to rotate and register it again. If successful, it assumes that the image was rotated.

Note: An image may be set in the Guess status. It means that we could not register the image to any template, but we were able to calculate the document and page number using ambient information. For instance, in two-paged documents with sequential order we assume that the page that follows the first page is the second page even if it was not registered.

> Detect pages rotated 90° and 270°: Specifies whether a registration process should detect images rotated clockwise and counterclockwise. If this property is set to Yes, all images will be verified for being rotated  90° or 270° and will be pre-rotated to be orientated vertically before registration, if necessary.

> Maximum horizontal shift: Establishes the maximum horizontal shift allowed for form registration. Forms that are shifted more than the allowed value will not be registered. The higher the shift is set, the higher the number of forms that will be registered, but the less accurate the registration. The smaller the allowed shift, the more accurate the registration, but the fewer the forms that are registered. Choose the shift value so that it corresponds to a typical average shift for accurately scanned forms. The default value, which corresponds to the Not Set option, equals 200 pixels. The value must be within 0 and 500 range.

> Maximum vertical shift: Establishes the maximum vertical shift allowed for form registration. Forms that are shifted more than the allowed value will not be registered. The higher the shift is set, the higher the number of forms that will be registered, but the less accurate the registration. The smaller the allowed shift, the more accurate the registration, but the fewer the forms that are registered. Choose the shift value so that it corresponds to a typical average shift for accurately scanned forms. The default value, which corresponds to the Not Set option, equals 200 pixels. The value must be within 0 and 500 range.

> Use image size: Specifies whether an image size should be used in the registration procedure. When image size differs from the size of sample image by more than 10% in any direction, the image will not be registered. If set to Yes, image size will be taken into account during registration.

> Use image resolution: Specifies whether an image resolution should be used in the registration procedure. When image resolution does not match the resolution of sample image, the image will not be registered. If set to Yes, image resolution will be taken in account during registration.

Important Note: The captured images/documents may fail during Registration and/or Extraction if the resolution (DPI) and Height/Width do not match the settings in the Image Properties.

Barcode Recognizer

> Barcode recognizer:

> Barcode type:

OCR Properties

> OCR Module: Defines the OCR module that will be used. The allowed values are FineReader, CuneiForm or None (default).

> Recognize whole page: If set to True the entire page is recognized by page OCR. The default value is False.

> OCR Language: Defines one of the following languages – English, German, French, Russian, Portuguese (Brazil), or Spanish (Latin America). To use these languages you should have corresponding license.

Page Registration Properties

To preview and specify registration properties for a certain page of a form in the form definition, click on the page name in the Form Definition Layout window. Registration properties will be displayed in the right window on the screen. In Figure 4-5 you can see the registration properties set for Page in the form definition my_definition:

You may specify the following registration properties for a page in your form definition:

General

> Name: Specifies the page name. The property length must be at least 1 character. The property length must not exceed 32 characters.

> Description: Specifies the page description. This property is for your reference only. The property length must not exceed 255 characters.

> Page type: Specifies the page description. This property is for your reference only.

> Use OCR in dynamic field location:

Registration

> Registration mode: Specifies the image registration mode. The following options are available:

– Accept page without registration: If selected allows accepting a page without registration.

– Use registration zones: For registration will be used registration zones.

– Use zones with rescale: The registration mode that takes into account changes in image scale.

– HCFA/CMS-1500: Defines specific HCFA 1500 forms for medical applications, which do not require any zones to be defined.

– UB-04/CMS-1450: Defines specific HCFA 1450 forms for medical applications.

– Check (financial document): Defines a check that is supported by CheckPlus or CheckUltra applications.

> Page must have a barcode: Indicates if page has a barcode. Use this information for document separation. Default is No.

> Page must be an envelope: Indicates if page has an envelope. Use this information for document separation. Default is No.

Document Separation

> Number pages in document: Defines minimum and maximum number of instances of this page in document. To set these values click the small angle item in right bottom corner of the field.

> Starts document in batch: Indicates whether an occurrence of page is a flag to start a new document.

> Ends document in batch: Indicates whether an occurrence of page is a flag to finish document with this page.

Verification

> Keying type: Defines the type of keying that applies to all fields on the page.

> Double keying: A number of operators that have to process fields on this page. The value may be Single, Double or Triple.

> Keyer tip (page keying): Defines a help string to display to the operator during page keying.

Using Recognition for Registration

To speed up the process of creating form definition you can use recognition while registering test images.

To enable recognition during the registration:

Step 1: Specify the registration zone on the form.

Step 2: Run form registration as described below in the section Running the Form Registration. Note that unless you have completed form registration, you will not be able to proceed with the next step of specifying and recognizing fields.

Step 3: Specify a field (known as the registration field). To do this click the New Registration Field button. The Generic Fields context window appears.

Step 4: Select a field type and then define the field position on the image using your  mouse.

Note: Specifying the registration field is similar to specifying other fields on the form.

To specify the registration field:

Step 1: Select Registration on the Navigation pane to return to the Registration stage.

Step 2: Rename the sample image in the Form Definition Layout window so that its name is identical to the text written in the registration field.

Step 3: Run the form registration – see the section Running the Form Registration below.

Running Form Registration

After setting the registration zone(s), you can run the process of registration.

To run the process of registration:

Step 1a: To process multiple images from the deck, click the Register button on the Tool Bar. All images in the deck are processed.

Step 1b: To process a single image from the deck, select that image by highlighting it and then click the Validate button on the Tool Bar.

Step 1c: To validate the registration parameters using selected images in the current test deck click Validate.

Checking the Results

To check the process of registration:

Step 1: Make sure that most or all images were registered successfully. A registered image is identified by a green check mark to the left of its name. Images that did not register have a red cross mark to the left of their names.

The Registered column displays the results of form registration for every test image (see Figure 4-7).

Step 2: If you are not satisfied with the results, go to the next section of the document to modify the registration zone(s), create an additional zone, or create a new zone.

Modifying the Registration Zone

If for some reason registration failed for all or most of the images, it is recommended that you modify the registration zone.

To modify the registration zone:

Step 1: To modify the existing registration zone, select the registration zone on the sample image and change its location or size and shape.

Step 2: To add another registration zone, follow the instructions in the section Setting Up the Registration Zone.

Step 3: To delete the existing registration zone: right-click on the registration zone click the Delete icon on the Tool Bar.

Step 4: To create a new zone, follow the instructions in the section Setting Up the Registration Zone.

Step 5: After modifying or adding a registration zone, process the images again. To do this, follow the instructions in Running the Form Registration.

Step 6: Once you have created registration zones and processed the test images successfully, you can proceed with the next steps of designing the form definition.

Drop-out Template

Usually the HCFA form is printed in appropriate color. In this form, a user as a rule prints own data in the black font. When scanning the user information is only recognized and all the lines and separators are removed keeping in mind their color. It allows recognizing and keying information more suitable. However, when archiving it has a sense to keep forms with all separators for more comfortable viewing.

The purpose of this functionality is to archive batch image as it was before scanning, including image parts that were dropped out by the scanner. A separate Drop-out Template image may be added to every Sample Image (clone). Scaling and alignment of the Drop-out Template with respect to the Sample Image may be adjusted to achieve a maximum possible match between Sample Image and Drop-out Template. ILINX Advance Capture Workflow Engine uses batch image registration results (document, page, and clone), Template image, its scaling and alignment properties to superimpose registered batch image and the Drop-out Template. This superimposition occurs only when batch images are archived.

Adding Drop-out Template to a Sample Image

Step 1: Being in the Form Definition Layout window right-click a sample image.

Step 2: Select Add Drop-out Template from the context menu. The Open Images window displays from which choose the corresponding image as your template. Drop-out Template appears underneath a sample image.

Drop-Out Template Properties

If you select a Drop-out Template item in the Form Definition Layout window, the window Properties of Drop- Out Template appears as pictured below.

Properties of Drop-out Template have the following meaning:

General:

> Name: A name of Drop-out Template. You can change this name. It must not exceed 32 characters. By default – Template_n.

> Description: Any text which must not exceed 255 characters.

> Drop out: Defines whether a drop-out background contained in the template’s sample image will be removed during scanning.

Notes:

1. Use values less then 1 to squeeze and values greater then 1 to expand the template. Possible value range for these properties is from 0.2 to 5.

2. ILINX Advanced Capture Studio functionality such as Copy/Paste, Drag-and-Drop, Undo/Redo, fly-over hints is supported with Drop-out Templates where possible as appropriate.

Registration Window Tools

Appearance of Tool Bar depends on the window selected on Navigation pane.

Registration tab

For the Registration tab of Registration window it looks as follows.

The icons of Tool Bar have the following meaning:

Image tab

For the Image tab of Registration window it looks as follows.

The icons of Tool Bar have the following meaning:

---

Fields

The next stage in designing the form definition is specifying fields. Select the Fields item in the Form Definition Development Cycle window menu list to proceed to this stage.

The Fields screen is designed to let you easily specify a new field, define its location on the form, and set field properties and preprocessing options. The workspace is divided into following panes:

> Tool Bar pane: Appearance of Tool Bar depends on the window selected on Navigation window. For information about Tool Bar for Fields Layout and Image tabs of the Fields window see Fields Window Tools.

> Fields Layout: Displays the hierarchical representation of folders, tables, and fields defined on a document page of a form definition;

> Properties: Reserved to specify properties of the fields defined on the document page.

> Image Preview: Displays the sample image and allows you to specify the location of fields on it.

Figure 5-1 shows the initial Specify Fields screen view. Any fields already defined are listed in the tree structure within the Field Layout window.

Specifying Fields Window allows you to do the following:

> Creating a New Field

> Creating a New Table

> Viewing Field Types Description

While working within the Field window, you can also do the following with images by clicking the Image tab toolbar:

> Center an image in the image window by clicking the button Center Image.

> Display the rulers around an image by clicking the button Rulers. If clicked once more the rulers disappear.

> Increase the image size displayed on the screen by selecting Zoom in.

> Decrease the image size displayed on the screen by selecting Zoom out.

> Adjust the zoom ratio so that the image becomes fully visible by selecting Fit to Window.

> Adjust the zoom ratio so that the entire window width is used by selecting Fit to Width.

> Adjust the zoom ratio so that the entire window height is used by selecting Fit to Height.

Creating a New Field

You are allowed to create new fields of more than twenty types. Refer to the section titled Appendix A: Field Types for a detailed description of the field types provided in the ILINX Advanced Capture Studio and for tips on how to choose the correct field types.

Important Note: The field name identified in the General section of the Properties must match the Index field name configured in the Doc Type of the Batch Profile.

To create a new field:

Step 1: Click the field button on the Tool Bar.

Step 2: Select the desired field type from the context list of field ILINX Advanced Capture Studio provides over 20 commonly encountered field-types, so that you can easily select a proper field type for any field that is encountered on your form. When creating the Address fields, you also have to select the proper format of address field that matches the format of the Address field on your form.

Step 3: Using the image displayed in the Image Preview pane, put the crosshairs in the upper left comer of the desired field. Hold down the left mouse button, drag a marquee around the field area, and release the mouse button. A field zone, highlighted in light green, is created on the image. For complex address field types, several field components may be created.

Step 4: In the field name box of the Field Layout window, type a name for the field.

Notes:

1. The system automatically creates for a check the following fields: Check type, Check amount, Check date, Check number, and Check MICR line.

2. The system automatically creates for a signature the following properties: Signature Style, Unify Resolution, Signatures Expected, Genuine Threshold, and Forgery Threshold.

3. If a check has other format you can change bounders of these fields.   

Creating a New Table

You can create a table structure for forms containing multiple rows and columns of data. Using table structures for homogeneous data organized as tables on the forms is more efficient and easier than defining each table cell as a separate field.

To create a new table:

Step 1: Click the Table button on the Tool Bar. The Insert Table window opens on the screen.

Step 2: Put the mouse pointer in the upper left comer of the table and drag it across this window. While dragging the current row and column numbers will be displayed in the upper left corner of the Insert Table window.

Step 3: Release the mouse button. Using the image displayed in the Image Preview window put the crosshairs in the upper left comer of the future table. Hold down the left mouse button, drag a marquee around the table area, and release the mouse button. The fields you defined, highlighted in yellow, will be shown in the corresponding cells of the table.

Step 4: Using the mouse change the size of table cells to correspond to the size of fields on your form.  New field Table appears in the Field Layout window.

Note: You do not have to define all cells in the table as new fields. The field definition created for the first cell in each column or row will be automatically applied to all cells in the corresponding column or row.

Setting the Field Properties

When you select a field in Fields window, its properties are displayed in the Properties pane. You can fill out the fields directly in the pane. To do this, double click the corresponding field.

To define field properties:

Step 1: Select the field you want to specify properties for in the Fields window.

Step 2: Specify the properties for this field in the Properties pane.

You can use the following field properties:

> General Properties

> Location and OCR Properties

> Image Cleanup and Preprocessing Properties

> Recognition Properties

> Validation Properties

General Properties – You may specify the following general properties:

> Name: Specifies the field name. The property length must be at least 1 character. The property length must not exceed 32 characters.

> Description: Specifies the field description. This property is for your reference only. The property length must not exceed 255 characters.

> Enabled: Specifies whether the field should be processed (Yes) or not (No).

> Keying only: Specifies whether the field is not recognized but keyed only.

Location Properties – You may specify the following location properties:

> Position – Defines the type of field position. May be one of the following: Fixed, Dynamic, Recipient Address, Sender Address, Boxed, Payor Block.

If in Location properties you select Dynamic from the Position property, the system automatically expands Location properties folder. The following additional properties appear:

> Location Type – Specifies the type of dynamic field location:

– Keyword only: Specified a field that doesn’t have value.

– Keyword and value: Specifies a field that uses Keyword to determine the location of the Value.

– Value only: Specifies Value which could be found without a Keyword. The context of the Value (Field Type, Template or Regular Expression) should allow to distinguish the Value from all other text elements in the Location Area.

– Anchor: Similar to Keyword Only. Specifies a field that must be present on the page, doesn’t have value and is used to help to define locations of other fields.

– Optional Keyword: Specified a Value which may have a Keyword. The engine will first try to locate the Value as Keyword and Value and if it fails then as Value Only.

> Location Area – A position of the field on the page. Specifies one of the following regions: Whole page, Right half, Left half, Top half, Bottom half, Top-left quadrant, Top-left quadrant, Bottom-left, Bottom-right, Inside field zone.

> Answer priority – Specifies which answer to select as the best from competing candidates. Specifies one of the following priorities: Highest confidence, Rightmost, Leftmost, Topmost, Bottommost, Longest. The Longest value is used if the dynamic field contains more than one word or defines the longest answer.

> Required – Specifies whether the field and its value must be present on the form.

If Location Type is not Value Only, the following properties are available in Location properties:

> Keyword vocabulary – A vocabulary that contains field keywords. Here you can create a new Vocabulary or DB Vocabulary, or select one of the existing vocabularies which can be edited. Using vocabularies allow you to increase the recognition quality. For further information refer to the section Using the Vocabulary Editor.

> Keyword reject vocabulary – A vocabulary that contains words that cannot be field keywords.

> Keyword language – It defines the language for dynamic field’s keywords. All dynamic fields on the page should have the same language for keywords. You can specify one of the following languages: English, German, French, Russian, Portuguese (Brazil), and Spanish (Latin America). To use these languages, you should have the corresponding license.

> Keyword writing style – Defines one of the following writing styles for the keyword – Machine-print, Hand-print.

If Location Type is set to Keyword and Value, Value Only, or Optional Keyword, the following additional properties are available in Location properties:

> Value position – A position of the field value relative to its keyword. May be one of the following: N (top position from keyword), E (right position from keyword), S (bottom position from keyword), W (left position from keyword) SW (bottom-left position from keyword), SE (bottom-right position from keyword), NW (top-left position from keyword), NE (top-right position from keyword) and KW (in the same cell with keyword). You can select several values.

> Value Vertical Proximity – Expected distance between field value and its keyword in vertical direction. Minimum and Maximum distances are measured by numbers of keyword rectangle height. If distance between closest edges of located keyword and value will be outside of specified range then the dynamic field answer will have low confidence. To define Minimum and Maximum distance click this field and set corresponding values. You can leave this column as empty one.

> Value Horizontal Proximity – Expected distance between field value and its keyword in horizontal direction. Minimum and maximum distances are measured by numbers of keyword rectangle height. If distance between closest edges of located keyword and value will be outside of specified range then the dynamic field answer will have low confidence. To define Minimum and Maximum distance click this field and set corresponding values. You can leave this column as empty one.

Note: You can specify several rules per each part of fields. Rules are combined using disjunction (OR) rule.

> Field relations – This field defines relationship rules between this and other fields on the page. Each rule contains the name of the related field, type of reference element (Keyword or Value) for this and related fields, the distance between reference elements, and their relative position.

Click Field relations to display the table in the top part of a window. The table with the relation rules has the following columns:

> Field Element – Defines type of reference element (Keyword or Value) of this field. To define type of reference element click this column and select Keyword or Value from drop-down list.

> Related Field – Defines related field. To define this field, click this column and select corresponding field from drop-down list.

> Related Field Element – Defines type of reference element (Keyword or Value) of related field. To define type of reference element click this column and select Keyword or Value from drop-down list.

> Horizontal Distance – Defines Minimum and Maximum distance between reference elements of this and neighbor fields in horizontal direction. The distance is measured by numbers of keyword rectangle width. If distance will be outside of specified range, this element will not be available. To define Minimum and Maximum distance click this column and insert corresponding values. You can leave this column as empty one.

> Vertical Distance – Defines Minimum and Maximum distance between reference elements of this and neighbor fields in vertical direction. The distance is measured by numbers of keyword rectangle height. If distance will be outside of specified range, this element will not be available. To define Minimum and Maximum distance click this column and insert corresponding values. You can leave this column as empty one.

> Position – A position of the Related Field Element to Field Element. May be one of the following: N (top position from Field element), NW (top-left position from Field element), NE (top-right position from Field element), E (right position from Field element), S (bottom position from Field element), W (left position from Field element), SW (bottom-left position from Field element), SE (bottom-right position from Field element).

Important! Some rules you are recommended to apply practically while working with dynamic fields:

1. You can use dynamic fields if it essential for your work. Keep in mind that dynamic fields can significantly slow down a recognition process.

2. If necessary to use a dynamic field, associate it with a fixed field or keyword though the latter case is a better choice.

3. You are recommended to point location a field relatively keyword as precisely as possible using such parameters as Value position and Proximities.

4. A keyword should consist of 2 – 3 words. Avoid using too short or too long keywords.

5. If all the values for the field are known, i.e. if there is appropriate vocabulary, and the keywords cannot be defined, select one of the following ways:

> Recognize without keywords. Vocabulary will be used on recognition stage only. For example, you can select Vocabulary (1) from Recognition properties.

> Use a Vocabulary for keywords. In this case set property Keyword only in Yes value.

Notes:

1. If vocabulary item consists from several words, you are recommended to select the second way from point 5.

2. If you will stick to proposed rules the results of recognition will be more reliable and predictable. There are in view a distance between reference elements, and their relative position at the first.

When you insert changes in parameters mentioned above Location Properties will be looked as follows:

Image Cleanup and Preprocessing – Specify operations that have to be performed on an image before recognition. These include image rotation, filtering to remove extraneous elements like lines, combs, boxes, or intrusions from field images, and also the aligning of field images (see description below). These properties are available for modification in the Fields and Recognition windows.

> Rotate image – Specifies whether the field must be pre-rotated before recognition and the rotation angle desired. Select from the following options:

– No rotation and no flipping – Fields are recognized as is.

– 90° rotation without flipping – Field image will be rotated 90 degrees clockwise before recognition.

– 180° rotation without flipping – Field image will be rotated 180 degrees before recognition.

– 270° rotation without flipping – Field image will be rotated 270 degrees clockwise or 90 degrees counter-clockwise before recognition.

– No rotation following by a horizontal flip – Fields are flipped horizontally.

– 90° rotation following by a horizontal flip – Field image will be rotated 90 degrees clockwise and flipped horizontally before recognition.

– 180° rotation following by a horizontal flip – Field image will be rotated 180 degrees clockwise and flipped horizontally before recognition.

– 270° rotation following by a horizontal flip – Field image will be rotated 270 degrees clockwise and flipped horizontally before recognition.

– No rotation following by a vertical flip – Fields are flipped horizontally.

– 90° rotation following by a vertical flip – Field image will be rotated 90 degrees clockwise and flipped vertically before recognition.

– 180° rotation following by a vertical flip – Field image will be rotated 180 degrees clockwise and flipped vertically before recognition.

– 270° rotation following by a vertical flip – Field image will be rotated 270 degrees clockwise and flipped vertically before recognition.

– Default – No rotation and no flipping.

> Remove boxes and/or combs – Specifies whether the boxes or combs are deleted from field image. When you select outside option, all the data from outside boxes or combs are deleted also; otherwise only boxes or combs lines are deleted. Select from the following options:

– No – Removal of boxes and combs is turned off.

– Remove box – Removes boxes from fields.

– Remove box and clean around the box – This option removes boxes and other lines, words, or intrusions that occur outside of the boxes.

– Remove combs – Removes combs from around individual characters.

– Default – No.

> Remove lines – Specifies whether the horizontal and/or vertical lines are deleted from field image. Select from the following options:

– No – Line removal is turned off.

– Remove vertical lines – All vertical lines will be removed from the field image.

– Remove horizontal lines – All horizontal lines will be removed from the field image.

– Remove vertical and horizontal lines – Both vertical and horizontal lines will be removed from field image.

– Remove vertical and horizontal dash lines – Both vertical and horizontal dash lines will be removed from field image.

– Default – (No).

> Remove template – Specifies whether the sample image template will be removed from the field image. Turns on (Yes) and off (No) the removal image template. Default – (No).

> Remove pre-printed text – Turns on (Remove from top or Remove from bottom) and off (No) the removal of pre-printed text and other pre-printed information from the field image. You must have a blank sample image to be used as a pattern for image cleaning when removing pre-printed text. This method of cleaning is highly effective for field snippets that include pre-printed form material. If a blank sample image is unavailable, you can use the built-in image editor to clean all written information from the sample image. Default – (No).

> Remove noise – Specifies whether the small isolated noise and intrusion lines are deleted from field image. Turns on (Yes) and off (No) the removal of small noise objects from the field image. Noise objects are stray lines or curves that extend into the snippet area from above, below or the sides. Default – (No).

> Align text – Specifies whether the field text is horizontally aligned before recognition. Turns on (Yes) and off (No) the horizontal alignment of text in a field image. If the mode is on, places the image on a straight, horizontal plane. Corrects writing that slants upward or downward. Default – No.

Note: Setting Horizontally align field text option to Yes may lead to misinterpretation for some barcode types if those barcodes are rotated to some specific angles.

Recognition Properties – Specify for the recognizer what information to expect in the field (style of writing, date format, value limit, etc.) and how to recognize the field (with or without vocabulary, whether to use an alias table, database, etc.). The set of properties available for a field depends on the field type. Only properties available for a particular field type are displayed in the Properties of window for the corresponding field. Field properties are also available for modification in Fields and Recognition windows. The list of Recognition Properties is provided below.

> Field type – The field type is a primary description of field content. A field type specifies the format of the field, the set of characters that may occur in the field, and therefore the way it will be recognized. Thus, it is important that you choose an appropriate field type for each field that is encountered on your form. Select from these values:

– Alpha – The field contains a generic alpha text.

– Numeric – The field contains a generic numeric text.

– Alpha-numeric – The field contains a generic alpha-numeric text. Either vocabulary or template should be specified for this field to get recognized.

– Address – The field contains a USA address in various formats.

– Age – The field contains a person’s age.

– Amount – The field contains a dollar amount (including cents).

– Checkbox – The field contains a checkbox.

– Credit card number – The field contains a number of credit card.

– Date – The field contains a date in various formats.

– Empty – The field is recognized as blank or non-blank.

– First Name – The field contains a person’s first name.

– Full Name – The field contains a person’s full name.

– Last Name – The field contains a person’s last name.

– Middle Name – The field contains a person’s middle name initial.

– Phone number – The field contains a USA phone number.

– Social Security Number – The field contains a USA social security number.

– Length – The field contains a length measurement expressed in feet, inches, or both.

– Measure – The field contains a numeric value and an abbreviation (such as “115 “). The measurement unit (lbs., oz., kg., mm., and so on) should not exceed four characters.

Notes:

> Russian language is supported for the following field types: Alpha, Alpha-numeric, First Name, Last Name, Middle Initial, and Numeric.

> French language is supported for the following field types: Alpha, Alpha-numeric, and Numeric for Cursive writing style.

> Language – The following languages are supported: English, German, French, Russian, Portuguese (Brazil), and Spanish (Latin America). To use these languages you should have corresponding license.

When working with the Date or Numeric field types you are allowed to select one of the available languages – English, German, Portuguese (Brazil), or Spanish (Latin America). The last two languages are also supported by Alpha and Credit card number fields.

Because recognition of German numbers is only realized for hand-written text, an error of recognition arises, if you simultaneously select German for the Language property and Machine-Print for the Writing Style property: The current combination of language and recognition settings is not supported.

The same error arises, if you simultaneously select German for the Date field type and define one of the following date formats: DD MMMM YYYY or MMMM DD, YY.

Default – English.

> Writing style – Specifies the field text writing style. Select from these values:

– Cursive – The text is handwritten; the letters in the words are generally linked together.

– Hand-print – The text is hand-printed without any constraints.

– Boxed hand-print – The text is hand-printed within the confines of special boxes or combs.

– Machine-print – The text is machine-printed.

– Auto-detect – The style of writing is unknown. The application will automatically detect the style of writing by selecting between machine-printed and unconstrained hand-printed styles.

Notes:

> Machine-print and OCR-A writing styles are not supported for Russian language recognition.

> French language is supported for the following field types: Alpha, Alpha-numeric, and Numeric for Cursive writing style.

– Default – Hand-print.

> Speed/Accuracy – Selects whether the speed or accuracy is a priority during recognition. Select from these values:

– Accuracy – The recognition is more accurate.

– Speed – The recognition takes less time.

With the speed setting, recognition is faster but might not be as accurate, especially for fields containing names written with cursive or unconstrained handprint. With the accuracy setting, the recognition process includes an additional step, which makes it slower but more accurate. Consider using the accuracy mode for first name, last name, and full name field types and the speed mode for other field types.

– Default – Accuracy.

> Vocabulary (1) – Specifies the recognition vocabulary. Using a vocabulary – even when it’s optional – can significantly improve recognition accuracy. Refer to the section Setting Vocabulary to learn how to select an existing vocabulary or create a new one.

Default – Not Set.

> Vocabulary (n) – Specifies additional vocabulary files to be concatenated with the main vocabulary. This mechanism simplifies the use of large vocabulary files. It eliminates the need to physically create a large vocabulary. Instead it allows you to work with smaller vocabulary files, which contain the components of the words from the desired vocabulary.

Additional vocabularies are used to recognize fields, which could contain more than one word. If, for example, you have a field “City, Country”, you could set vocabulary (1) to the list of Cities, and vocabulary (2) to the list of Countries.

Select one of the vocabulary files defined in the object table. You can specify up to 4 additional vocabulary files. Refer to the section Setting Vocabulary below to learn how to select an existing vocabulary or create a new one.

Default – Not Set.

> Alias table – Specifies the recognition alias table. After recognition, all recognition answers are converted from aliases to their standard form. Select one of the alias tables defined in the object table. Refer to the section Setting Alias Table below to learn how to select an existing alias table or create a new one.

Default – Not Set.

> Special charset – Specifies additional characters to be added to the field default character set (alpha, numeric, or alpha-numeric). Type here any characters you want to recognize within this field (comma, period, etc). For example, add a hyphen to the special character set when recognizing account numbers, which are otherwise a numeric field type.

Default – Not Set.

> Template – ILINX Advance Capture allows you to define field format through a template. A template is a string of characters and symbols that represents both an input and (partially) an output string. Templates may include any of the following special characters:

# matches numerals

@ matches alpha characters

$ matches special symbols

* matches any symbol

The set of symbols depends on field type. For example, if you are recognizing insurance application numbers that always consist of three alpha characters followed by four numeric characters, you would specify a template of @@@####. For more information about field types refer to the section Appendix A: Field Types.

Default – Not Set.

> Exclude charset – Specifies additional characters which cannot be used in the current field. This character set is enabled for Alpha-numeric fields only. For example, it can contain the letter “O” to avoid confusion with digit “0”.

> Valid length – Specifies the valid range of lengths of recognition answers. There are several formats to provide the answer length range (N, N1, N2 are the numerals):

N: The answer should be exactly N characters long,

N1-N2: The answer length should be between N1 and N2 characters inclusively,

N-: The answer length should be N or more characters long,

-N: The answer length should be N or less.

Notes:

> If a fixed length is set and the recognizer finds that the length of the current field is less than this value, the result is considered confidently empty.

> Property ‘Valid length’ affects the acceptance of recognition results.

> Valid values: Specifies the valid range of values of numeral recognition answers. There are several formats to provide the answer value range (N, N1, N2 are the numerals):

N: The answer value should be equal to N,

N1-N2: The answer value should be between N1 and N2 inclusively,

N-: The answer value should be N or more,

-N: The answer value should be N or less.

> Additional vocabulary: This property for ‘Full Name’ field type supports non-name entries like SAME, NONE, etc. Designed to replace ‘Name or Alpha’ field type.

> Date format: This parameter is only applicable for the Date field type. Allows you to select one of the available formats: MM/DD/YYYY, MMMM DD YYYY, DD/MM/YYYY and so on.

> Valid dates: This parameter is only applicable for the Date field type. Defines a valid range for the date.

> Two digit year cutoff: This parameter is only applicable for the Date field type. Defines a cutoff year for interpreting two-digit years as four-digit years.

> Blank delimited: This parameter is only applicable for the Date field type. Specifies whether the field components are separated by blanks.

> Multi-line zone: Can be set for fields that contain a few lines of machine-printed text, for example when the name and address are typed on separate lines. Specifies what line (or lines) within the field must be recognized. It works best for machine-printed text, but could also be used for any multi-line field. Select from the following possible options:

– Top: Only the top line of the multi-line field will be recognized.

– Second: Only the second line of the multi-line field will be recognized.

– Third: Only the third line of the multi-line field will be recognized.

– Fourth: Only the fourth line of the multi-line field will be recognized.

– Bottom: Only the bottom line of the multi-line field will be recognized.

– SecondFromBottom: Only the second line from the bottom line of the multi-line field will be recognized.

– ThirdFromBottom: Only the third line from the bottom line of the multi-line field will be recognized.

– FourthFromBottom: Only the fourth line from the bottom line of the multi-line field will be recognized.

– AllBeneathTop: All the lines located below the top line in the multi-line field will be recognized.

– AllAboveBottom: All the lines located above the bottom line in the multi-line field will be recognized.

– DefaultAllAboveBottom: Default option.

Note: The options AllBeneathTop, and AllAboveBottom make sense for address fields only.

> Threshold: Specifies the recognition threshold. Recognition answers whose confidence levels are equal or higher than threshold are accepted and rejected otherwise. Refer to the section Setting Acceptance Thresholds to read how to set the recognition threshold for the field. The value must be within 0 and 101 range. Default – 101.

Default – Not Set.

OCR Properties – Specify how third party OCR results should be merged with the recognition answer.

> Merge mode: Specifies the merge mode. May be one of the following:

– Ignore: For recognition of field is only used internal ILINX Advance Capture engine. This is default value.

– Replace: For recognition of field is only used page OCR.

– Merge: For recognition of field is used internal ILINX Advance Capture engine as well as page OCR and then to merge results.

Validation Properties – Specify parameters of validation for the field recognition results. The parameters apply to recognition answers in the order they appear in the Validation section in the Properties window. Validation properties are available for modification in Fields and Recognition windows. The list of Validation Properties is provided below.

> Remove charset: Specifies characters to be removed from the field answer after recognition.

Default – Not Set.

> Alias table: Specifies an alias table to be used to convert recognition results. Select one of the alias tables defined in the object table, edit it or create a new one. Refer to the section Using Alias Editor to learn how to select an existing alias table or create a new one.

Default – Not Set.

> Allow blank fields: Specifies whether the blank recognition answers are allowed or not. If set to No and the field was recognized as empty, the field will be rejected.

Default – Yes. A field image is enlarged.

> Valid values: Specifies a vocabulary that contains valid field values. If a recognition answer is not in this vocabulary, the field will be rejected. Select one of the vocabulary files defined in the object table, edit it or create a new one. Refer to the section Using Vocabulary to learn how to select an existing vocabulary or create a new one.

Default – Not Set.

> Invalid values: Specifies a vocabulary that contains invalid field values. If a recognition answer is in this vocabulary the field will be rejected. Select one of the vocabulary files defined in the object table, edit it or create a new one. Refer to the section Using Vocabulary to learn how to select an existing vocabulary or create a new one.

Default – Not Set.

Verification Properties – Specify parameters which tune process of keying.

> Document field: Specifies whether the field data may be entered from any page in the document.

> Keying type: Defines a type of keying that applies to the field.

> Double keying mode: A number of operators that have to process fields on this page. The value may be Single, Double or Triple.

> Enlarge field image: Specifies whether the field visible area should be extended by 50% in each direction for the field keying.

> Remove white margin: Specifies whether the white margin should be removed from the field image during field keying.

> Keyer tip (page keying): Defines a help string to display to the keyer during field keying.

> Prompt, if cannot validate: Specifies whether the dialog box prompting invalid field value should be displayed to keyer when his/her answer cannot be validated.

> Template: A field keying template.

> Valid length: A minimum and maximum length of the keying answer.

> Valid values: A minimum and maximum value for the numeric keying answer.

> Vocabulary: A vocabulary to use during keying.

> Threshold: A keying threshold. Shows recognition answer to keyer if recognition answer equals current field answer and confidence level is greater than or equal to the threshold.

Fields Window Tools

Appearance of Tool Bar depends on the window selected on Navigation pane.

Fields Layout tab

For the Fields Layout tab of Fields window it looks as follows.

The icons of Tool Bar have the following meaning:

Image tab

For the Image tab of Fields window Tool Bar looks as follows.

The icons of Tool Bar have the following meaning:

---

Recognition

ILINX Advanced Capture Studio enables recognizing and fine-tuning of fields in your form definition to maximize their potential for recognition in your application. To access the Recognition screen, select Recognition on the Navigation window (see Figure 6-1).

Important! From recognition point of view there is no difference between structured and semi-structured pages. The same recognition technique is applicable to semi-structured pages as they are applicable now to structured pages.

The Recognition window contains of following panes:

> Tool Bar pane: Appearance of Tool Bar depends on the window selected on Navigation window. For information about Tool Bar for Recognition and Image tabs of the Recognition window see Recognition Window Tools.

> Navigation pane: Displays the list of items which define the stages of ILINX Advanced Capture Studio working.

> Fields Layout pane: Displays the current form definition and recognized fields.

> Statistics pane: Displays the field recognition statistics for the processed set of test images.

> Snippets pane: Displays the field snippets cut from the test images as well as recognition results along with confidence values (refer to the section Setting Acceptance Thresholds for more information on the confidence value and its use), truth values for the field (when truth data are available), and the number of the corresponding sample image. This window also contains three different decks, which help you to tune form definition more quickly.

> Properties pane: Displays the list of properties for the field selected in the Field Layout window and allows you to change the field property settings. The current field will be processed every time the settings are changed, so you can observe how these changes affect the field recognition results. The following sections describe the field properties and available values:

– Location Properties

– Image Cleanup and Preprocessing Properties

– Recognition Properties

– OCR Properties

– Validation Properties

Generally speaking ILINX Advanced Capture Studio allows you to use multiple test image decks. Their purpose is to facilitate definition tuning in Recognition window. One of the possible scenarios is to have two decks with test images, one small and one large. Small deck can be used for quick (course) tuning of the definition and large deck for more fine tuning, thus saving overall tuning time.

You can use also Test Bag deck to recognize several images separately. In this case select several images from any deck, right click on selection and choose the command Send to Test Bag.

Any change you make to a field (changing the field snippet size or location, resetting the field properties) results in re-processing the fields and producing new recognition statistics. The following sections describe fine-tuning various field properties in more details.

TIP: For faster processing of the initial tuning of new fields, use a small batch of 100 or fewer images. Once initial settings for the field have been established, fine-tune the form definition with a sample size of 500 or more form images, ideally 1000 or more.

Setting Acceptance Thresholds

The Snippets section of the Recognition window displays snippets from the selected field on all forms in the batch of test images as well as an answer for each field, along with a confidence value that ranges from 0 to 100. A confidence value is a number indicating how confident the recognition process is about a particular answer. The higher the confidence value the higher the probability that the answer is correct.

A certain confidence value can be chosen to serve as a threshold for recognition of a particular field. The higher the threshold, the lower the number of accepted answers. However, increasing the threshold also increases the accuracy of accepted answers. Whatever threshold you choose, an answer with a confidence level less than or equal to the threshold will be rejected. Similarly, an answer with a confidence level greater than the chosen threshold will be accepted. The threshold range is 0 to 101. Default values are set automatically for all fields but may be adjusted as desired in accordance with the need for accuracy for a given field. If the threshold is set to 80, the recognition confidence value for an answer must be 81 or greater for the answer to be accepted. If the threshold is set to 0, all recognition answers will be accepted, while at 101 no recognition answers are accepted.

Set the desired threshold value in the Recognition section of the Properties window. Refer to the section Threshold Tuning below to read how to tune the threshold.

Once you have selected the basic settings for the form processing options, you are ready to add truth values and start tuning the fields for recognition processing.

Threshold Tuning

Threshold tuning is critical for reducing recognition errors. The goal of threshold tuning is to set the Field, Empty Field, and Character thresholds at their optimum levels.

To perform threshold tuning, first run recognition processing on a field. When the processing is complete, click in the Threshold row of the Properties table. Click the button that appears to the right of the Field threshold value in this row. This opens the Recognition Statistic graph for that field.

The recognition statistic chart displays three values for the field:

> The green line shows the percentage of recognized fields that were accepted.

> The red line shows the field error rate.

> The vertical violet line shows the current threshold setting for this field.

Threshold tuning is accomplished by changing the threshold to the optimal value. Change the threshold setting by dragging the pointer on the top of the chart or by modifying the value in the Tune Field screen.

Adding Truth Information

Testing and tuning a form definition requires a truth data file to be prepared for the set of form images used for evaluation. Truth data can be added by loading a text file that contains correct values for each field in the batch of forms. ILINX Advanced Capture Studio offers a convenient Truth Editor for rapid keying and editing of truth data.

Once the field recognition parameters are defined and the truth file is loaded, you can test the recognition results achieved for this field. After obtaining the recognition results, you can tune the parameters of recognition more precisely to achieve the best possible results. ILINX Advanced Capture Studio displays field recognition result received for an individual form and also displays field recognition statistics graph for a batch of forms. This graph shows acceptance rate for the recognized fields, character error rate, and field error rate, as well as the current threshold setting, which can be easily adjusted in accordance with the accuracy requirements for a given field.

To begin evaluating your form definition it is critical to have truth data for all fields in the test set of form images. Truth data can be added by loading a text file that contains the correct field values for each form in the set of test form images or by manually adding data directly in the ILINX Advanced Capture Studio. Recognition truth data are stored in a comma-delimited text file.

Note: The word “Clone” in the title row of the truth file is a synonym of “Sample Image”.

Note: You can switch between a few form definitions, each of which has a specific truth table. A form definition, its test images, and the truth table are integrated. Every time you start working with an existing form definition, the right test images and the right truth table are automatically selected. Therefore, there is no need to select test images and reload the truth table every time you switch the form definition. You always have the correct truth table associated with a particular form definition and you do not have to worry about corrupting it.

Opening a Truth File

If truth information for the test images exists, you can load the truth file. To do this, select Truth window on Navigation Bar then choose Open Truth from drop-down arrow in the left top corner of the screen. Select the text file from the dialog box and click Open. Truth values will now appear under the snippets.

Note: If truth values do not load, the truth file may not be set up properly to match the structure of the fields in ILINX Advanced Capture Studio. The truth file should be a comma-delimited text file in which the first row contains the words “FILE”, “DOCNAME”, “PAGE”, “CLONE” and the exact field names of the fields as they appear in ILINX Advanced Capture Studio. Each subsequent row contains the file name, document name, page number, clone number, and the truth values for one form in each field in the order in which the field names are listed. Any deviation from the correct field names will cause the truth file to fail to load. Use only a .txt file extension for truth files.

Manually Adding Truth Data

If you do not have a truth file to load for the test images, manually add truth values to the field snippets. When no truth has been set, the results cell for truth will have the text “NO TRUTH”.

To add truth manually, proceed as follows:

Step 1a: If the recognition answer (shown below the snippet) is correct:

> Select snippets if their recognition answers are correct. Right-click in these snippets and select Copy Answer(s) to Truth. Truth values appear in the rows under the snippets.

Step 1b: If the recognition answer is incorrect or blank:

> Type the correct truth value for this snippet. Hitting Enter after typing the truth value will automatically advance the cursor to the next snippet’s truth cell.

Saving Truth Files

After providing truth information for your snippets, you can save these values in a truth file by choosing Save Truth from drop-down arrow in the left top corner of the screen. The truth data you have entered will be stored as part of a comma-delimited text file.

Removing Truth Data

You can remove truth data for a specific field image. To do this, right-click in the cell that contains truth data to be deleted and select Remove Truth from the context menu. The cell for truth will now have the text “No Truth”.

Similarly, you can remove truth data for all images. First, select all images in the Snippets. Then proceed as described above.

Sorting Snippets

During the tuning process, it may be helpful to sort the snippets. To change the snippet order click one of the buttons at the top of Recognition window.

The following possibilities are available:

> Sort by Errors: To use this possibility click the Errors button on Tool Bar. An error is an answer, which is incorrect but was accepted. If the confidence of the answer is lower than the threshold setting for this field, this answer is not considered to be an error. This option is only available when truth values are present for the snippets. With this option, snippets with errors are displayed first.

> Sort by Text: To use this possibility click the Text button on Tool Bar. The default setting, which displays the snippets in the order of the files in the test set.

> Sort by Confidence: To use this possibility click the Confidence button on Tool Bar. This option orders the snippets by recognition confidence. Snippets are displayed in descending order of confidence, so that the snippets with the highest confidence level are shown first.

> Sort by Image: To use this possibility click the Image button on Tool Bar. Snippets are grouped in accordance with the number of the corresponding sample image. The number of the sample image is displayed in the right corner of the row that contains the recognition answer and its confidence value.

Viewing Options

Image & Template Preview

For each field snippet displayed in the Snippets window, you can see the form with the selected field highlighted on it. To switch to the Form Image Preview mode, select the snippet, for which you want to see the form page it was cut from, and click the Image button on the Tool Bar. The form page image with the field highlighted in light blue displays in the preview window.

You can also see the image template in the Image Preview window, if click the Template button on the Tool Bar.

Recognition Results Preview

After performing the recognition, ILINX Advance Capture produces two kinds of answers, both in ASCII form: (1) a main recognition answer and (2) a list of probable recognition results.

The main recognition answer includes a confidence level and is shown in the Snippets window.

The list of probable recognition results includes answer versions generated during the recognition, each of which includes a similarity score. The score ranges from 0 to 1500 and indicates to what extent a result represents the input image.

Answer for non-address fields

To see the list of recognition results, click on the snippet for which you want to see the recognition results. These results displays In the Field Answer window. You can select the List, Characters or Picture tab. The List tab (default) presents a table of answers along with similarity scores.

The Characters tab displays recognition results obtained for every character in the main recognition answer. The result for each character includes a confidence value and the coordinate values of the character.

Using the Picture tab, you can observe the segmentation of words into characters and Tracker color you can change.

Note: Character answers do not always coincide with the corresponding characters in the main answer received on a field basis.

Answer for address fields

The recognition answer for address field includes a confidence level and is shown in the Snippets window. Example of this answer presented below.

Here the answer parts have the following meaning:

98 – Confidence

32909763302 – ZIP code:

32909 – Main ZIP5

7633 – Additional ZIP

02 – House number (may be apartment pointer)

PALM BAY – City

FL– State

S – Address type identifier (S – street address, P – PO Box, R – rural route)

14 – Street number, PO Box number, or rural box number

02 – House/building number

HERO ST – Street name, PO Box, or rural route name

SE – Apartment name and number

Notes:

> The address answer has only first three components when only City-State-ZIP line is recognized.

> The apartment name and number component is valid only for street address type answers when apartment number is recognized.

Fine-tuning the Recognition Process

Image Cleanup & Preprocessing Tuning

Start the fine-tuning process by looking at the snippets cut from the test images. The individual snippets are displayed in the Snippets pane. You can return to the Fields window and see the recognized field you are tuning highlighted in red border. Adjust the field size by dragging the field marquee (see figure below).

Next, tune the image cleanup options. A detailed description of all cleanup and preprocessing properties is provided in the table in the section Image Cleanup and Preprocessing properties. Experiment with the Cleanup Filters changing them in the Properties window until as much of the extraneous snippet information as possible is removed. Examine the snippets in the Snippets window. The goal is to have only the data to be recognized displayed in black in the snippets. Avoid removing any portion of the snippet data.

Note: Changes to the snippet size may require changes in the cleanup settings. If tuning or changes to the form definition alter the size and position of the field marquee, repeat snippet cleanup tuning.

Recognition Tuning

When you are satisfied with your snippet cutting and cleanup processes, proceed with the Recognition fine-tuning.

IMPORTANT REMINDER: Recognition tuning must be performed using truth values for the fields being tuned. Tuning without comparison to truth values may actually increase the error rates. Tuning against truth values will allow you to minimize recognition errors while maintaining acceptance of correct recognition answers.

Recognition Statistics are displayed in the Statistics pane in the bottom of the screen (see figure below).

The information displayed in the Statistics pane is described below.

All Fields

> Total: The number of fields processed by the recognizer.

> Accepted: The number of fields accepted with a confidence value greater than the field confidence threshold.

Fields with Truth

> Total: The number of fields that have truth information.

> Accepted: The number of fields accepted that also match the truth.

> Valid: The number of fields processed that also match the truth.

> Error: The number of fields that were accepted (the confidence value is above the threshold), but yielded results that didn’t match the truth.

Not Blank Fields

> Total: The number of fields that contain information (are not blank) and have truth information.

> Accepted: The number of fields accepted that contain information (are not blank) and have truth information.

> Valid: The number of fields processed that also contained information (are not blank) and have truth information.

> Error: The number of fields that were accepted (the confidence value is above the threshold), but yielded blank results.

Evaluating Results

By analyzing the field confidence, the recognition answer, and the truth value for that field, you can evaluate the results for individual fields. Use the following guidelines:

> The recognition answer displayed below each snippet is colored either blue or red. Blue marks the accepted answers, while red marks the rejected answers.

> The confidence value and the answer are blue if the answer has been accepted. Answers are accepted only when the recognition confidence level is at or above the field’s confidence threshold.

> The confidence value and the answer are red if the answer has not been accepted. Note that in this case the field confidence value is below the value set for the field threshold.

> The truth value is blue if the answer matches the truth. The truth value is red if the answer does not match the truth.

IMPORTANT! If you see blue values on the left (confidence value and answer) and red values under the recognition answer (truth value), incorrect answers are being accepted. To prevent incorrect answers from being accepted, use threshold tuning. See the section Setting Acceptance Thresholds.

For evaluating results, sorting the snippets will be extremely helpful. Refer to the section Sorting snippets for information on the sorting options.

If you are not satisfied with your recognition results, continue tuning until you feel that you have maximized the recognition potential for the set of test images. Essentially, the goal is to find the settings for the field whereas many correct answers as possible are returned from the recognition process. Adjusting field properties, field cleanup filters, and the size and placement of the field marquee on the Sample Image will all affect recognition performance: Increases in recognition rate can be seen in an increase in the confidence levels of fields or, most directly, in an increase in the number of correct recognition answers.

During recognition tuning, do not try to correct field or character errors. These errors will be addressed by threshold tuning. See the section Setting Acceptance Thresholds

Recognition Window Tools

Appearance of Tool Bar depends on the window selected on Navigation pane.

Recognition tab

For the Recognition tab of Recognition window it looks as follows.

The icons of Tool Bar have the following meaning:

Image tab

For the Image tab of Recognition window it looks as follow.

The icons of Tool Bar have the following meaning:

---

ILINX Capture Connection

Once you have completed the Registration, Field configuration, and Recognition, you are now ready to connection the Form Definition to ILINX Capture workflow.

Step 1: Within ILINX Advanced Capture Studio, select the blue arrow at the top left of the screen to open the dropdown menu.

Step 2: Select the Save Definition As option. (If you have made changes to an existing file, use the Save Definition option)

Step 3: Enter the file name and save to a accessible location. The file type should default to Form Definition files (*.xpbd).

Step 4: Open the ILINX Capture workflow for the desired batch profile. Add an Advanced Capture IXM from the ILINX Extension Modules list.

Step 5: Click the configuration button under the Configure menu option to open the Advanced Capture Configuration window.

a. Select the Browse button to open the Upload form definition project window.

b. Select the *.xpbd file you saved in Step 3 and click Open.

c. Mapping options:

– Map doc types defined in Capture Studio into ILINX Capture if name matches – To use this option The ILINX Capture Doc Type name must match the Form Definition name to ensure document identification and separation. Also each Index field name you wish to extract from the image must match the field name configured within the Form Definition.

– Direct all doc types in Capture Studio to a single ILINX Capture doc type – Enables 2 dropdowns to be selected from.

» Select the ‘Doc type’ to direct the return values to.

» Select the ‘Index field’ name.

Note: Will also auto map any index fields returned from ILINX Capture Studio if the name matches.  If there is a data type mismatch at runtime, an exception will be thrown and move the batch into the system Errors queue.

– Use alternate separation algorithm.

Step 6: Select OK when you are finished making changes.

Step 7: Save & Publish the workflow.

You now ready to capture an Unclassified Tiff or PDF image to run through the Advanced Capture Studio.

Note: This process must be repeated each time changes are made to the Form Definition.

Troubleshooting Connection Issues

Workflow Example

Depending on business processes, the workflow configurations may vary. When configuring your workflow to use ILINX Advanced Capture, it is important to have the AdvancedCapture IXM right after the Start (if file types other than TIFF or PDF are being captured, it may be necessary to use the Format Converter tool to convert the images.)

In the above example, the FlowDecisionEx IXM is configured to review the confidence level that is returned from the Form Definition. By selecting the built in option Confidence is equal or higher than expected from the Field dropdown, then selecting the Value of Yes, this will review each returned document for the configured level of confidence. The Expected confidence is set when configuring the Doc Type. The default is set to 70%.

Add two HumanActivity IXMs to the Flowchart to return the condition configured within the FlowDecision. The True HumanActivty (labeled as ‘Returns High %’ in the figure above) will have met the condition as expected. The False HumanActivity IXM will contain those batches that are below the configured confidence level.

Once the Workflow is Saved & Published, be sure to refresh IE. Capture a TIFF image using the Unclassified doc option to process through the defined Form Definition. Once the batch has returned, open the batch to review the field level values returned.

ILINX Advanced Capture classification information and operations is added to the document entity. The information that is tracked at the document level is the same as that tracked at the extracted index field level. The ILINX Capture interface has a new document context menu option called “Mark doc as verified” that allows the user to indicate that the document was correctly classified by Advanced Capture. This should clear the low confidence icon.

There are three different icons that will be displayed within the Index panel for the batch:

> Good extraction: [value] – Displayed when the confidence level is higher than configured confidence levels.

> Corrected and verified by [user] on 2024 OR Verified by [user] on 2024 – This icon is displayed after a user has verified the extracted value and may have made a correction to the if necessary.

> Low confidence extracted value: [confidence value] – This icon is displayed when the confidence level is below the configured expected levels and may require user attention.

Selecting Tab or Enter while verifying the extracted data within the index fields.

---

Shared Objects

This section will walk you through the additional features you can use while working with ILINX Advanced Capture Studio. It will focus particularly on the use of the ILINX Advance Capture components presented below.

First of all you have to select Shared Object on the Navigation window. You are allowed to work with the shared objects listed below. To do so click the New button on the left top corner of your screen. You can select the following objects:

> Vocabulary: Using Vocabulary Editor you can open an existing vocabulary or create a new one and assign it to a selected vocabulary object. For more information see the section Using Vocabulary Editor.

> Alias Table: Using Alias Editor, you can open an existing alias or create a new one and assign it to a selected alias table object. For details refer to the section Using Alias Editor.

> Database Alias Table: Using DB Alias Editor you can also set corresponding parameters. To do this see the section Setting Database Alias Table below.

> Database Vocabulary: Using DB Vocabulary Editor you can also set corresponding parameters. To do this see the section Setting Database Vocabulary below.

> String: Allows you to add a string of rules.

> Data Source: When selected allows to change provider and connection string. For details see the section Setting Data Source Info.

Using the Alias Editor

The Alias Editor is a utility that allows you to create and modify alias lists to be used during recognition. An alias list defines words that are spelled differently but have the same meaning. The word pair “Center”/“Centre” is an example of this. Use an alias list to include all possible spellings of a word. Using an alias list can significantly improve recognition accuracy.

The recognition process produces a list of words (probable answers) ordered by similarity to an input image. The top word in the list is the recognition answer. To calculate a confidence value for this answer, the recognizer analyzes the list of probable recognition answers. If at least one word in the answer list is as similar to the input image as the best answer, this word competes with the best answer, and the recognizer cannot be confident about the recognition result.

In this case, the confidence value of the answer will be low. Words having similar spelling appear in the recognition answer list and are the first to compete. However, such words often have the same meaning, and the application does not differentiate which word will be chosen as an answer. An alias list can resolve this situation.

To open an existing alias list:

Step 1: Select one of the Alias Table ID’s.

Step 2: The Shared Object Editor window appears at the right.

To import an existing alias list:

Step 1: Select the Alias Table Click the Paste Special button in the left top corner of your screen. The Open Alias File dialog box appears.

Step 2: You can open an alias list with the .als, .xals or .fxa extension. The Shared Object Editor window will be contained you alias list.

Important: Alias lists with the .als, .xals or .fxa extension have different formats.

Managing an Alias List

You can create an alias list by:

> Adding or deleting words directly in Alias Editor,

> Creating an alias list in an ASCII (text) file and importing it into Alias Editor.

Creating an Alias List Using Alias Editor

Use Alias Editor if you are creating an alias list for the first time and do not have a database or text file containing aliases. Otherwise, consider importing an ASCII file into Alias Editor. Refer to section Creating an Alias List in an ASCII file below for more information.

To add entries to an alias list:

Step 1: Type the word and the alias, changing between the cells with Tab. After you have typed the alias, press Enter or Tab again to move to the next cell.

Step 2: Repeat these steps for each alias you want to add to the alias list. A color highlight marks the line you are currently on.

In such a way you can create multiple entries for words and corresponding aliases. Each word appears in the left column, while the associated alias appears to the right of the word.

To edit a word or alias:

Step 1: Highlight the entry to be edited.

Step 2: Double-click the desired field (alias or word), or press F2 to start editing from the alias field.

Step 3: Modify the alias or word as desired, then press Enter. The modified alias and word appear in the list on the Shared Object Editor window.

To delete a string with word and alias:

Step 1: Highlight the entry to be deleted.

Step 2: Press the Del button on your keyboard.

Setting Database Alias Table

When you select Database Alias Table from drop-down list of Shared Objects, the Shared Object Editor appears on the screen. You have to set the following parameters:

> Data Source: From drop-down list you can select existing data source or create a new one.

> Table: A table name.

> Word column name: A name of the Word column.

> Alias column name: A name of the Alias column.

> Code page: Code page.

To create database alias table:

Step 1: First of all you have to create a database with the name of the Table field. You can use practically any database.

Step 2: Then it’s necessary to create a Data Source for this database (in our example – AliasNames) in the Release window. Consequently the Connection string presented in Figure 9-5. Example of Connection string will be as a result. How to create a new Data Source refer to the section Export Release Results to Database.

Step 3: Using created Data Source you can set the fields Table, Word column name, and Weight column name as shown on picture below.

Using the Vocabulary Editor

The Vocabulary Editor is a utility that allows you to import, create, or modify a vocabulary that will be used during recognition. A vocabulary contains a list of words or numbers likely to occur in a certain field. For example, a vocabulary might contain a list of client names for an insurance company. Using a vocabulary can significantly improve recognition accuracy.

To open an existing vocabulary:

Step 1: Select one of the Vocabulary ID’s.

Step 2: The Shared Object Editor window appears at the right.

To import an existing vocabulary:

Step 1: Select the Vocabulary tab. Click the Paste Special button in the left top corner of your screen. The Open Vocabulary File dialog box appears.

Step 2: You can open a vocabulary with the .voc, .xvoc or .fxv extension. The Shared Object Editor window will be contained your vocabulary.

Important: Vocabularies with the .voc, .xvoc or .fxv extension have different formats.

Managing a Vocabulary

You can create a vocabulary by:

> Adding or deleting words directly in Vocabulary Editor,

> Creating a vocabulary in an ASCII (text) file and importing it into Vocabulary Editor.

Creating a Vocabulary Using Vocabulary Editor

Use Vocabulary Editor if you are creating a vocabulary for the first time and do not have a database or text file containing vocabularies. Otherwise, consider importing an ASCII file into Vocabulary Editor. Refer to section Creating a Vocabulary in an ASCII file below for more information.

To add entries to a vocabulary:

Step 1: Type the word and its weight, changing between the cells with Tab. After you have typed the weight, press Enter or Tab again to move to the next cell.

Step 2: Repeat these steps for each word you want to add to the vocabulary. A color highlight marks the line you are currently on.

In such a way you can create multiple entries for words and corresponding weights. Each word appears in the left column, while the associated weight appears to the right of the word.

To edit a word or weight:

Step 1: Highlight the entry to be edited.

Step 2: Double-click the desired field (weight or word), or press F2 to start editing from the weight field.

Step 3: Modify the weight or word as desired, then press Enter. The modified weight and word appear in the list on the Shared Object Editor window.

To delete a string with word and weight:

Step 1: Highlight the entry to be deleted.

Step 2: Press the Del button on your keyboard.

Setting Database Vocabulary

When you select Database Vocabulary from drop-down list of Shared Objects, the Shared Object Editor appears on the screen. You have to set the following parameters:

> Data Source: From drop-down list you can select existing data source or create a new one.

> Table: A table name.

> Word column name: A name of the Word column.

> Weight column name: A name of the Weight column.

> Code page: Code page.

To create database vocabulary:

Step 1: First of all you have to create a database with the name of the Table field. You can use practically any database.

Step 2: Then it’s necessary to create a Data Source for this database (in our example – Vocabulary Animals) in the Release window. Consequently the Connection string presented in Figure 9-5. Example of Connection string will be as a result. How to create a new Data Source refer to the section Export Release Results to Database.

Step 3: Using created Data Source you can to set the fields Table, Word column name, and Weight column name as shown on picture below.

Setting Data Source Info

When you select Data Source from drop-down list of Shared Objects, the Shared Object Editor appears on the screen.

To set Data Source Info it’s necessary to set corresponding export properties – Provider and Connection string as well as a Data Source name.

Note: You can set data source info for several database types. However, the procedure presented below is relevant for a SQL database.

To set Data Source Info:

Step 1: To set Provider select it from the list displayed by clicking the down arrow at the right of this string. The default provider is System.Data.OLEDb.

Step 2: Click the … button at the right of Connection string. The Data Link Properties dialog, as pictured below, Note that the provider is already highlighted.

Step 3: Select the Connection tab or click Next. The Connection tab is displayed. Select the Use connection string option and click the Build button at the right. The Select Data Source dialog appears.

Step 4: Here you can select an existing DSN name or set a new one by clicking New. In the latter case the first screen of the Create Data Source Wizard, as shown below, appears.

Step 5: Select SQL Server and click Next. Then follow the instructions of the wizard. The window displayed below appears.

Step 6: Click Finish. After this process is completed, the following window appears:

Step 7: To test the Data Source click the Test Data Source button. The system should issue the following message: TESTS COMPLETED SUCCESSFULLY!

Step 8: Then the system returns to the Shared Object Editor, the Connection string will look as follows:

---

License Manager

The ILINX Advanced Capture Engine software is protected from unauthorized use. To use the software you need a valid license installed on your computer. The license is designed to protect and control the distribution of the ILINX Advanced Capture Engine software.

The ILINX Advanced Capture Engine license defines the modules that are available for you to use and the number of documents/fields you can process.
The ILINX Advanced Capture Engine License Manager lets you request a license, insert the license on your computer, view information about the license, and transfer the license to another computer if necessary.

IMPORTANT! If you are using Windows Vista or Windows 7, you must run the ILINX Advanced Capture Engine License Manager with a full administrator access token at least once. Otherwise, the software will not function correctly.

To run an application one time with a full administrator access token:

1. Locate the program icon or a shortcut in Windows Explorer.
2. Right-click the program icon or shortcut, and then click Run as administrator.

Windows 7 does not display Run as administrator option by default. Extended popup menu that has that option appears on Shift + Right-click.

Request a License

The first time you run the License Manager you will not have the license installed. The screen will show that your license is invalid.

If your license is invalid, or has expired, or you have exceeded your license quota and need a new license you will have to request a new license from your vendor.

To request a license you need to generate a machine key and send it to your vendor. Your vendor will issue you the license.

To request a license:

1. Click Request. This will generate a machine key. If you have already requested a license from this computer, but did not install it, the dialog will appear prompting to you to restore the old key.
2. In the dialog that appears, copy the machine key or save it to the file having .slc extension.
3. Send the key by submitting a ticket to ImageSource Support Department. The Support Department will issue you the license. The license will be in a file having .psk extension.

IMPORTANT! You can generate a machine key only if the ILINX Advanced Capture Engine is configured to use local license provider. You cannot request a license while using network or OEM license providers.

Review Parascript.FormXtra.Licensing.dll.config to ensure that the local provider is being used. Also make sure that the connection string in the View group is blank.

Installing the License

Once you have received the license file from your vendor, install it your computer.

You need to install the license on the same computer you made a request from. Do not try to install the license more than once.

To install the license:

1. Click Install.
2. In the open file dialog that appears, select .psk file you have received from the ImageSource Support Department.
3. Click Open.
4. The license information will appear in the License Manager.

IMPORTANT! You can install a new license only if the ILINX Advanced Capture Engine is configured to use local license provider. You cannot install a license while using network or OEM license providers.

Review Parascript.FormXtra.Licensing.dll.config to ensure that the local provider is being used. Also make sure that the connection string in the View group is blank.

Uninstall the License

Sometimes due to hardware or software failures the license becomes corrupted and you need to uninstall the current license to request a new one.

To uninstall the license:

1. Click Uninstall.
2. In the file dialog that appears, click Yes.

IMPORTANT! You cannot restore the license after you uninstall it. The license information will be lost.
IMPORTANT! You can uninstall a license only if the ILINX Advanced Capture Engine is configured to use local license provider. You cannot uninstall a license while using network or OEM license providers.

Review Parascript.FormXtra.Licensing.dll.config to ensure that the local provider is being used. Also make sure that the connection string in the View group is blank.

View License Information

The ILINX Advanced Capture Engine License Manager lets you view information about the license.

The license information is displayed on the application main screen. The help area at the bottom of the screen displays detailed information about selected item.

You can use the ILINX Advanced Capture Engine License Manager to view the license installed on remote computer running ILINX Advanced Capture Engine License Server. To connect to the ILINX Advanced Capture Engine License Server specify the connection string in the View group. The connection string must be in the following format: hostname[:port_number] , where hostname is a name or ip address of the computer you attempt to connect to, and port_number is a number of the port used by the ILINX Advanced Capture Engine License Server.

When the connection string in the View group is blank, the ILINX Advanced Capture Engine License Manager uses configuration settings from Parascript.FormXtra.Licensing.dll.config to connect to the license.

Transfer the License

You can transfer the license from one computer to another. Not all licenses can transferred.

If your license is non-transferrable, or has expired, or you have exceeded the number of transfers you will have to request a new license from your vendor.

To transfer the license to another computer:

1. Start the License Manager on target computer.
2. Click Request. This will generate a machine key. If you have already requested a license from the target computer, but did not install it, the dialog will appear prompting to you to restore the old key.
3. In the dialog that appears, copy the machine key or save it to the file having .slc extension.

4. Start the License Manager on source computer.
5. Click Transfer. In the dialog that appears, paste the machine key or select .slc file you have created.
6. Click Transfer. In the save file dialog that appears, save the license to the file having .psk extension. This will remove the license from the source computer.
7. Transfer “package” consists of up to 3 files: License (.psk), log (.lpd) and financial log (.fpd). Information from all 3 files has to be restored at destination computer. It is normal if one or all log files not generated.

8. Start the License Manager on target computer.
9. Click Install. In the open file dialog that appears, select .psk file you have saved.
Log (.lpd) and financial log (.fpd) will be installed if files are present in the same directory and have the same name as a license (.psk) file.
10. Click Open.
11. The license information will appear in the License Manager.

IMPORTANT! You can transfer a license only if the ILINX Advanced Capture Engine is configured to use local license provider. You cannot transfer a license while using network or OEM license providers.

Review Parascript.FormXtra.Licensing.dll.config to ensure that the local provider is being used. Also make sure that the connection string in the View group is blank.

Write Usage Report

ImageSource may ask you to provide the software usage report from time to time.

To create the report:

1. Click Write Report.
2. In the dialog that appears, select a file having .pdr extension.
3. Click Save.
4. Send the report to ImageSource requestor.

IMPORTANT! You can write a license report only if the ILINX Advanced Capture Engine is configured to use local license provider. You cannot write a license report while using network or OEM license providers.

Review Parascript.FormXtra.Licensing.dll.config to ensure that the local provider is being used. Also make sure that the connection string in the View group is blank.

---

Truth Files

To begin working with Truth Files select Truth on Navigation window. In this section, you will learn about the way to create and modify some portions of your form definition project, such as editing the truth information in Truth window.

Except of Navigation pane this screen consists of the following panes:

> Tool Bar – Appearance of Tool Bar depends on the window selected on Navigation pane. For information about Tool Bar for Truth window see the Truth Window Tools.

To open a truth file:

Step 1: Click the down arrow in the left top corner of your  screen.

Step 2: Select Open Truth from the drop-down menu. The Open Truth window appears.

Step 3: Select the truth file corresponding your form definition and click Open.

To edit individual cell:

Step 1: In the truth file select a cell you are going to edit.

Step 2: Click this cell once more and change cell content.

To save a truth file:

Step 1: Click the down arrow in the left top corner of your screen.

Step 2: Select Save Truth from the drop-down menu. The Save Truth window appears.

Step 3: Select the directory to save your truth file and click Save.

Truth Window Tools

Appearance of Tool Bar depends on the window selected on Navigation pane. For the Truth window it looks as follow.

The icons of Tool Bar have the meaning presented below:

---

Image Libraries

When working with any form definition you can also use so-called Image Libraries. Image Library helps to organize user’s images and drastically accelerates the process of adding test images to form definition. The purpose of Images Library window is to organize your image libraries. You can open an existing library, create a new one or delete it. You can also rename an image library if double click on it.

The Image Libraries Window contains the following panes:

> Tool Bar Pane – Appearance of Tool Bar depends on the window selected on Navigation pane. For information about Tool Bar of the Image Libraries window see Image Libraries Window Tools.

> Image Libraries – Contains Image Libraries.

> Thumbnails – Contains the thumbnails of images from the selected Image Library.

> Image Properties – Displays image properties for selected image thumbnail.

To create a new library:

Step 1: Select Image Libraries item on Navigation pane.

Step 2: Right-click on the Image Library window and select New from the context menu. The Image Library n appears in this window. You can immediately rename it.

To open test images:

Step 1: Right-click on the Image Library pane and select Open from the context menu. The Open Images window displays.

Step 2: Select appropriate images from your disk or other source and click Open.

To save test images:

Step 1: Right-click on the Image Library pane and select Save As from the context menu. The Save Images window displays.

Step 2: Select appropriate images from your disk or other source and click Save.

---

Configuration

To begin working with Configuration window select Configuration on Navigation pane.

In this section you can do the following:

> Copy files from old repository (if corresponding check box is checked). The default location for storing the Repositories is under the logged in user profile. To share the Projects among multiple users: Created a shared location on the server (ie. C:\FormDef\…). Each user must access the Configuration window and set the Repository Folder to the new shared location.

> Set Collaboration server for keeping common form definitions.

> Define the Postal Database path.

> Create a backup copy of the form definition every time it changes (if corresponding check box is checked).

> Display names of the fields (if corresponding check box is checked).

> Define OCR and Classification Cache value.

To define Postal Database path:

Step 1: Click the Browse button on the Configuration window. The Open Postal database window appears.

Step 2: Set appropriate directory and select the last version of database with .cfg extension.

---

Appendix A: Field Types

Each field on the form contains a certain type of data. For instance, a phone number or street address is a type of field. Each field has a certain pattern and certain characteristics. For example, numbers should not be present in an “eye color” field. Thus, the type of a field may determine what kind of data it should contain. This information helps the program read the data in the field correctly.

To provide a high recognition rate, the module called the “recognizer” should also “know” what to expect in the field. This information is given to it through defining the field type. A field type is a primary description of the field content. It specifies the format of the field and the set of characters that may occur in the field. The field type for each field that is encountered on your form is set in the form definition.

After the field type is specified, its name determines what additional information is available to adjust the recognizer for processing a stream of field images. A properly chosen field type simplifies further context description.

Notes:

> Russian language is supported for the following field types: Alpha, Alpha-numeric, First Name, Last Name, Middle Initial, and Numeric.

> French language is supported for the following field types: Alpha, Alpha-numeric, and Numeric for Cursive writing style.

The list of field types includes:

Most fields on an Application for Driver License form can be described with these specific field types (see the Figure below)

In reality, it is impossible to reserve a special type for all variations of field data that may be encountered on the forms. To work with generalized fields, FieldScript has generalized field types. Such field types as Alphanumeric, Alpha, and Numeric are generalized types. On our form we have a few fields that should be described as generalized field types (see the Figure below).

---

Appendix B: Field Types for Invoice Forms

Invoice type of form is characterized by the following:

> It has a set of predefined fields to choose from, and

> No field location on these forms is specified.