Help > Knowledge Processing > Knowledge Processing Explorer > Knowledge Processing Explorer

Knowledge Processing Explorer

Knowledge Processing Explorer is a separate application that provides information about the layout data points (LDPs) that are used for online "learning" of classification and/or interpretation.

This is very useful when Knowledge Processing is implemented in your XBOUND system. For example:

Note: Knowledge Processing Explorer must be installed. Please see XBOUND Installation Guide for details.

Start Knowledge Processing Explorer by selecting Start menu > ReadSoft > Capture Components > Knowledge Processing Explorer.

Use these panes and settings:

Service host

The ReadSoft Knowledge Processing Service host from which data is displayed. After selecting a host, click Refresh above the Active LDPs by document type list.

Solution

The solutions, where Knowledge Processing has learned classification and interpretation, that are present in the database. Select a solution from the drop-down list to work with LDPs and classes found in this solution.

Find LDPs

Use this to search for a specific LDP.

Active LDPs by document class

Document classes are listed by name.

  • Click the colored circle to select a class without expanding the entire list of class LDPs.

  • Click the up/down arrow next to the colored circle to expand/collapse to see ALL LDPs in the class list and to then select a specific LDP. (Up to 100 active LDPs can be displayed for each document class.)

  • If you want to move an LDP to a different class, right-click the LDP and select Change document class. Start typing, and a list of all matching classes is displayed. You can then select the correct class from the list. Refresh the Class neighborhood to view the new changes.

  • Any class that contains one or more LDPs with learned rubberbanded fields is proceeded by a icon. When the class is expanded, these specific LDPs are marked with the icon. ClosedExample:

LDPs are listed by GUID (Globally Unique Identifier).

LDP selected for analysis

A visualization of the LDP that is selected in the list. You can see the words, fields, their positions, and the document class.

If you want to extract a field based on its position, you can rubberband the position on the LDP and add the field. This field then appears under Field details.

Rubberbanded field positons that are inherited from a father LDP are Closedshown with a dotted red outline.

Note: The LDP selected for analysis might also contain fields surrounded by solid lines. Only fields with dotted lines are inherited.

LDP selected for comparison

A visualization of the LDP that is selected in the LDP neighborhood pane. You can see the words and their positions, as well as the document class. Learned, rubberbanded field positons are shown here Closedshown with a solid red outline

Note: The LDP selected for analysis might also contain fields surrounded by solid lines. Only fields with dotted lines are inherited.

This is useful for examining documents of two types (represented by two different colors in the LDP neighborhood pane) to understand why Knowledge Processing considers them similar. Use Zoom in and Zoom out to get a closer look.

When two documents are displayed side by side, words displayed in green are the ones that Knowledge Processing considers to be similar.

Any word selected under Unstructured class match is shown with a red rectangle around it.

Unstructured class match

A detailed breakdown of the words found on the selected LDP showing:

  • all the words found on the LDP.

  • the Weight of the words - how unique they are to this LDP. 1.0 means they are exclusive to the selected LDP. The greater the weight, the higher up the list a word is.

  • whether or not they match the class that the LDP belongs to.

When you click a word, the word is shown with a red rectangle around it in LDP selected for analysis.

LDP neighborhood

An icon with binoculars represents an LDP that is selected for analysis. A visualization of this LDP is displayed in the LDP selected for analysis pane.

If an LDP selected for analysis contains inherited, rubberbanded fields, the father LDP is displayed with a document information icon.

The "neighborhood" consists of icons which represent LDPs that are similar, from the Knowledge Processing perspective, to the selected LDP.

Color is significant:

  • Icons with the same color as the icon with the binoculars represent documents that were recognized as the same document class.

  • Icons with a different color than the icon with the binoculars represent documents that were recognized as a different class.

The distance between the icons corresponds to how similar the LDPs are.

You can click these icons to display the associated document in the LDP selected for comparison pane.

If the Knowledge Processing database contains no other active LDPs that are similar to the selected one, then only the icon with the binoculars is displayed.

Class neighborhood

The "neighborhood" is displayed when you click a colored circle for a class under Active LDPs by document class. When you expand a class (by clicking the arrow next to the colored circle) with a large number of LDPs, a progress bar is shown while the Knowledge Processing Explorer retrieves data from the Knowledge Processing Service. You may, at any time, select or expand a new document class, even if class neighborhood processing is not complete.

The icons, grouped around a center point, show which LDPs are similar to the class in question. Icons that are not the same color/class, and are present in the vicinity, may be a cause for concern.

Color is significant:

  • The center point is the color of the class currently being analyzed.

  • Icons that are the same color as the center point represent documents recognized as the same document class.

  • Icons that are a different color than the center point represent documents recognized as a different class.

  • When the number of different classes is very large, the colors representing them may need to be duplicated. However, Closedat the top of each icon you can see text representing the name of the class. A tool tip with the whole class name is displayed when you hover over each document.

The distance between the icons corresponds to how similar the LDPs are.

You can click these icons to display the associated document in the LDP selected for comparison pane.

Class match

Shows how closely a selected LDP matches (Certain) or does not match (Uncertain) each document class, including a Score of how certain the match is. The higher the score, the more certain the match.

Click the class name to see a detailed breakdown of the selected LDP in the Unstructured class match view.

Field details

A green icon in the left column indicates that this is a Learned value. This is followed by the LDP name, and the value of the field itself, which you can edit here. To delete a field, click the red "X".

If you want to extract a field based on its position, you can rubberband the position on the LDP and add the field. This field then appears here, under Field details. See Adding and Deleting field positions in the Knowledge Processing Explorer for more details.

How and when are layout data points added?

Changing the document class