Decision Knowledge Trees

Decision trees are often related to dichotomy (a division or the process of dividing into two mutually exclusive or contradictory groups). Dichotomous keys are often available to help non-experts diagnose a particular problem. Because of the ease-of-use and structured format of a dichotomous key, K-Vision approaches knowledge acquisition in the same way.

K-Vision uses what could be called a multi-dichotomous key or a decision tree. The difference between how K-Vision and a conventional dichotomous key work is that the Knowledge Engineer (KE) can have more than two paths from a preceding step, and answers are not limited to YES and NO. Certain actions can also be executed, such as displaying a picture or playing a video clip to show the user how to execute an important step in a procedure. K-Vision also allows for "custom" and "numeric" answers.

K-Vision uses decision trees to quickly and easily build Decision Support Expert Systems. K-Vision decision trees are developed with ease on the screen in front of you. Thus, you can readily identify branches that need responses and build the appropriate solutions.

Keeping track of your activity while working within the tree is easy, as the cursor changes design with each type of action. Zoom in and zoom out commands allow you to view all of the tree or only the portion on which you are currently working.

Questions, conclusions, actions, help tasks, or procedures can be added to the end of the answer line to represent the next step in the tree when that particular answer is chosen. It is easy to keep track of the different steps, as each type of step is represented by a particular geometric shape. An answer line without a step attached represents incomplete knowledge and visually prompts you to build a conclusion.

Sample Tree

We are interested in your comments, questions, and suggestions.
Please E-mail us at

The GINESYS Corporation, 211 Providence Way, Rutherfordton, NC 28139
Phone: (800) 277-8338 --- Outside U.S.: 011-828-288-1112 --- FAX: (828) 288-1109

This page last updated on Sepetmber 26, 2000.