DRAWING UNDERSTANDING SYSTEM INCORPORATING RULE GENERATION SUPPORT WITH MAN-MACHINE INTERACTIONS

The present study describes using the state transition type of drawing understanding framework to construct a multi-purpose drawing understa nding system. This new system employs an understanding process that co mplies with the understanding rules, which are easily obtained by the user. The same set of user-provided rules must be used for the same ty pe of target drawings, but for slightly different ones, fine tuning is required to obtain understanding rules. To overcome this inherent dra wback in constructing drawing understanding systems, we extended the s ystem using a newly constructed understanding rule generating support system. The resultant integrated system is based on a man-machine coop eration type interface, and can automatically generate rules from user -provided simple interactions using a graphical user interface (GUI). To obtain efficient rule generation, the system employs an inductive i nference method as a learning algorithm. Map-drawing experiments were successfully carried out, and an evaluation based on a rule learning e rror criterion subsequently revealed an efficient rule generation proc ess.