Confusion set recognition of on-line Chinese characters by artificial intelligence technique

Abstract In this paper, a knowledge based on-line recognition system for the square written Chinese similar characters is presented by utilizing Chinese character knowledge and artificial intelligence techniques. The reference patterns are represented by the heuristic knowledge describing them. The knowledge consists of the knowledge primitives designed in the system. For the pattern pair matching made on an unknown pattern, the matching recognition is made by utilizing problem reduction strategy, goal-driven inferencing like matching, and AND-OR tree searching algorithm. The method is developed and concatenated to a first on-line recognition system specially for recognizing similar characters. The recognized similar characters to be researched in the experimental system are chosen from the part of the error recognition results of the first system, totally 96 similar characters. The average recognition rate of 99.9% can be achieved. The results suggest that the knowledge can present efficient distinguishing ability and the method is reasonable and feasible.