In this paper, the validity of syntactic pattern recognition by context-free grammar with error-correcting is made sure from a theoretical point and results on a simulation for hand-written Katakana characters. Firstly, we define three types of syntax errors to debase strings and propose a distance between one syntactic pattern (a string) and a group of syntactic patterns (a set of strings) showing the smallest number of errors to parse. Secondly, we introduce an algorithm on the error-correcting parser for context-free grammar to find the distance, and apply the algorithm to the pattern recognition of hand-written characters. Finally, we show the results on a simulation for hand-written Katakana characters.