摘要 |
PROBLEM TO BE SOLVED: To improve the precision of rough classification without increasing the number of candidate character categories given to detailed classification. SOLUTION: The precision can be improved without increasing the number of candidate character categories given to a detailed classification part 2 by outputting more candidate character categories than usual from a rough classification part 1 and narrowing down them to candidates suitable as a language by a language processing part 21 for rough classification. At this time, a language dictionary 11 used by the language process part 3 is used also as a language dictionary for the language processing part 21 for rough classification, so the addition of a new dictionary accompanied by a great increase of resources is eliminated. The language processing part 21 for rough classification is much faster than the rough classification part 1 and the detailed classification part 2, so the total decrease in character recognition speed is very small.
|