摘要 |
PURPOSE:To add a learning function and to improve a recognition ratio by retrieving the dictionary of a wooden structure and adding a path to a correct character to a retrieving path when the retrieved recognition result is erroneous. CONSTITUTION:When a recognition control part 2 inputs, for instance, the quantity of the feature of an unknown character, a down retrieval is carried out along the wooden structure of the dictionary 1, and a finally obtained recognition result is displayed by a character, for instance, and outputted. For instance, only when an operator decides a correction, he applies correct character information to the recognition control part 2 only at the time of an error. Thereby, the recognition control part 2 stores the dictionary 1 so as to apply the retrieving path to the correct character in the wooden structure of the dictionary 1 to learn so as not to have the error to a next character retrieval.
|