PURPOSE: To remarkably reduce labor required for a similarity decision, and also, to prevent an erroneous decision as much as possible by providing the data being extremely effective for the similarity decision of each character string data.

CONSTITUTION: In the case character-strings subjected to similarity decision are, for instance, 'ARUSHIN' and 'ARUSAN', ' SHI ' and ' SA ' being different parts are fixed, respectively, and on the other hand, 'ARU N' being a coincident part is replaced with a character '?' for showing an arbitrary character, and its retrieval is executed, respectively. Moreover, from both retrieval results, those which coincide as to the '?' part, for instance, 'MARUSHIN' and 'MARUSAN' are retrieved and they are compared and outputted.