To solve the problem that, in general, when speech uttered by a user is recognized and converted into a character string, the character string after conversion is presented to the user for correction if there is an error, and thereby, incorrect conversion is prevented, but the user's one by one input is required in this method, and it takes long time to correctly recognize the speech.

A link character string candidate which is obtained by linking to a determined character string, is compared with a recognition character string candidate generated from a signal for character recognition, and the character string which is coincident (or similar) is made the determined character string, in an input support system of the invention. Moreover, the link character string candidate which is to be compared with the generated recognition character string candidate, is made into a postposition link character string candidate which is postposed to the determined character string candidate which is already selected, and/or preposition link character string candidate which is preposed.