To correct a voice-recognized character string through an easy operation.
In a dictionary 13 for dictation, words for dictation recognition are previously registered. A dictation voice that a speaker utters is inputted to a voice input device 11, analyzed by a voice recognition part 12, and converted into a character string by referring to the dictionary 13 for dictation. A word registration/retrieval processing part 15 displays it on a display device 16. At the same time, all the words included in the displayed character string are registered in a command dictionary 14 for retrieval. Then when the speaker looks at the display on the display device 16 and utters a corresponding word to correct the character string, the voice recognition part 12 analyzes and converts it into a character string, by referring to the command dictionary 14 for retrieval. The word registration/retrieval processing part 15 points out the display position of the converted character string on the display device 16.