An input device which includes plural inputting units, each accepting external information, including a unit for recognizing an input time of input information accepted via each of the plural inputting means; a unit for dividing or merging the input information into a predetermined primitive analysis unit, to obtain a recognition result; a unit for estimating the input time of a unit recognition result, by using an estimating method predetermined for each of the inputting units; and a unit for collecting some of the recognition results whose estimated input times are close to one another, and outputting the collected information as a semantic analysis unit. An input content is specified by collecting plural pieces of input information being signals input from plural inputting units.