PROBLEM TO BE SOLVED: To improve the accuracy of the automatic selection of a task through the sound in a communication network. SOLUTION: A selected task is discriminated through the naturally spoken sound by a user who makes a selection. The meaningful phrase is selected from the sound for test through the use of the meaningful test (a meaningful phrase selecting device 10). The selected meaningful phrase is clustered (a meaningful phrase clustering device 15). The information on the meaningful phase cluster is supplied to a meaningful phrase cluster detecting device 25 and a classifying processor 30. The meaningful phrase cluster detecting device 25 judges whether or not the meaningful phrase cluster is present in the inputted sound recognized by an inputted speech recognizing device 20. Then, the classifying processor 30 determines the task kind based on the meaningful phrase cluster recognized in the inputted sound.
展开▼