Dialogue system and dialogue method

1. A dialogue system comprising:

a convolutional neural network including at least an input layer; and

circuitry configured to:

estimate a satisfaction of a user based on a speech input from the user by applying an electrical signal into the input layer, wherein the input layer creates a multi-dimensional vector by using the Fourier transformation to convert the electrical signal as the speech input;

estimate a dialogue state with the user based on the speech input from the user and the estimated satisfaction of the user;

update the estimated dialogue state based on a previous dialogue state stored in the circuitry; and

determine a behavior towards the user based on the updated dialogue state.

2. The dialogue system according to claim 1 , wherein the circuitry is further configured to estimate the satisfaction of the user by applying input data that is based on the speech input to a learnt statistical model that is created by performing supervised learning using learn-specific speech input with a label indicative of a satisfaction.

3. The dialogue system according to claim 1 , wherein the satisfaction is expressed as a probability distribution.

4. The dialogue system according to claim 1 , wherein the circuitry is further configured to determine the behavior towards the user by applying input data that is based on the dialogue state to a learnt statistical model that is created by performing reinforcement learning.

5. The dialogue system according to claim 4 , wherein a reward used in the reinforcement learning is set according to a satisfaction of a user who uttered a learn-specific speech input, estimated based on the learn-specific speech input.

6. A dialogue system comprising:

a convolutional neural network including at least an input layer; and

circuitry configured to:

estimate a satisfaction of a user based on a speech input from the user by applying an electrical signal into the input layer, wherein the input layer creates a multi-dimensional vector by using the Fourier transformation to convert the electrical signal as the speech input;

estimate a dialogue state with the user based on an input from the user;

update the estimated dialogue state based on a previous dialogue state stored in the circuitry; and

determine a behavior towards the user based on the estimated satisfaction and the updated dialogue state.

7. A dialogue system comprising:

a convolutional neural network including at least an input layer; and

circuitry configured to:

estimate a satisfaction of a user based on a speech input from the user by applying an electrical signal into the input layer, wherein the input layer creates a multi-dimensional vector by using the Fourier transformation to convert the electrical signal as the speech input;

estimate a dialogue state of a dialogue with the user based on at least an input from the user;

update the estimated dialogue state based on a previous dialogue state stored in the circuitry;

determine a behavior towards the user based on the updated dialogue state; and

determine whether or not to discontinue the dialogue with the user based on at least the estimated satisfaction.

8. The dialogue system according to claim 7 , wherein the circuitry is further configured to determine whether or not to discontinue the dialogue with the user based on the dialogue state and the estimated satisfaction.

9. The dialogue system according to claim 7 , wherein the circuitry is further configured to automatically hand over the dialogue with the user to an operator when the dialogue with the user is determined to be discontinued.