Information processing apparatus

1. A conversation agent for conversing with a user, the conversation agent comprising:

an output device for outputting a question to a user;

an audio signal acquisition section adapted to acquire and analyze an audio signal from the user in response to the question;

a time identification section adapted to identify, within the audio signal, a total thinking time and a total response time; and

an agent processing section for:

determining if the audio signal contains an answer;

determining an answer classification for the answer as an anticipated answer or as an unanticipated answer; and

comparing a difference between a standard time and the total thinking time to a predetermined threshold to determine an evaluation value Vrc for the answer,

wherein the standard time is a predetermined necessary time for the user to respond to the question;

selecting an audio response-determined by the evaluation value Vrc and the answer classification; and

outputting the audio response using the output device,

wherein Vrc is calculated using n evaluation values from Vr( 1 ) to Vr(n) acquired by n responses in the past according to a formula:

Vrc

=

∑

i

=

1

n

⁢

w

⁡

(

i

)

·

Vr

⁡

(

i

)

n

where w(i) is the weight assigned to each past evaluation value and calculated according to:

w ( i )= e −α(i-1)   

where α is a positive constant.

2. The conversation agent of claim 1 , wherein

the total filler time includes at least the time during which the user is producing a meaningless utterance, and

wherein the total thinking time further includes a time for which speech content recognition by the agent processing section failed.

3. The conversation agent of claim 1 , further comprising:

a holding section adapted to hold a parameter value used to determine content of the output produced by agent processing section, wherein

the parameter value held by the holding section is updated in accordance with the total thinking time.

4. The conversation agent of claim 3 , wherein

the agent processing section changes the audio response in accordance with the parameter value held by the holding section.

5. The conversation agent of claim 1 , wherein

the agent processing section outputs a statement directed to the user ahead of identification of the determining if the audio signal contains an answer,

the time identification section identifies the total thinking time before the user responds to the statement, and

the agent processing section changes an output after the audio response in accordance with the total thinking time.

6. The conversation agent of claim 5 , wherein

the output section outputs a statement selected from among a plurality of possible statements in accordance with an identified evaluation target time after the response.

7. The conversation agent of claim 6 , wherein the agent processing section outputs the statement selected in accordance with a plurality of selection criteria including the total thinking time.

8. An information processing method for a conversation agent comprising:

outputting a question to a user from a speaker of the conversation agent;

acquiring and analyzing an audio signal from the user in response to the question;

identifying , within the audio signal, a total thinking time and a total response time,

determining if the audio signal contains an answer;

determining an answer classification for the answer as an anticipated answer or as an unanticipated answer; and

comparing a difference between a standard time and the total thinking time to a predetermined threshold determine an evaluation value Vrc for the answer,

wherein the standard time is a predetermined necessary time for the user to respond to the question;

selecting an audio response determined by the evaluation value Vrc and the answer classification; and

outputting the audio response using the output device,

wherein Vrc is calculated using n evaluation values from Vr( 1 ) to Vr(n) acquired by n responses in the past according to a formula:

Vrc

=

∑

i

=

1

n

⁢

w

⁡

(

i

)

·

Vr

⁡

(

i

)

n

where w(i) is the weight assigned to each past evaluation value and calculated according to:

w ( i )= e −α(i-1)   

where α is a positive constant.

9. A non-transitory computer-readable information storage medium storing a program for a computer, the program comprising:

outputting a question to a user from a speaker of the conversation agent;

acquiring and analyzing an audio signal from the user in response to the question;

identifying , within the audio signal, a total thinking time and a total response time,

determining if the audio signal contains an answer;

determining an answer classification for the answer as an anticipated answer or as an unanticipated answer; and

comparing a difference between a standard time and the total thinking time to a predetermined threshold determine an evaluation value Vrc for the answer,

wherein the standard time is a predetermined necessary time for the user to respond to the question;

selecting an audio response determined by the evaluation value Vrc and the answer classification; and

outputting the audio response using the output device,.

wherein Vrc is calculated using n evaluation values from Vr( 1 ) to Vr(n) acquired by n responses in the past according to a formula:

Vrc

=

∑

i

=

1

n

⁢

w

⁡

(

i

)

·

Vr

⁡

(

i

)

n

where w(i) is the weight assigned to each past evaluation value and calculated according to:

w ( i )= e −α(i-1)   

where α is a positive constant.