Voice interaction apparatus and voice interaction method

1. A voice interaction apparatus configured to have a conversation with a user by using a voice, comprising:

a speech acquisition unit configured to acquire user speech, the user speech being speech given by the user;

a syntactic analysis unit configured to perform a syntactic analysis for linguistic information on the acquired user speech;

a response generation unit configured to generate a first response according to the user speech;

a voice output unit configured to output a voice for the user;

a non-linguistic information analysis unit configured to analyze non-linguistic information for the acquired user speech, the non-linguistic information being different from the linguistic information and including at least one of prosodic information on the user speech and history information about a second response generated by the response generation unit;

a filler length determination unit configured to determine a length of a filler output by the voice output unit according to a non-linguistic information analysis result, the non-linguistic information analysis result being a result of an analysis by the non-linguistic information analysis unit; and

a filler generation unit configured to generate a filler having a length corresponding to a result of a determination by the filler length determination unit, wherein

the filler length determination unit determines that a long filler should be generated when a syntactic analysis result needs to be used to generate the first response and determines that a short filler should be generated when the syntactic analysis result does not need to be used to generate the first response, the syntactic analysis result being a result of an analysis by the syntactic analysis unit, and

the voice output unit outputs the first response generated by the response generation unit after outputting the filler.

2. The voice interaction apparatus according to claim 1 ,

wherein the filler length determination unit determines whether or not the acquired user speech is a question put to the voice interaction apparatus, and

wherein when the filler length determination unit determines that the acquired user speech is a question put to the voice interaction apparatus, the filler length determination unit determines that a long filler should be generated;

the voice output unit outputs the long filler generated by the filler generation unit;

the response generation unit generates an answer to the question as the first response by using the syntactic analysis result; and

the output unit outputs the generated answer.

3. The voice interaction apparatus according to claim 1 ,

wherein the filler length determination unit determines whether or not the acquired user speech is a question put to the voice interaction apparatus, and

wherein when the filler length determination unit determines that the acquired user speech is not a question put to the voice interaction apparatus, the filler length determination unit determines that a short filler should be generated;

the voice output unit outputs the short filler generated by the filler generation unit;

the response generation unit generates a response for guiding the conversation to a different topic without using the syntactic analysis result; and

the output unit outputs the generated response for guiding the conversation to a different topic.

4. The voice interaction apparatus according to claim 1 , wherein the filler length determination unit determines the length of the filler output by the voice output unit based on a comparison between at least one feature quantity included in the non-linguistic information analysis result and a predetermined threshold corresponding to the feature quantity.

5. The voice interaction apparatus according to claim 1 , wherein the filler length determination unit determines the length of the filler by determining whether or not a feature indicated in the non-linguistic information analysis result corresponds to a necessity to use the syntactic analysis result to generate the first response by using a determination model that is generated in advance through mechanical learning.

6. A voice interaction method performed by using a voice interaction apparatus configured to have a conversation with a user by using a voice, comprising:

acquiring user speech, the user speech being speech given by the user;

performing a syntactic analysis for linguistic information on the acquired user speech;

generating a first response according to the user speech analyzing non-linguistic information for the acquired user speech, the non-linguistic information being different from the linguistic information and including at least one of prosodic information on the user speech and history information about a second response generated by the voice interaction apparatus;

determining whether or not a syntactic analysis result needs to be used to generate the first response according to a non-linguistic information analysis result, the syntactic analysis result being a result of the syntactic analysis, the non-linguistic information analysis result being a result of the analysis of the non-linguistic information;

generating and outputting a long filler when it is determined that the syntactic analysis result needs to be used to generate the first response, and generating and outputting a short filler when it is determined that the syntactic analysis result does not need to be used to generate the response; and

outputting a voice corresponding to the first response generated according to the user speech after outputting the filler.