Response generation method, response generation apparatus, and response generation program

• US 9,653,078 B2
• Filed: 08/14/2015
• Issued: 05/16/2017
• Est. Priority Date: 08/21/2014
• Status: Active Grant

• Current Assignee: Toyota Jidosha Kabushiki Kaisha (Toyota Motor Corporation)
• Original Assignee: Toyota Jidosha Kabushiki Kaisha (Toyota Motor Corporation)
• Inventors: Watanabe, Narimasa
• Primary Examiner(s): COLUCCI, MICHAEL C
• Application Number:
  US14/826,280
• Publication Number:
  US 20160055849A1
• Time in Patent Office: 641 Days
• Field of Search: 704235, 704241, 704244, 704251, 704275, 704 3, 704 9, 379 8818, 455411
• CPC Class Codes: G10L 15/00 Speech recognition G10L17/00 takes precedence
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques
• CPC Class Codes: G10L 17/22 Interactive procedures; Man-machine interfaces

Abstract

A response generation method includes a step of recognizing a voice of a user, a step of analyzing a structure of the recognized voice, a step of generating a free response sentence in response to the voice of the user based on the analyzed voice structure and outputting the generated free response sentence, a step of generating the recognized voice of the user as a repeat response sentence, and a step of outputting the generated repeat response sentence before outputting the free response sentence based on the voice structure.

Claims

1. A response generation method comprising:

a step of recognizing a voice of a user by performing voice recognition processing based on the voice of the user acquired though a microphone and generating character-string information of the voice of the user;

a step of analyzing a structure of the recognized voice for the character-string information indicating recognized voice information of the user by using a structure analysis unit; and

a step of generating a free response sentence in response to the voice of the user based on the analyzed voice structure by using a response generation unit and outputting the generated free response sentence by using a speaker, wherein

the response generation method further comprises;

a step of generating a repeat response sentence by extracting a keyword and its part of speech from the character-string information of the recognized voice of the user, selecting an additional word corresponding to the extracted keyword and its part of speech based on an additional information stored in a memory, and adding the selected word to the extracted keyword;

a step of analyzing a phoneme of the voice of the user by detecting a change in a volume level or a frequency of the voice of the user acquired through the microphone to infer a break in the voice of the user;

a step of generating a responsive response to the voice of the user based on an analysis result of the phoneme; and

a step of outputting the generated repeat response sentence before outputting the generated free response sentence based on the voice structure by using the speaker, and

the generated responsive response is output before the generated repeat response sentence is output.

2. The response generation method according to claim 1, wherein

a plurality of response candidates for responding to the voice of the user are generated based on the analyzed voice structure,

among the plurality of generated response candidates, a response candidate that is identical to the generated repeat response sentence or the generated responsive response is excluded, and

a response candidate selected from among the response candidates, for which the exclusion has been performed, is used as the free response sentence.

3. The response generation method according to claim 2, wherein the repeat response sentence is generated by extracting a keyword and its part of speech from the recognized voice of the user, selecting, based on additional information about correspondences among a plurality of keywords, their respective parts of speech and their respective additional word endings, the additional word ending corresponding to the extracted keyword and its part of speech, and adding the selected additional word ending to the extracted keyword.

4. A response generation apparatus comprising:

voice recognition means for recognizing a voice of a user by performing voice recognition processing based on the voice of the user acquired though a microphone and generating character-string information of the voice of the user;

structure analysis means for analyzing a structure of the voice recognized by the voice recognition means for the character-string information indicating recognized voice information of the user; and

response output means for generating a free response sentence in response to the voice of the user based on the voice structure analyzed by the structure analysis means and outputting the generated free response sentence by using a speaker, wherein

the response generation apparatus further comprises;

repetition generation means for generating a repeat response sentence by extracting a keyword and its part of speech from the character-string information of the recognized voice of the user, selecting an additional word corresponding to the extracted keyword and its part of speech based on an additional information stored in a memory, and adding the selected word to the extracted keyword;

phoneme analysis means for analyzing a phoneme of the voice of the user by detecting a change in a volume level or a frequency of the voice of the user acquired through the microphone to infer a break in the voice of the user; and

responsive-response generation means for generating a responsive response to the voice of the user based on an analysis result of the phoneme,

the response output means outputs the repeat response sentence generated by the repetition generation means before outputting the generated free response sentence based on the voice structure by using the speaker, and

the generated responsive response is output before the generated repeat response sentence is output.

5. A non-transitory computer readable medium storing a response generation program for causing a computer to execute:

a process of recognizing a voice of a user including performing voice recognition processing based on the voice of the user acquired though a microphone and generating character-string information of the voice of the user;

a process of analyzing a structure of the recognized voice for the character-string information indicating recognized voice information of the user by using a structure analysis unit;

a process of generating a free response sentence in response to the voice of the user based on the analyzed voice structure by using a response generation unit and outputting the generated free response sentence by using a speaker;

a process of generating a repeat response sentence by extracting a keyword and its part of speech from the character-string information of the recognized voice of the user, selecting an additional word corresponding to the extracted keyword and its part of speech based on an additional information stored in a memory, and adding the selected word to the extracted keyword;

a process of analyzing a phoneme of the voice of the user by detecting a change in a volume level or a frequency of the voice of the user acquired through the microphone to infer a break in the voice of the user;

a process of generating a responsive response to the voice of the user based on an analysis result of the phoneme; and

a process of outputting the generated repeat response sentence before outputting the generated free response sentence based on the voice structure by using the speaker,

wherein the generated responsive response is output before the generated repeat response sentence is output.