Conversation control apparatus, conversation control method, and programs therefor

• US 20050144013A1
• Filed: 11/19/2004
• Published: 06/30/2005
• Est. Priority Date: 11/20/2003
• Status: Active Grant

• Current Assignee: Universal Entertainment Corporation
• Original Assignee: P B Limited (Balfour Beatty Plc), Universal Entertainment Corporation
• Inventors: Katukura, Hiroshi, Huang, Shengyang, Fujimoto, Jun
• Granted Patent: US 7,676,369 B2
• Time in Patent Office: Days
• US Class Current: 704/277
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/227 of the speaker; Human-factor methodology

Abstract

The present invention is a conversation control apparatus for carrying out conversation control based on both speech content and conversation state and information acquired from outside, and carrying out natural responses. This apparatus comprises a conversation database storing a plurality of items of conversation topic specifying information, a conversation control unit for selecting a reply sentence using conversation topic specifying information determined using the conversation history and conversation information, and an environment recognition unit for acquiring environment recognition information, wherein the environment recognition unit generates event information based on acquired environment recognition information, and the conversation control unit selects a reply sentence based on the event information.

Claims

1. A conversation control apparatus for outputting a response in reply to speech from a user, comprising:

a conversation database unit storing a plurality of items of conversation topic specifying information;

a conversation control unit for storing a conversation history determined using previous conversations and previous replies and selecting a reply sentence using conversation topic specifying information determined using the conversation history and conversation information; and

an environment recognition unit for acquiring environment recognition information, wherein

the environment recognition unit generates event information based on acquired environment recognition information, and the conversation control unit selects a reply sentence based on the event information.

2. The conversation control apparatus according to claim 1, wherein the event information includes at least one of emotion estimation information, gender estimation information, age information and speaker estimation information.

3. The conversation control apparatus according to claim 1, wherein the event information includes emotion estimation information which indicates speaker emotion estimated from the speaker'"'"'s image output.

4. The conversation control apparatus according to claim 1, wherein the event information contains gender estimation information which indicates a speaker'"'"'s gender from the speaker'"'"'s image output.

5. The conversation control apparatus according to claim 1, wherein the event information contains speaker estimation information which is used to specify respective speakers when plural speakers simultaneously uses the conversation control apparatus.

6. The conversation control apparatus according to claim 5, wherein the speaker estimation information is generated on the basis of at least one of the speaker'"'"'s voiceprint and the speaker'"'"'s position.

7. The conversation control apparatus according to claim 1, wherein the event information contains speaker proximity notification information generated when a particular speaker is close enough to the conversation control apparatus.

8. A conversation control apparatus for outputting a response in reply to speech from a user, comprising:

a conversation database unit storing a plurality of items of conversation topic specifying information;

a conversation control unit for storing a conversation history determined using previous conversations and previous replies and selecting a reply sentence using conversation topic specifying information determined using the conversation history and conversation information;

an environment recognition unit for acquiring environment recognition information; and

an emotional condition information management unit for storing and updating emotional condition information representing an emotion of at least one of a character portrayed by the conversation control unit and a speaker, wherein;

the environment recognition unit generates event information based on acquired environment recognition information, the emotional condition information management unit updates emotional condition information in response to event information, and the conversation control unit selects a reply sentence based on emotional condition information.

9. The conversation control apparatus according to claim 8, wherein the emotional condition information includes at least one of character emotional condition information representing the emotion of the character, and speaker emotional condition information representing the emotion of the speaker.

10. The conversation control apparatus according to claim 9, wherein the conversation control apparatus displays at least one of the expression on the face and the movement the character changeably on the basis of the character emotional condition information.

11. A conversation control method for outputting a response in reply to speech from a user, comprising the steps of:

previously storing a plurality of items of topic specifying information, one or a plurality of conversation titles determined for each topic specifying information, one or a plurality of reply sentences determined for each conversation title, and event information flags determined in response to the reply sentences;

acquiring environment recognition information from an external environment and generating event information from the environment recognition information; and

collating conversation topic specifying information determined using conversation history determined using previous conversations and previous replies to acquire at least a conversation title determined for any of the conversation topic specifying information, and selecting a reply sentence corresponding to an event information flag corresponding to the event information from a plurality of reply sentence corresponding to that conversation title.

12. The conversation control method according to claim 11, wherein the event information includes at least one of emotion estimation information, gender estimation information, age information and speaker estimation information.

13. The conversation control method according to claim 11, wherein the event information includes emotion estimation information which indicates speaker emotion estimated from the speaker'"'"'s image output.

14. The conversation control method according to claim 11, wherein the event information contains gender estimation information which indicates a speaker'"'"'s gender from the speaker'"'"'s image output.

15. The conversation control method according to claim 11, wherein the event information contains speaker estimation information which is used to specify respective speakers when plural speakers simultaneously uses an apparatus carrying out the conversation control method.

16. The conversation control method according to claim 15, wherein the speaker estimation information is generated on the basis of at least one of the speaker'"'"'s voiceprint and the speaker'"'"'s position.

17. The conversation control method according to claim 11, wherein the event information contains speaker proximity notification information generated when a particular speaker is close enough to an apparatus carrying out the conversation control method.

18. A conversation control method for outputting a response in reply to speech from a user, comprising the steps of:

previously storing a plurality of items of topic specifying information, one or a plurality of conversation titles determined for each topic specifying information, and one or a plurality of reply sentences determined for each conversation title;

acquiring environment recognition information from an external environment and generating event information from the environment recognition information;

updating emotional condition information representing emotion of at least one of a character or a speaker according to the event information; and

collating conversation topic specifying information determined using conversation history determined using previous conversations and previous replies to acquire at least a conversation title determined for any of the conversation topic specifying information, and selecting a reply sentence corresponding to the event information and emotional condition information from a plurality of reply sentence corresponding to that conversation title.

19. The conversation control method according to claim 18, wherein the emotional condition information includes at least one of character emotional condition information representing the emotion of the character, and speaker emotional condition information representing the emotion of the speaker.

20. The conversation control method according to claim 19, wherein the conversation control method further includes a step of displaying at least one of the expression on the face and the movement the character changeably on the basis of the character emotional condition information.

21. A program product for implementing a conversation control method for outputting a response in reply to speech from a user, comprising the steps of:

previously storing a plurality of items of topic specifying information, one or a plurality of conversation titles determined for each topic specifying information, one or a plurality of reply sentences determined for each conversation title, and event information flags determined in response to the reply sentences;

acquiring environment recognition information from an external environment and generating event information from the environment recognition information; and

collating conversation topic specifying information determined using conversation history determined using previous conversations and previous replies to acquire at least a conversation title determined for any of the conversation topic specifying information, and selecting a reply sentence corresponding to an event information flag corresponding to the event information from a plurality of reply sentence corresponding to that conversation title.

22. A program product for A conversation control method for outputting a response in reply to speech from a user, comprising the steps of:

previously storing a plurality of items of topic specifying information, one or a plurality of conversation titles determined for each topic specifying information, and one or a plurality of reply sentences determined for each conversation title;

acquiring environment recognition information from an external environment and generating event information from the environment recognition information;

updating emotional condition information representing emotion of at least one of a character or a speaker according to the event information; and

collating conversation topic specifying information determined using conversation history determined using previous conversations and previous replies to acquire at least a conversation title determined for any of the conversation topic specifying information, and selecting a reply sentence corresponding to the event information and emotional condition information from a plurality of reply sentence corresponding to that conversation title.