Conversation analyzing device, conversation analyzing method, and program

• US 9,934,779 B2
• Filed: 02/28/2017
• Issued: 04/03/2018
• Est. Priority Date: 03/09/2016
• Status: Active Grant

• Current Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Original Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Inventors: Mizumoto, Takeshi, Hagiya, Shunichi, Sahata, Tomoyuki
• Primary Examiner(s): Ortiz Sanchez, Michael
• Application Number:
  US15/444,556
• Publication Number:
  US 20170263245A1
• Time in Patent Office: 399 Days
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

A conversation data acquiring unit acquires conversation data indicating speech of each speaker in a conversation. A conversation state analyzing unit analyzes an amount of speech of each speaker in the conversation and a degree of influence of the speech of each speaker on the conversation on the basis of the conversation data. A role determining unit determines a role of each speaker in the conversation on the basis of the amount of speech and the degree of influence of the speaker.

Claims

1. A conversation analyzing device comprising:

a conversation data acquiring unit, implemented via at least one processor, configured to acquire conversation data indicating speech of each of a plurality of speakers in a conversation;

a conversation state analyzing unit, implemented via the at least one processor, configured to analyze an amount of speech of each of the plurality of speakers in the conversation and a degree of influence of the speech of each of the plurality of speakers on the conversation based on the conversation data;

and a role determining unit, implemented via the at least one processor, configured to determine a role of each of the plurality of speakers in the conversation based on the amount of speech and the degree of influence of the speaker,

wherein the degree of influence includes a facilitator level index value, the facilitator level index value being indicative of a degree to which one of the speakers facilitates speech of other speakers in the conversation and includes as a component thereof a speech amount correction level, the speech amount correction level being indicative of a degree to which a deviation in the amount of speech among the plurality of speakers is lessened and a degree to which the amount of speech of other speakers is corrected by the one speaker from a natural state to an ideal state, wherein the natural state is an average amount of speech of the other speakers in a previous conversation and the ideal state is a state in which amounts of speech of each of the plurality of speakers are equal to each other in a current conversation,

the role determining unit determines whether the role of the one speaker is a facilitator based on the facilitator level index value;

and a display unit to output image data including data representing the determination of whether the role of the speaker is a facilitator.

2. The conversation analyzing device according to claim 1, wherein the facilitator level index value includes as a component thereof a conversation facilitation frequency which is a speech frequency for facilitating the conversation.

3. The conversation analyzing device according to claim 1, wherein the degree of influence includes an idea provider level index value including a conversation activity increasing rate which is a degree to which the conversation is activated by speech as a component, and

the role determining unit determines whether the role is an idea provider based on idea provider level index value.

4. The conversation analyzing device according to claim 3, wherein the idea provider level index value includes a conclusion mention level which is a mention frequency of a conclusive element of the conversation as a component.

5. The conversation analyzing device according to claim 1, wherein the degree of influence includes a dominator level index value indicating an interruption state of speech of another speaker, and

the role determining unit determines whether the role is a dominator on the basis of the dominator level index value.

6. The conversation analyzing device according to claim 1, further comprising a display data acquiring unit, implemented via at least one processor and a display, configured to output display data, via the display, including a diagram collectively indicating magnitudes of index values of the degree of influence, including at least the facilitator level index value, and a diagram illustrating a ratio of an amount of speech for each of the plurality of speakers.

7. The conversation analyzing device according to claim 1, further comprising:

a sound collecting unit configured to acquire a plurality of channels of voice signals; and

a sound source separating unit, implemented via the at least one processor, configured to separate voice signals associated with speech of each speaker from the plurality of channels of voice signals.

8. A conversation analyzing method in a conversation analyzing device, implemented via at least on one processor, the method comprising:

a conversation data acquiring step of acquiring conversation data indicating speech of each of a plurality of speakers in a conversation;

a conversation state analyzing step of analyzing an amount of speech of each of the plurality of speakers in the conversation and a degree of influence of the speech of each of the plurality of speakers on the conversation based on the conversation data,

wherein the degree of influence includes a facilitator level index value, the facilitator level index value being indicative of a degree to which one of the speakers facilitates speech of other speakers in the conversation and includes as a component thereof a speech amount correction level, the speech amount correction level being indicative of a degree to which a deviation in the amount of speech among the plurality of speakers is lessened and a degree to which the amount of speech of other speakers is corrected by the one speaker from a natural state to an ideal state, wherein the natural state is an average amount of speech of the other speakers in a previous conversation and the ideal state is a state in which amounts of speech of each of the plurality of speakers are equal to each other in a current conversation;

a role determining step of determining a role of each of the plurality of speakers in the conversation based on the amount of speech and the degree of influence of the speaker, including determining whether the role of the one speaker is a facilitator based on the facilitator level index value;

and presenting on a display output data including data representing the determination of whether the role of the speaker is a facilitator.

9. The conversation analyzing method of claim 8, further comprising:

a display data acquiring step of outputting display data, via a display, including a diagram collectively indicating magnitudes of index values of the degree of influence, including at least the facilitator level index value, and a diagram illustrating a ratio of an amount of speech for each of the plurality of speakers.

10. A program stored on a non-transitory computer readable medium which, when executed by a processor of a computer, causes the computer to perform:

a conversation data acquiring sequence of acquiring Conversation data indicating speech of each of a plurality of speakers in a conversation;

a conversation state analyzing sequence of analyzing an amount of speech of each of the plurality of speakers in the conversation and a degree of influence of the speech of each of the plurality of speakers on the conversation based on the conversation data, wherein the degree of influence includes a facilitator level index value, the facilitator level index value being indicative of a degree to which one of the speakers facilitates speech of other speakers in the conversation and includes as a component thereof a speech amount correction level, the speech amount correction level being indicative of a degree to which a deviation in the amount of speech among the plurality of speakers is lessened and a degree to which the amount of speech of other speakers is corrected by the one speaker from a natural state to an ideal state, wherein the natural state is an average amount of speech of the other speakers in a previous conversation and the ideal state is a state in which amounts of speech of each of the plurality of speakers are equal to each other in a current conversation;

a role determining sequence of determining a role of each of the plurality of speakers in the conversation based on the amount of speech and the degree of influence of the speaker, including determining Whether the role of the one speaker is a facilitator based on the facilitator level index value;

and presenting on a display output data including data representing the determination of whether the role of the speaker is a facilitator.

11. The program according to claim 10, further comprising:

a display data acquiring sequence of outputting display data, via a display, including a diagram collectively indicating magnitudes of index values of the degree of influence, including at least the facilitator level index value, and a diagram illustrating a ratio of an amount of speech for each of the plurality of speakers.