JPWO2018015999A1 - Discussion support system, discussion support method, and discussion support program - Google Patents

Abstract

  A discussion support system 10 that supports discussion on an agenda composed of a plurality of elements, which includes a speech acquisition unit 201 that acquires the content of a speech made by a speaker on the agenda, and the acquired speech as a plurality of words Based on the decomposed word analysis processing unit 203 and the decomposed word, the element corresponding to the acquired message is identified, and it is determined whether there is another element having a causal relationship with the identified element. And a question generation unit 210 that generates a question about the agenda based on the other element, and a question output unit 212 that outputs the generated question based on the other element. .

Description

  The present invention relates to a discussion support system, a discussion support method, and a discussion support program.

  In recent years, attempts have been made to create new ideas by utilizing a framework for market analysis and corporate strategy formulation. In meetings to create new ideas, the meeting moderator supports the creation of new ideas by proceeding with the agenda by gathering the opinions of the participants and asking the participants.

  One method for conducting such a conference is to apply system dynamics. System dynamics identifies elements (nodes) that affect system fluctuations and models causal relationships between the nodes. Then, by performing a simulation based on the model, characteristics that change with the passage of time of the system are clarified. In system dynamics, a causal loop diagram is used as a method for modeling a causal relationship between nodes. The causal loop diagram is composed of nodes and arrows representing the causal relationship. The arrow expresses a positive correlation that when one increases (or decreases) the other increases (or decreases), and when one increases (or decreases) the other decreases (or increases) That expresses negative correlation. The causal loop diagram is suitable for modeling a characteristic that changes with the passage of time because a node that takes time until an effect appears can be expressed.

  Applying this causal loop diagram to create a causal loop diagram with the elements constituting the conference agenda as nodes and visualizing the content of the agenda facilitates the understanding of participants on the conference agenda and the activation of the discussion It has been tried to do. As this related technology, there are Patent Literature 1 and Patent Literature 2. Patent Document 1 discloses that information is presented according to a scene in a conference. Further, Patent Document 2 discloses that a topic such as a related topic that is not discussed at a conference is presented to the user so that the discussion is developed.

JP 2014-85916 A Japanese Patent Laying-Open No. 2015-156099

  In Patent Document 1, a conference state change condition and a conference state change action are used in order to present information at an appropriate timing according to a meeting scene. The conference status change condition includes: silent time, frequency of occurrence of specific words, frequency of occurrence of specific contexts, time when the same speaker continues speaking, time when a specific person does not speak continuously, specific speech Or at least one selected from the group consisting of the content of specific comments. In addition, the conference state change action includes a message for prompting rest, a text display of speech, a display of frequently searched keyword internet search results, a display of search results of frequently used keywords in the document management system, or a text / image that brightens the atmosphere. Is defined. Then, when at least one of the conference state change conditions described above is satisfied, a conference state change action associated with the condition is executed. However, since the system of Patent Document 1 executes a conference state change action one after another when a predetermined condition is satisfied, when a participant of the conference inputs each request individually, When a participant designates the state of the conference and displays it on the screen, there is a possibility that the moderator cannot control the conference.

  In Patent Document 2, in order to present a topic of a conference to a user, first, a speech of an ongoing conference is converted into text and a keyword is extracted by morphological analysis. Then, topics are extracted from documents collected from an external public database in advance and stored in an external information DB (Database). Thereafter, a topic having a high similarity to the extracted keyword is extracted from the external information DB and presented on the monitor. However, in Patent Document 2, since the topic acquired from the external information DB is expressed by a word (word) similar to the extracted keyword, it may be insufficient as a motive for creating a new idea. For example, it is difficult for a party to create a new viewpoint and a new idea by combining unknown keywords.

  The present invention has been made in view of such circumstances, and an object thereof is to provide a discussion support system, a discussion support method, and a discussion support program capable of presenting a new development discussion.

  In order to solve the above-described problem, one aspect of the present invention is a discussion support system that supports discussion on an agenda composed of a plurality of elements, and obtains the content of a statement made by a speaker on the agenda An acquisition unit; a speech analysis processing unit that decomposes the acquired speech into a plurality of words; and the element corresponding to the acquired speech is identified based on the decomposed word, and has a causal relationship with the identified element It is determined whether there is another element, and when it is determined that there is the other element, a question generation unit that generates a question about the agenda based on the other element, and the generated question And a question output unit for outputting.

  According to the present invention, a new development discussion can be presented.

FIG. 1 is a diagram illustrating an example of the configuration of the discussion support system according to the present embodiment. FIG. 2 is a diagram for explaining an example of the hardware configuration of the discussion support apparatus 100. FIG. 3 is a diagram for explaining functions provided in the discussion support apparatus 100. FIG. 4 is a diagram illustrating an example of the utterance information 202. FIG. 5 is a diagram illustrating an example of the natural language classification dictionary 204. FIG. 6 is a diagram illustrating an example of the evaluation expression dictionary 205. FIG. 7 is a diagram illustrating an example of the utterance analysis result information 206. FIG. 8 is a diagram illustrating an example of the agenda element information 208. FIG. 9 is a diagram illustrating an example of a causal loop diagram. FIG. 10 is a diagram illustrating an example of the agenda related case information 209. FIG. 11 is a flowchart illustrating an outline of processing performed by the discussion support apparatus 100. FIG. 12 is a flowchart illustrating details of the question sentence generation process. FIG. 13 is a flowchart illustrating details of the question sentence generation process. FIG. 14 is an example of the discussion screen.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<< System configuration >>
FIG. 1 is a diagram illustrating an example of the configuration of the discussion support system according to the present embodiment. As shown in the figure, the discussion support system 10 is introduced into a conference room, a hall or the like in which a plurality of participants 13 including a speaker 11 and a moderator 12 participate and discuss.

  Each participant 13 discusses an agenda composed of a plurality of elements (hereinafter also referred to as nodes). That is, this agenda is composed of a plurality of topics or themes, and each participant 13 discusses each topic or theme.

  The discussion support system 10 is provided with a screen 14 that displays the contents of the agenda and the comments of the participants 13 so that all the participants 13 can confirm the contents of the discussion.

  On the screen 14, a diagram showing a causal relationship between elements (nodes) constituting the agenda (hereinafter referred to as a causal loop diagram) is displayed. Details of the causal loop diagram will be described later.

  The screen 14 is a monitor such as a liquid crystal display, for example. In this case, the information is displayed on the screen 14 by connecting an information processing apparatus (server, personal computer, or the like) storing the information on the conference material to the screen 14, for example. The screen 14 may be a conference room wall or a whiteboard. In this case, the information is displayed on the screen 14 by using, for example, a projector to which the information processing apparatus is connected.

  As shown in FIG. 1, the discussion support system 10 is provided with a discussion support device 100.

<< Hardware Configuration of Discussion Supporting Device 100 >>
FIG. 2 is a diagram for explaining an example of the hardware configuration of the discussion support apparatus 100. As shown in the figure, the discussion support device 100 is an information processing device (computer), and as a hardware, a central processing device 101 including a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM ( A storage device 102 such as a read only memory (NVRAM) or a non-volatile RAM (NVRAM), a hard disk (Hard Disk Drive), an SSD (Solid State Drive), a CD-ROM (Compact Disc-Read Only Memory) drive, etc. The storage device 103 includes an input device 104 including a keyboard, a mouse, and a touch panel, an output device 105 including a monitor (display), and a communication device 106 such as a LAN (Local Area Network) card. These devices are communicably connected by a bus 107.

<< Functions of Discussion Supporting Device 100 >>
Next, functions provided in the discussion support apparatus 100 will be described.
FIG. 3 is a diagram for explaining functions provided in the discussion support apparatus 100. As shown in the figure, the discussion support apparatus 100 includes functions of a speech receiving unit 201, a speech analysis processing unit 203, an idea creation support processing unit 207, a question sentence generation processing unit 210, and a display processing unit 211.

  In addition, the discussion support apparatus 100 stores information of speech information 202, a natural language classification dictionary 204, an evaluation expression dictionary 205, speech analysis result information 206, agenda element information 208, and agenda related case information 209.

  The speech receiving unit 201 (speech acquisition unit) acquires the content of a speech made by a speaker on the agenda item. Specifically, for example, the speech receiving unit 201 digitizes the speech (voice) of each participant 13 acquired from a microphone device connected to the speech receiving unit 201 or a device having a speech recognition function. The contents of the speech made by the participant 13 are acquired. Alternatively, the speech receiving unit 201 acquires the content of the speech from the participant 13 by acquiring the speech input to the information processing device (for example, an electronic tablet) used by the participant 13 via the communication device 106. May be. Note that the speech receiving unit 201 displays the content of the acquired speech on the screen 14 via communication means such as the communication device 106.

  The message receiving unit 201 stores the acquired content of the message in the message information 202.

<Remark information 202>
Here, the comment information 202 will be described.
FIG. 4 is a diagram illustrating an example of the utterance information 202. As shown in the figure, the utterance information 202 is a database storing the contents of utterances, and includes items of an utterance ID 301, a parent utterance ID 302, a corresponding node ID 303, an utterance 304, an utterance date 305, and a utterer 306. It is composed of at least one record.

  The speech ID 301 stores an identifier for identifying the speech of the participant 13 (hereinafter referred to as a speech ID), and the parent speech ID 302 includes a speech that causes the speech of the speech ID 301 (hereinafter referred to as a parent speech). An identifier for identification (hereinafter referred to as a parent speech ID) is stored. As for the parent speech ID, for example, the moderator 12 specifies the parent speech of the speech of the speech ID 301 while listening to the speech of the participant 13, and sets the speech ID of the identified parent speech in the parent speech ID 302. Further, for example, the speech receiving unit 201 analyzes the speech time and content of each speech of the participant 13, identifies the parent speech based on the context of each speech and the citation status of each speech, and identifies the identified parent speech Is set as the parent message ID 302.

  Corresponding node ID 303 stores an identifier (hereinafter referred to as a node ID) that identifies a node (hereinafter referred to as a corresponding node) corresponding to the statement of statement ID 301, and statement 304 stores the content of the statement of statement ID 301. In the utterance date and time 305, information of the date and time when the utterance of the utterance ID 301 is acquired (hereinafter referred to as the utterance date and time) is stored, and in the speaker 306, information for identifying the participant who has spoken (for example, the name of the speaker) And the identifier (hereinafter referred to as a speaker ID). In these items, for example, a character string or a numerical value is stored.

  Note that the corresponding node ID 303, for example, when the presenter 12 is in a meeting using a causal loop diagram, heard a statement (a remark of the participant 13) in which the remark ID is set in the remark ID 301. The node ID corresponding to the statement is specified above, and the specified node ID is set as the corresponding node ID 303.

  In addition, the discussion support apparatus 100 may acquire the content of the utterances of the participant 13 by using the electronic sticky note function by providing the electronic sticky note function. For example, the chairperson 12 displays the remarks of the participant 13 as an electronic sticky note on his / her electronic tablet or the like, and displays the displayed contents on the screen of the electronic tablet or the like held by each participant 13 or the screen 14. It may be displayed as an electronic sticky note at a predetermined position. In this case, the discussion support apparatus 100 may determine the correspondence between the speech of the participant 13 and the corresponding node based on the output positional relationship based on the electronic sticky note function. For example, when the chairperson 12 places (outputs) the speech of the participant 13 at an arbitrary node in the causal loop diagram on the screen or in the vicinity thereof, the discussion support apparatus 100 determines that the speech and the arbitrary Associate a node.

  Next, the speech analysis processing unit 203 analyzes the content of the speech acquired by the speech receiving unit 201. For example, the speech analysis processing unit 203 refers to the speech acquired by the speech receiving unit 201 while referring to the natural language classification dictionary 204 and the evaluation expression dictionary 205, morphological analysis, syntax analysis, semantic analysis, case grammar, emotion analysis, etc. Analysis using natural language processing technology.

  Specifically, the speech analysis processing unit 203 decomposes the speech acquired by the speech receiving unit 201 into a plurality of words. In the present embodiment, the speech analysis processing unit 203 decomposes the acquired speech into at least one morpheme (a word having a predetermined meaning).

  In addition, the speech analysis processing unit 203 acquires a viewpoint of the decomposed speech word. That is, the utterance analysis processing unit 203 classifies the viewpoint indicated by the utterance into one of a plurality of types of viewpoints. In the present embodiment, the viewpoint includes a subject viewpoint, a subject viewpoint, a time viewpoint, a place viewpoint, a reason viewpoint, and a method viewpoint. That is, the said viewpoint shall be six kinds of viewpoints of what is called 5W1H (Who, What, When, Where, Why, How).

  In addition, the speech analysis processing unit 203 acquires the evaluation information of the acquired speech. Specifically, the utterance analysis processing unit 203 analyzes the acquired evaluation information of the utterance, and stores the analyzed evaluation information in association with the acquired utterance. In the present embodiment, the evaluation information includes emotion polarity information indicating the degree of positive meaning or the degree of negative meaning of the element.

  The speech analysis processing unit 203 stores a speech analysis result in speech analysis result information 206 described later.

  Here, the natural language classification dictionary 204 and the evaluation expression dictionary 205 referred to by the speech analysis processing unit 203 will be described.

<Natural language classification dictionary 204>
FIG. 5 is a diagram illustrating an example of the natural language classification dictionary 204. As shown in the figure, the natural language classification dictionary 204 is a database storing morpheme information, and includes morpheme ID 401, morpheme 402, prototype 403, part of speech 404, synonym ID 405, broader word ID 406, narrower word ID 407, and related items. It is composed of at least one record including each item of the word ID 408. Each of these items stores, for example, a character string or a numerical value.

  The morpheme ID 401 stores an identifier for identifying the morpheme (hereinafter referred to as morpheme ID), the morpheme 402 stores the contents of the morpheme ID 401, and the prototype 403 stores the prototype of the morpheme. . A morpheme prototype is a morpheme that represents the common semantic content when the semantic content is common among multiple different types of morphemes (for example, when multiple morphemes are used). It is.

  The part of speech 404 includes the type of morpheme of the morpheme ID 401 (eg, noun, verb, adjective, adverb, conjunction, particle, auxiliary verb, exclamation, and collocation, etc. The type of the morpheme) is stored, and the synonym ID 405 has a different word (for example, different word form) from the morpheme of the morpheme ID 401 but has a similar semantic content (for example, a morpheme that can be replaced depending on the situation). An identifier (hereinafter referred to as a synonym ID) is stored, and an identifier (hereinafter referred to as a broad word ID) for identifying a morpheme of a higher concept of the morpheme ID 401 is stored in the broad word ID 406, and a narrow word ID 407 is stored. Stores an identifier for identifying a morpheme of a lower concept of the morpheme of the morpheme ID 401 (hereinafter referred to as a narrow word ID). In the word ID 408, an identifier (hereinafter referred to as a related word) of an expression (for example, a synonym, a co-occurrence word, an associative word, or an associative word) that is often handled together with the morpheme of the morpheme ID 401. , Referred to as related word ID).

  By referring to the above natural language classification dictionary 204, the speech analysis processing unit 203 can acquire the prototype of the specified morpheme, part of speech, synonym, broader word, narrower word, and related word. Although not shown, the natural language classification dictionary 204 also stores an antonym for each morpheme.

  In this embodiment, a common identifier is assigned to each synonym ID and morpheme ID, and when searching for a synonym for a morpheme of a certain morpheme ID, the record of the record storing the morpheme ID is stored. By acquiring a synonym ID and acquiring a record of a morpheme ID having the same identifier as the acquired synonym ID, information on the synonym can be acquired.

  Further, in this embodiment, a common identifier is assigned to each of the broad word ID and the morpheme ID, and when searching for a broad word for a morpheme of a certain morpheme ID, the record of the record storing the morpheme ID is stored. By acquiring the broad word ID and referring to the record of the morpheme ID having the same identifier as the acquired broad word ID, it is possible to acquire the information of the broad word.

  In this embodiment, a common identifier is assigned to each of the narrow-sense word ID and the morpheme ID. When searching for a narrow-sense word for a morpheme of a certain morpheme ID, the record of the record storing the morpheme ID is stored. By acquiring the narrow-sense word ID and referring to the record of the morpheme ID having the same identifier as the acquired narrow-sense word ID, information on the narrow-sense word can be acquired.

  In this embodiment, a common identifier is assigned to each of the related word ID and the morpheme ID. When searching for a related word for a morpheme of a certain morpheme ID, the record of the record storing the morpheme ID is stored. By acquiring the related word ID and referring to the record of the morpheme ID having the same identifier as the acquired related word ID, the related word information can be acquired.

<Evaluation expression dictionary 205>
Next, the evaluation expression dictionary 205 will be described.
FIG. 6 is a diagram illustrating an example of the evaluation expression dictionary 205. As shown in the figure, the evaluation expression dictionary 205 is a database storing morpheme evaluation information (emotion polarity) information, and includes a main word ID 501, a main word 502, a subordinate word ID 503, a subordinate word 504, and an evaluation polarity value. It is composed of at least one record including 505 items. Each of these items stores, for example, a character string or a numerical value.

  The main word ID 501 stores a morpheme ID. Note that the morpheme ID stored in the main word ID 501 is a morpheme (hereinafter referred to as a main word, for example, a noun) in which the morpheme indicated by the morpheme ID can be the subject (for example, the subject) of a sentence.

  The main word 502 stores the contents of the morpheme of the main word ID 501, and the subordinate word ID 503 corresponds to the morpheme of the main word ID 501 (hereinafter referred to as subordinate word. For example, verb, An identifier (hereinafter referred to as a dependent word ID) for identifying an adjective or an adjective verb is stored, and the dependent word of the dependent word ID 503 is stored in the dependent word 504.

  The evaluation polarity value 505 stores information (hereinafter referred to as a polarity value) that expresses the emotional polarity of a word that is a combination of the main word with the main word ID 501 and the subordinate word with the subordinate word ID 503. For example, “1” is stored if the meaning of the word is positive, “−1” is stored if the meaning of the word is negative, and “0” is stored if the meaning of the word is neutral. For example, in the word “fruit is sweet” in which the main word is “fruit” and the subordinate word is “sweet”, the meaning of the word is positive, and therefore the evaluation polarity value 505 has “1”. Stored. On the other hand, in the word “check is sweet” where the main word is “check” and the subordinate word “sweet”, the meaning of the word is negative, and therefore the evaluation polarity value 505 has “−1”. Stored. In addition, the word “umbrella” is the main word “umbrella”, the subordinate word is “yes”, and the meaning of the word is neither positive nor negative. Stores “0”.

  By referring to the evaluation expression dictionary 205 described above, the speech analysis processing unit 203 can acquire the polarity value of the designated morpheme in consideration of the relationship with the corresponding dependent word.

  The contents of the evaluation expression dictionary 205 may be input in advance by the user, or may be automatically created by the discussion support apparatus 100 based on a natural language processing technique or the like. The content of the evaluation polarity value 505 may be based on a polarity value (for example, a real number in the range of −1 or more and 1 or less) acquired from a predetermined emotion dictionary created manually or automatically.

  As shown in FIG. 3, the speech analysis processing unit 203 stores the result of speech analysis performed based on the natural language classification dictionary 204 and the evaluation expression dictionary 205 in speech analysis result information 206.

<Speech analysis result information 206>
FIG. 7 is a diagram illustrating an example of the utterance analysis result information 206. As shown in the figure, the speech analysis result information 206 includes speech ID 601, speech morpheme ID 602, speech morpheme 603, sentence evaluation polarity value 604, subject information 605, subject information 606, state information 607, time information 608, and position information 609. , And reason information 610, and includes at least one record. For example, a character string or a numerical value is stored in each of these items.

  The speech ID 601 is the same as the speech ID 301 of the speech information 202. The speech morpheme ID 602 stores the morpheme ID of the morpheme constituting the speech of the speech ID 601, and the speech morpheme 603 stores the contents of the morpheme of the speech morpheme ID 602.

  The sentence evaluation polarity value 604 stores the message polarity value of the message ID 601. For example, when the utterance of the utterance ID 601 is a positive expression, “1” is stored in the sentence evaluation polarity value 604, and when the utterance of the utterance ID 601 is negative, the sentence evaluation polarity value 604 contains When “−1” is stored and the statement of the statement ID 601 is neither a positive expression nor a negative expression, “0” is stored in the sentence evaluation polarity value 604. Note that when a real number of −1 or more and 1 or less is stored in the evaluation polarity value 505 of the evaluation expression dictionary 205, the value may be used.

  In the subject information 605, the target information 606, the state information 607, the time information 608, the position information 609, and the reason information 610, information from the viewpoint of the morpheme is stored.

  The subject information 605 stores information indicating whether or not the morpheme of the speech morpheme ID 602 represents the subject. For example, when the morpheme of the speech morpheme ID 602 represents the subject, “1” is stored in the subject information 605, and when the morpheme of the speech morpheme ID 602 does not represent the subject, “0” is stored in the subject information 605. Is stored. Note that representing the subject corresponds to, for example, a so-called 5W1H Who (who?).

  The target information 606 stores information indicating whether or not the morpheme of the speech morpheme ID 602 represents a target (for example, a target object to be operated). For example, when the morpheme of the speech morpheme ID 602 represents the target, “1” is stored in the target information 606, and when the morpheme of the speech morpheme ID 602 does not represent the target, “0” is stored in the target information 606. Is stored. It should be noted that expressing the object corresponds to, for example, what is called 5W1H What.

  The state information 607 stores information indicating whether or not the morpheme of the speech morpheme ID 602 represents a method. For example, when the morpheme of the speech morpheme ID 602 represents a method, “1” is stored in the state information 607, and when the morpheme of the speech morpheme ID 602 does not represent a method, “0” is stored in the state information 607. Is stored. Note that the expression of a method corresponds to, for example, a so-called 5W1H How.

  The time information 608 stores information indicating whether or not the morpheme of the speech morpheme ID 602 represents time. For example, when the morpheme of the speech morpheme ID 602 represents time, “1” is stored in the time information 608, and when the morpheme of the speech morpheme ID 602 does not represent time, “0” is stored in the time information 608. Is stored. Note that the expression of time corresponds to, for example, the so-called 5W1H When.

  The position information 609 stores information indicating whether or not the morpheme of the speech morpheme ID 602 represents a place. For example, when the morpheme of the message morpheme ID 602 represents a place, “1” is stored in the position information 609, and when the morpheme of the message morpheme ID 602 does not represent a place, “0” is stored in the position information 609. Is stored. In addition, representing a place corresponds to, for example, a so-called 5W1H Where (where?).

  The reason information 610 stores information indicating whether or not the morpheme of the speech morpheme ID 602 represents the reason. For example, when the morpheme of the utterance morpheme ID 602 indicates the reason, “1” is stored in the reason information 610, and when the morpheme of the utterance morpheme ID 602 does not indicate the reason, “0” is displayed in the reason information 610. Is stored. Note that expressing the reason means that, for example, it corresponds to the so-called 5W1H Why (why?).

  By referring to the above-described speech analysis result information 206, the speech analysis processing unit 203, for the speech of the designated participant 13, each morpheme that constitutes the speech, the polarity value of the speech, and each of the speech that constitutes the speech A morphological perspective can be obtained.

  As illustrated in FIG. 3, the idea creation support processing unit 207 controls the speech analysis processing unit 203, the question sentence generation processing unit 210, and the display processing unit 211.

<Question sentence generation processing unit 210>
Next, the question sentence generation processing unit 210 (question generation unit) specifies the element corresponding to the acquired utterance by the utterance reception unit 201 based on the decomposed word by the utterance analysis processing unit 203, and the specified It is determined whether or not there is another element having a causal relationship with the element, and when it is determined that there is the other element, the other element is specified as an object of the discussion, and the specified object is determined. Based on this, a question about the object is generated.

  When the question generation unit determines that there is no other element, the question generation unit selects at least one of the viewpoints of the plurality of viewpoints of the specified element, and the viewpoint of the selected element To generate a question about the agenda.

  In addition, the question generation unit determines whether there are a plurality of elements that cause the identified element as the other elements, and determines that there are a plurality of elements that cause the plurality of elements. When a question about the agenda is generated based on the selected element among the elements of the above and it is determined that there are not a plurality of the causative elements, the question about the agenda is based on the element that is the result of the identified element. Generate a question.

  Further, the question generation unit acquires the evaluation information of the identified element by the speech reception unit 201, and determines whether or not the acquired evaluation information of the speech and the acquired evaluation information of the element correspond to each other When it is determined that the two do not correspond to each other, the word having evaluation information different from the evaluation information of the element is extracted from the decomposed words, the word different from the extracted word, A question about the agenda is generated based on other factors.

  The question generation unit determines whether the element corresponding to the statement is an initial element in the discussion, and only when the element is determined not to be an initial element, It is determined whether or not the evaluation information corresponds to the evaluation information of the identified element.

  Further, the question generation unit determines whether or not the other element having a causal relationship with the element corresponding to the statement is important based on information indicating the importance in the discussion set for each of the elements. If the other element is determined to be important, a question about the agenda is generated based on the other element, and if it is determined that the other element is not important, A question about the agenda is generated based on the decomposed words.

  In addition, the question generation unit determines whether or not the identified element is an element within the authority of the person who is performing the discussion, and only when the element is determined to be an element within the authority Generate questions about the agenda.

  The question generator stores the elements used for generating the question and the viewpoints of the elements.

<Agenda Element Information 208>
Here, the agenda element information 208 will be described.
FIG. 8 is a diagram illustrating an example of the agenda element information 208. As shown in the figure, the agenda element information 208 is a database storing information on the causal relationship between elements (nodes), and includes a node ID 801, a node name 802, a node morpheme ID 803, a node morpheme 804, and a viewpoint attribute 805. , Node evaluation polarity 806, control availability 807, cause node ID 808, result node ID 809, time series label 810, loop ID 811, and important node label 812, each of which includes at least one record. Each of these items stores, for example, a character string or a numerical value.

  The node ID 801 stores the node ID, the node name 802 stores the name of the node ID 801, the node morpheme ID 803 stores the morpheme ID of the morpheme constituting the node of the node ID 801, and the node morpheme The morpheme content of the node morpheme ID 803 is stored in 804, and the viewpoint of the morpheme of the node morpheme 804 is stored in the viewpoint attribute 805. Note that if there is no viewpoint of the morpheme of the node morpheme 804, nothing is stored in the viewpoint attribute 805.

  The node evaluation polarity 806 stores the polarity value of the node having the node ID 801. For example, “1” is stored if the event represented by the node with the node ID 801 is preferable, and “−1” is stored if the event is not preferable. For example, if the event is “amount of harvest”, “1” is stored, and if the event is “the number of traffic accidents”, “−1” is stored. Note that the content of the node evaluation polarity 806 may be input (set) by the user in advance, or the discussion support apparatus 100 provides data (for example, news or the like) related to the event indicated by the node with the node ID 801. ) From other devices or networks, the acquired data is analyzed (for example, positive / negative evaluation for the event is analyzed), and the analysis result is converted into a polar value and stored in the node evaluation polarity 806. You may do it.

  The control permission / prohibition 807 stores information indicating whether the participant 13 has authority for the event indicated by the node with the node ID 801. For example, if the event indicated by the node with the node ID 801 is an event within the authority of the conference participant 13 (for example, within the business authority), “1” is stored, and the event indicated by the node with the node ID 801 is the conference participant 13. If the event is out of authority (for example, out of business authority), “−1” is stored. For example, when the participant 13 is engaged in the production of goods, if the event indicated by the node with the node ID 801 is “production amount”, the event is within the authority of the participant 13, so that control is possible 807. “1” is stored, but when the event indicated by the node with the node ID 801 is “declining birthrate and aging”, “1” is stored in the control permission / inhibition 807 because the event is outside the authority of the participant 13. Is done. Note that the contents of the control permission / inhibition 807 may be set in advance, or may be set at any time during the conference based on the speech of the participant 13. Note that the participant 13 has authority is, for example, that the participant 13 has authority that can directly participate in the promotion or suppression of the event indicated by the node. For example, if the participant 13 belongs to a private company, but the event indicated by the node is a government measure, a natural phenomenon, or the like, the participant 13 has no authority.

  The cause node ID 808 stores the node ID of the node that is the cause of the node with the node ID 801 (that is, the node that represents the cause in the causal relationship, hereinafter referred to as the cause node). If the cause node does not exist, “0” is stored.

  The result node ID 809 stores a node ID of a node that is a result of the node with the node ID 801 (that is, a node that represents a result in a causal relationship, hereinafter referred to as a result node). If there is no result node, “0” is stored.

  The time series label 810 stores time series information assumed by the agenda indicated by the node with the node ID 801. Specifically, the time-series label 810 stores information when a plurality of nodes including the node of the node ID 801 form a causal relationship, and the plurality of nodes that circulate (hereinafter referred to as a circulation node). Time-series information (hereinafter referred to as node time-series) in a series of discussions indicated in FIG. For example, the time series label 810 stores information such as “initial”, “transition period”, “mature period”, “present”, “5 years later”, “10 years later”, and the like. The time series label 810 stores values such as “1” for “initial”, “2” for “transition”, “3” for “mature”, etc. Good. Thereby, the participant 13 can grasp which stage of the element is being discussed in time series.

  The loop ID 811 stores the identifier of the circulating node to which the node with the node ID 801 belongs (hereinafter referred to as a loop ID).

  The important node label 812 stores information indicating the importance of the node having the node ID 801 (hereinafter referred to as an important node label). Specifically, when the event indicated by the node with the node ID 801 is an event to be deeply clarified in the discussion, “1” is stored in the important node label 812, and the event indicated by the node with the node ID 801 is the discussion. In the case of an event that should not be deeply cultivated, “0” is stored in the important node label 812. Note that the contents of the important node label 812 may be set by a user (moderator 12 or the like), respectively, or may be set based on a predetermined rule. For example, it is important when the node with the node ID 801 is a node constituting two different circulation nodes, or when the node time series of the node with the node ID 801 is a node time series different from the cause node or the result node. “1” may be set in the node label 812.

  By referring to the agenda element information 208 described above, the speech analysis processing unit 203 specifies the morpheme constituting the designated node, the viewpoint of each morpheme constituting the designated node, the polarity value of the designated node, The authority of the specified node, the cause node or result node of the specified node, the time series of the specified node, the importance of the specified node, and the circular node constituted by the specified node can be acquired. .

  Here, a causal loop diagram output to the discussion support system 10 based on the agenda element information 208 will be described.

<Causal loop diagram 700>
FIG. 9 is a diagram illustrating an example of a causal loop diagram. As shown in the figure, the causal loop diagram 700 includes a plurality of nodes 701 (701 (1) to (7)) indicating a node and a link 702 connecting the two nodes 701. The

  A link 702 indicates that there is a causal relationship between the two nodes 701. In the present embodiment, the link 702 is indicated by an arrow. That is, the node 701 connected to the end point of the arrow of the link 702 is a result node, and the node 701 connected to the start point of the arrow indicated by the link 702 is a cause node. Each link 702 is represented by a solid or broken line arrow. When the arrow of the link 702 is a solid line, the relationship between the two nodes 701 connected to the link 702 is positively correlated. When the arrow of the link 702 is a dotted line, 2 is connected to the link 702. The relationship between the two nodes 701 is negatively correlated.

  In addition, the causal loop diagram 700 indicates that when a causal relationship is circulating among a plurality of nodes, a loop 703 indicates that. In the example shown in the figure, the node 701 (2) (node A) is the cause node of the node 701 (3) (node B), and the node 701 (3) (node B) is the node 701 (4) (node C). Node 701 (4) (node C) is the cause node of node 701 (5) (node D), and node 701 (5) (node D) is node 701 (2) (node A). Is the cause node.

  Corresponding to these, in the agenda element information 208 of FIG. 8, the cause node of the node 701 (5) (node D) is the node 701 (4) (node C) and the node 701 (5) (node D). Since the result node is the node 701 (2) (node A), “1” is stored in the loop ID 811. In addition, since the cause node of the node 701 (5) (node D) is the node 701 (7) (node F) and the result node is the node 701 (2) (node A), the loop ID 811 has “2”. Is stored.

  The causal loop diagram 700 may be output to, for example, the output device 105 of the discussion support apparatus 100 or may be output to the screen 14.

  For each element shown in the causal loop diagram 700 or the like, information related to a case related to each element (hereinafter referred to as a related case) is stored in the agenda related case information 209.

<Agenda related case information 209>
FIG. 10 is a diagram illustrating an example of the agenda related case information 209. As shown in the figure, the agenda related case information 209 is a database storing related agendas, and includes a node ID 901, a node configuration case ID 902, a node configuration case name 903, a node configuration case morpheme ID 904, a node configuration case morpheme 905, and It consists of at least one record including each item of the viewpoint attribute 906. Each of these items stores, for example, a character string or a numerical value.

  The node ID 901 stores the same information as the node ID 801, the node configuration case ID 902 stores the identifier of the related case of the node with the node ID 901 (hereinafter referred to as the related case ID), and the node configuration case name 903 , The contents of the related case of the node configuration case ID 902 are stored. For example, when the node ID 901 stores information on the phrase “price of domestic vegetables”, the node configuration example name 903 includes “production weather” or “ Stores information on the phrase “Import from overseas”.

  The node configuration case morpheme ID 904 stores an identifier (hereinafter referred to as a related case morpheme ID) of a morpheme (hereinafter referred to as a related case morpheme) constituting a related case of the node configuration case ID 902.

  The node configuration case morpheme 905 stores the contents of the related case morpheme, and the viewpoint attribute 906 stores information on the viewpoint of the related case morpheme of the node configuration case morpheme 905. If there is no corresponding viewpoint, nothing is stored in the node configuration case morpheme 905.

  From the agenda related case information 209 described above, the contents of the related case of the designated node, the morphemes constituting the related case of the node, and the viewpoints of the morphemes constituting the related case of the node are acquired.

  Returning to FIG. 3, the display processing unit 211 includes a question output unit 212 and a viewpoint output unit 213.

  The question output unit 212 outputs the question generated by the question generation processing unit.

  The viewpoint output unit 213 outputs the viewpoint of the element used for generating the question and the viewpoint of the acquired word.

  The functions of the discussion support device 100 described above are the programs stored in the storage device 102 or the secondary storage device 103 by the hardware of the discussion support device 100 or by the central processing unit 101 of the discussion support device 100. This is realized by reading and executing. The program may be recorded on a recording medium such as an IC card, an SD card, or a DVD.

<< Processing executed by the discussion support system 10 >>
Next, processing performed in the discussion support system 10 will be described.

<Overall processing>
FIG. 11 is a flowchart illustrating an outline of processing performed by the discussion support apparatus 100. Note that this process (hereinafter referred to as the overall process) is performed, for example, when the discussion support apparatus 100 is activated or when a predetermined timing set in the discussion support apparatus 100 arrives (for example, a predetermined time interval). In).

  As shown in the figure, first, the idea creation support processing unit 207 selects one node (S1001). Specifically, for example, the idea creation support processing unit 207 receives one input of the node ID stored in the agenda element information 208 from the moderator 12 or the node stored in the agenda element information 208 Among the IDs, a node ID that has not been selected so far and has the largest value is selected. In the following, the node with the selected node ID is referred to as a selected node.

  The speech receiving unit 201 waits for the participant 13 (the speaker 11 or the moderator 12) to acquire a speech (S1002: NO).

  When the utterance is acquired (S1002: YES), the utterance analysis processing unit 203 analyzes the utterance acquired in S1002 (S1003). Specifically, the speech analysis processing unit 203 uses the natural language processing techniques of morphological analysis, syntax analysis, semantic analysis, case grammar, and emotion analysis for the acquired speech, and the natural language classification dictionary 204, the evaluation expression dictionary Based on 205, the morpheme is extracted from the utterance of the participant or the presenter, a viewpoint for the extracted morpheme is given, polarity is given to the utterance, and these results are stored in the utterance analysis result information 206.

  The question sentence generation processing unit 210 generates a question sentence based on the content of the utterance acquired in S1002, the morpheme of the selected node, the importance of the selected node, the circular node to which the selected node belongs (hereinafter, the question sentence generation process) (S1004). Details of the question sentence generation process will be described later.

  Next, the display processing unit 211 outputs the question sentence created in S1004 (S1005).

  The speech receiving unit 201 determines whether or not the participant 13 has made a speech in response to the question message created in S1005 (S1006). Specifically, the message receiving unit 201 monitors whether or not a new message has been acquired from the participant 13 for a predetermined time.

  If a new message is acquired from the participant 13 (S1006: YES), the process returns to S1003. If a new message is not acquired from the participant 13 (S1006: NO), the process proceeds to S1007. move on.

  In step S1007, the idea creation support processing unit 207 determines whether or not to end the discussion regarding the selected node selected in step S1001. Specifically, for example, the idea creation support processing unit 207 receives an input from the participant 13 (for example, the moderator 12) whether to end the discussion on the selected node, and determines the content of the received input. .

  When the discussion about the selected node is ended (S1007: YES), the process proceeds to S1008. When the discussion about the selected node is not ended (S1007: NO), the process returns to S1004.

  In step S1008, the idea creation support processing unit 207 determines whether there is a node that has not generated a question sentence. If there is a node that has not generated a question sentence (S1008: YES), the idea creation support processing unit 207 sets the node as a selected node and repeats the process of S1001. On the other hand, if there is no node that has not generated a question sentence (S1008: NO), the entire process ends (S1009).

<Question sentence generation process>
Details of the question sentence generation process will be described.
FIGS. 12 and 13 are flowcharts for explaining the details of the question sentence generation process (for the sake of space, the drawing is divided into two parts).

  As illustrated in FIG. 12, the question sentence generation processing unit 210 determines whether or not the selected node is an initial element of discussion (S1101). Specifically, the question sentence generation processing unit 210 refers to the agenda element information 208 and determines whether or not the time-series label of the selected node is “1”.

  When the selected node is an initial element of discussion (S1101: YES), the process proceeds to S1102, and when the selected node is not the initial element of discussion (S1101: NO), the process is S1112 (FIG. 13). Proceed to

  In S1102, the question sentence generation processing unit 210 determines whether the selected node is important in the discussion. Specifically, the question sentence generation processing unit 210 refers to the agenda element information 208 and determines whether or not the important node label of the record of the selected node is “1”.

  If the selected node is important in the discussion (S1102: YES), the process proceeds to S1104. If the selected node is not important in the discussion (S1102: NO), the process proceeds to S1103.

  In S1104, the question sentence generation processing unit 210 determines whether a question sentence has been generated based on the morphemes of all viewpoints of the selected node. When the question sentence is generated based on the morphemes of all the viewpoints of the selected node (S1104: YES), the question sentence generation process ends (S1130), and the question sentences based on the morphemes of all the viewpoints of the selected node are generated. If not (S1104: NO), the process proceeds to S1109.

  In S1109, the question sentence generation processing unit 210 determines whether the cause node of the selected node or the result node of the selected node (hereinafter referred to as the preceding and following nodes) is important. Specifically, for example, the question sentence generation processing unit 210 acquires the cause node of the selected node and the result node of the selected node by referring to the agenda element information 208, and then the agenda element information 208 again. By referencing, it is determined whether or not the important node label of the cause node or the important node label of the result node is “1”.

  If the cause node of the selected node or the result node of the selected node is important, the process proceeds to S1110 (S1109: YES), and the cause node of the selected node and the result node of the selected node are not important. (S1109: NO), the process proceeds to S1111.

  In S1110, the question sentence generation processing unit 210 generates a question sentence on the agenda based on the important cause node or result node (hereinafter, referred to as an important node before and after).

  Specifically, the question sentence generation processing unit 210 refers to the agenda element information 208 and the agenda-related case information 209, so that the viewpoints that have not been used in the selected node among the viewpoints of the question sentences generated so far. select. Next, the question sentence generation processing unit 210 acquires the morphemes of the preceding and following important nodes and the related cases of the preceding and following important nodes (hereinafter referred to as unused viewpoint morphemes in this paragraph) having the selected viewpoint. Then, the question sentence generation processing unit 210 generates a question sentence for asking about unused viewpoint morphemes. Thus, the question sentence generation process ends (S1130).

  In S1111, the question sentence generation processing unit 210 selects a viewpoint of a selected node that has not been used for generating a question sentence from among a plurality of viewpoints of the selected node, and selects each of the selected nodes having the selected viewpoint. Based on the morpheme, generate a question sentence about the agenda.

  Specifically, the question sentence generation processing unit 210 refers to the agenda element information 208 to acquire the viewpoints of the selected nodes that have not been used in the question sentence generated so far among all the viewpoints of the selected nodes. . And the question sentence production | generation process part 210 acquires each morpheme of the selection node which has the acquired viewpoint, and produces | generates the question sentence which asks about each acquired morpheme. Thus, the question sentence generation process ends (S1130).

  On the other hand, in S1103, the question sentence generation processing unit 210 determines whether or not the selected node is an element within the authority of the participant 13. Specifically, the question sentence generation processing unit 210 refers to the agenda element information 208 to determine whether or not the selected node is an element within the authority of the participant 13.

  When the selected node is an element within the authority of the participant 13 (S1103: YES), the process proceeds to S1105, and when the selected node is not an element within the authority of the participant 13 (S1103: NO). The question sentence generation process ends.

  In S1105, the question sentence generation processing unit 210 refers to the utterance information 202 and the utterance analysis result information 206, so that the utterance can be made from the viewpoint of the utterance of the participant 13 so far among the multiple viewpoints of the selected node. Identify the most common viewpoints. And the question sentence production | generation process part 210 judges whether the question sentence based on the selected node which has the specified viewpoint was produced | generated so far. The “perspective with a lot of remarks” may be, for example, a viewpoint with the most remarks so far among the viewpoints of remarks by the participant 13, or a predetermined viewpoint among the remarks with respect to the remarks of the participant 13. All viewpoints in which more than the number of statements have been made may be used.

  If the question sentence has been generated (S1105: YES), the question sentence generation process ends (S1130). If the question sentence has not been generated (S1105: NO), the process proceeds to S1106.

  In S1106, the question sentence generation processing unit 210 determines whether the preceding and following nodes are important as in S1109. If the preceding and following nodes are important (S1106: YES), the process proceeds to S1107. If the preceding and following nodes are not important (S1106: NO), the process proceeds to S1108.

  In S1107, the question sentence generation processing unit 210 generates a question sentence about the agenda based on the preceding and following nodes.

  Specifically, for example, the question sentence generation processing unit 210 first acquires the viewpoint identified in S1105 that has a lot of remarks. In addition, the question sentence generation processing unit 210 refers to the agenda element information 208 and the agenda related case information 209, so that each morpheme of the front and rear important nodes and each related case of the front and rear important nodes (hereinafter referred to as front and rear important in this paragraph). Morpheme). Next, the question sentence generation processing unit 210 acquires a morpheme having the acquired many viewpoints among the front and rear important morphemes. Then, the question sentence generation processing unit 210 generates a question sentence for asking about each acquired morpheme. Thus, the question sentence generation process ends (S1130).

  In S1108, the question message generation processing unit 210 generates a question message about the agenda based on the remarks of the participant 13.

  Specifically, for example, the question sentence generation processing unit 210 first acquires the viewpoint identified in S1105. In addition, the question sentence generation processing unit 210 acquires the morpheme of the message acquired in S1003. Next, the question sentence generation processing unit 210 refers to the natural language classification dictionary 203 to acquire the related words of each acquired morpheme. And the question sentence production | generation part 210 produces | generates the question sentence which asks about the acquired related word. Thus, the question sentence generation process ends (S1130).

  As shown in FIG. 13, in S1112, the question sentence generation processing unit 210 determines whether the selected node is important in the discussion, as in S1102. If the selected node is important in the discussion (S1112: YES), the process proceeds to S1113. If the selected node is not important in the discussion (S1112: NO), the process proceeds to S1114.

  In S1113, the question sentence generation processing unit 210 determines whether a question sentence has been generated based on the morphemes of all viewpoints of the selected node, as in S1104. When the question text is generated based on the morphemes of all the viewpoints of the selected node (S1113: YES), the process proceeds to S1116, and the question text is not generated based on the morphemes of all the viewpoints of the selected node. (S1113: NO), the process proceeds to S1115.

  In S1116, the question sentence generation processing unit 210 determines whether or not the evaluation information of the utterance of the participant 13 corresponds to the evaluation information of the selected node. Specifically, for example, the question sentence generation processing unit 210 acquires the polarity value of the statement acquired in S1002 by referring to the statement analysis result information 206, and the polarity of the selected node by referring to the agenda element information 208 A value is acquired, and it is determined whether or not the two acquired polarity values match (or whether or not the two polarity values match).

  When both evaluation information does not correspond (S1116: NO), a process progresses to S1119, and when both evaluation information respond | corresponds (S1116: YES), a question sentence production | generation process is complete | finished (S1130).

  In S1119, the question sentence generation processing unit 210 determines whether there are a plurality of cause nodes of the selected node as nodes having a causal relationship with the selected node. Specifically, the question sentence generation processing unit 210 refers to the agenda element information 208 to determine whether there are a plurality of cause nodes of the selected node.

  If there are a plurality of cause nodes of the selected node (S1119: YES), the process proceeds to S1121, and if there are not a plurality of cause nodes of the selected node (one or no cause node of the selected node) (S1119: NO), The process proceeds to S1120.

  In S1121, the question sentence generation processing unit 210 selects one of a plurality of cause nodes, and generates a question about the agenda based on the selected cause nodes.

  Specifically, for example, the question sentence generation processing unit 210 first obtains the viewpoint of each morpheme of the utterance of the selected node by referring to the utterance information 202 and the utterance analysis result information 206. Next, the question sentence generation processing unit 210 has the same viewpoint as each of the acquired viewpoints, and the cause node that has not been selected so far among the plurality of cause nodes (or the cause node that has been selected a smaller number of times) ) And each related morpheme of the causal node (hereinafter referred to as the same viewpoint morpheme in this paragraph) are acquired by referring to the agenda element information 208 and the agenda related case information 209. Next, the question sentence generation processing unit 210 identifies a morpheme having a polarity value that is different (or has a different value) from the polarity value of the selected node among the morphemes of the speech acquired in S1003, and is an antonym of the identified morpheme ( Get a synonym that reverses polarity). And the question sentence production | generation process part 210 produces | generates the question which asks about the acquired same viewpoint morpheme and the acquired synonym. Thus, the question sentence generation process ends (S1130).

  On the other hand, in S1120, the question sentence generation processing unit 210 generates a question about the agenda based on the result node of the selected node.

  Specifically, for example, the question sentence generation processing unit 210 first acquires the viewpoint of the morpheme of the remark acquired in S1003. In addition, the question sentence generation processing unit 210 converts each morpheme of the result node of the selected node and each morpheme of the related case of the result node (hereinafter referred to as result morpheme in this paragraph) into the agenda element information 208 and the agenda related case information 209. Get by referring to. And the question sentence generation process part 210 acquires each morpheme which has the said acquired viewpoint among result morphemes. In addition, the question sentence generation processing unit 210 identifies a morpheme having a polarity value that is different (or different in value) from the polarity value of the selected node among the morphemes of the remark acquired in S1003, and is a synonym ( Get a synonym that reverses polarity). And the question sentence production | generation process part 210 produces | generates the question which asks about each acquired morpheme in the acquired synonym and the acquired result morpheme. Thus, the question sentence generation process ends (S1130).

  In S1114, the question sentence generation processing unit 210 determines whether the selected node is within the authority of the participant 13 as in S1103.

  If the selected node is within the authority of the participant 13 (S1114: YES), the process proceeds to S1122, and if the selected node is not within the authority of the participant 13 (S1114: NO), a question text is generated. The process ends (S1130).

  In S1122, the question sentence generation processing unit 210 refers to the utterance information 202 and the utterance analysis result information 206, as in S1105, so that the selected node has a plurality of viewpoints. From the viewpoint of the speech of the participant 13 so far, the viewpoint that has a lot of speech is specified. And the question sentence production | generation process part 210 judges whether the question sentence based on the selected node which has the specified viewpoint was produced | generated so far.

  If the question sentence has been generated (S1122: YES), the process proceeds to S1124. If the question sentence has not been generated (S1122: NO), the process proceeds to S1123.

  In S <b> 1123, the question sentence generation processing unit 210 generates a question sentence about the agenda based on the remarks of the participant 13.

  Specifically, for example, the question sentence generation processing unit 210 acquires the viewpoint that has been specified in S1122 and has many remarks. Next, the question sentence generation processing unit 210 acquires the morpheme of the speech acquired in S1003, and acquires the morpheme having the above-described viewpoint among the acquired morphemes. Then, the question sentence generation processing unit 210 generates a question sentence for asking about each acquired morpheme. Thus, the question sentence generation process ends (S1130).

  On the other hand, in S1124, the question sentence generation processing unit 210 determines whether or not the evaluation information of the utterance of the participant 13 and the evaluation information of the selected node correspond to each other as in S1116.

  When both evaluation information corresponds (S1124: YES), a question sentence production | generation process is complete | finished (S1130), and when both evaluation information does not respond | correspond (S1124: NO), a process progresses to S1125.

  In S1125, the question sentence generation processing unit 210 determines whether the cause node of the selected node or the result node (previous node) of the selected node is important as in S1106. If the cause node of the selected node or the result node of the selected node is important (S1125: YES), the process proceeds to S1127, and if the cause node of the selected node and the result node of the selected node are not important ( S1125: NO), the process proceeds to S1126.

  In S1126, the question sentence generation processing unit 210 extracts a morpheme having evaluation information different from the evaluation information that the selected node has from the morpheme of the speech of the participant 13 associated with the selected node, and the extracted morpheme Generate questions about agendas based on different morphemes.

  Specifically, for example, first, the question sentence generation processing unit 210 acquires the viewpoint of each morpheme of the utterance acquired in S1003. Next, the question sentence generation processing unit 210 converts each morpheme of the preceding and following nodes and related morphemes of the preceding and following nodes (hereinafter referred to as the same viewpoint morpheme in this paragraph) having the same viewpoint as each acquired viewpoint to the agenda element information. 208 and the agenda related case information 209 are referred to. In addition, the question sentence generation processing unit 210 acquires a morpheme having a polarity value that is different in polarity (or a value that is different) from the polarity value of the selected node among the morphemes of the message acquired in S1003. An antonym that reverses the polarity) is acquired by referring to the natural language classification dictionary 204. And the question sentence production | generation part 210 produces | generates the question sentence which asks about the acquired same viewpoint morpheme and the acquired synonym. Thus, the question sentence generation process ends (S1130).

  In S1127, the question sentence generation processing unit 210 generates a question sentence based on the selected node (S1127).

  Specifically, for example, the question sentence generation processing unit 210 first acquires a morpheme having a polarity value that is different (or different in value) from the polarity value of the selected node, among the morphemes of the message acquired in S1003. The acquired morpheme antonym (an antonym that reverses the polarity) is acquired by referring to the evaluation expression dictionary 205. Next, the question sentence generation processing unit 210 acquires a morpheme having a viewpoint with many utterances specified in S1122, and acquires related words of the acquired morpheme by referring to the natural language classification dictionary 204. And the question sentence production | generation part 210 produces | generates the question sentence which asks about the acquired synonym and the acquired related word. Thus, the question sentence generation process ends (S1130).

  In S1115, similarly to S1106, the question sentence generation processing unit 210 determines whether the cause node of the selected node or the result node of the selected node is important. If the cause node of the selected node or the result node of the selected node is important, the process proceeds to S1118. If neither the cause node of the selected node or the result node of the selected node is important, the process proceeds to S1117. Proceed to

  In S1118, the question message generation processing unit 210 generates a question message asking about the agenda based on the preceding and following important nodes, as in S1107. Thus, the question sentence generation process ends (S1130).

  On the other hand, in S <b> 1117, the question sentence generation processing unit 210 generates a question sentence about the agenda based on the remarks of the participant 13 as in S <b> 1108. Thus, the question sentence generation process ends (S1130).

  As described above, according to the discussion support system 10 of the present embodiment, the content of the speech uttered by the speaker is acquired and decomposed into a plurality of words, and the elements corresponding to the speech are determined based on the decomposed words. Identify and determine whether there are other elements that have a causal relationship with the identified element. If it is determined that there are other elements, generate and generate questions on the agenda based on the other elements. Therefore, it is possible to present a discussion of a new development to the discussion participant 13. Thereby, each participant 13 of discussion can create a new idea effectively based on the speech situation of the participant 13. For example, according to the discussion support system 10 of the present embodiment, the participant 13 can make a discussion based on a combination of creative keywords necessary for the creation of a new idea or an idea from a viewpoint different from the conventional one. it can.

  In addition, according to the discussion support system 10 of the present embodiment, when there are no other elements having a causal relationship, a question on the agenda is asked based on an element corresponding to a statement having viewpoints selected from a plurality of viewpoints. If there is no topic that has a causal relationship with the statement and it is not appropriate to discuss widely, the participant 13 can discuss the content of the statement in detail based on each viewpoint. Deep discussion can be done.

  Further, according to the discussion support system 10 of the present embodiment, when there are a plurality of causal elements as other elements having a causal relationship, a question about the agenda is based on the selected element among the plurality of elements. If there are no multiple elements that cause the problem, a question about the agenda is generated based on the resulting element. Discussion can be directed to that one topic (the topic that caused the speech), otherwise the discussion participant can be directed to the topic resulting in the speech. Thereby, the consciousness of the participant 13 in the discussion can be guided to an appropriate peripheral topic, and the participant 13 can effectively discuss the topic related to the agenda. Thereby, for example, a question sentence using a viewpoint that has not yet been discussed can be presented to the presenter 12, thereby making it possible to realize a discussion without omissions or omissions.

  In addition, according to the discussion support system 10 of the present embodiment, when the comment evaluation information and the element evaluation information do not correspond to each other, the evaluation information different from the element evaluation information is included in the word of the comment. Since a word is extracted and a question about the agenda is generated based on the word different from the extracted word and the other elements, if there is a discrepancy between the evaluation of the node and the evaluation of the participant 13, the node By presenting a question sentence that uses the content assumed by the MC (for example, the content assumed by the moderator 12), it is possible to focus discussions that incorporate a different viewpoint from the content of the speech of the participant 13 so far it can.

  Further, according to the discussion support system 10 of the present embodiment, it is determined whether or not the element corresponding to the statement is an initial element in the discussion, and the evaluation of the statement is performed only when it is determined that the element is not the initial element. Since it is determined whether or not the information and the element evaluation information correspond to each other, for example, the speech of the participant 13 is given priority in the initial stage of the discussion, and the speech of the participant 13 in the discussion is advanced in the stage where the discussion is advanced. Discussion can be made by adding the viewpoint of the content assumed by the node (for example, the content assumed by the moderator 12) to the content. Thus, the discussion can be flexibly guided in an appropriate direction according to the degree of progress of the discussion of the participant 13.

  Further, according to the discussion support system 10 of the present embodiment, when other elements having a causal relationship with an element corresponding to a statement are important, a question on the agenda is generated based on the other elements, When the elements of are not important, questions about the agenda are generated based on the words of the remarks, so that the participants 13 discuss the topics to be discussed according to the importance of the topic indicated by each node. Can be discussed.

  Further, according to the discussion support system 10 of the present embodiment, the question about the agenda is generated only when the element is an element within the authority of the speaker. You can use your own authority to solve the problem. Thereby, the participant 13 can perform a practical discussion.

Finally, the screen output by the discussion support system 10 will be described.
<Discussion progress screen>
FIG. 14 is an example of a screen (hereinafter referred to as a discussion progress screen) that the viewpoint output unit 213 outputs to the screen 14. As shown in the figure, the discussion screen 1201 is a diagram showing the degree of progress of the discussion. The causal loop display unit 1202, the attention node viewpoint display unit 1203, the important node viewpoint display unit 1204, and the question sentence presentation management unit 1205. Each output item. The question sentence presentation management unit 1205 includes output items of a question sentence display unit 1206 and a question sentence display result operation unit 1207.

  A causal loop diagram is output to the causal loop display unit 1202. In this case, the causal loop display unit 1202 may output a node having a causal relationship with the selected node. Further, only the selected node may be output with emphasis (for example, display with a thick frame or enlarged display), or the important node may be output with emphasis (for example, display with a double line). And enlarged view). Thus, the participant 13 can easily determine the node currently being discussed from the nodes set as the agenda.

  The attention node viewpoint display unit 1203 outputs each morpheme of the speech uttered by the participant 13 and the viewpoint of the morpheme. Thereby, the participant 13 can confirm the viewpoint (5W1H) of the speech that has been said a lot about the selected node, and the viewpoint that has not yet been said.

  The important node viewpoint display unit 1204 outputs each morpheme of the output question text and the viewpoint of the morpheme. Thereby, the participant 13 can confirm the viewpoint that has been discussed a lot and the viewpoint that has not been discussed yet.

  A question sentence generated with respect to the selected node is output to the question sentence presentation management unit 1205. The output question text may be a sentence such as “What is XX (subject) of □□□ (location)?” Or “□□□ (location), XX (subject)”. It may be a combination of words.

  The question sentence display result operation unit 1207 includes input fields for a display 1208, a correction 1209, and a cancel 1210.

  When the display 1208 is pressed, the question text output to the question text display unit 1206 is output to another information processing apparatus (for example, the information processing apparatus included in each participant 13).

  When the correction 1209 is pressed, an input (correction) of the question text output to the question text display unit 1206 is accepted.

  When cancel 1210 is pressed, the question text output to the question text display unit 1206 is hidden, or another question text is output.

  Note that the display 1208, the correction 1209, or the cancellation 1210 may be output only to the information processing apparatus included in the moderator 12.

  As described above, the viewpoint output unit 213 outputs the viewpoints of the elements used for generating the question and the viewpoints of the morpheme of the speech, so that the participant 13 can speak or discuss any viewpoint. You can easily know whether you are.

  The above description of the embodiments is intended to facilitate understanding of the present invention and is not intended to limit the present invention. The present invention can be changed and improved without departing from the gist thereof, and the present invention includes equivalents thereof.

DESCRIPTION OF SYMBOLS 10 Discussion support system, 13 Participants, 100 Discussion support apparatus, 201 Speech receiving part, 203 Speech analysis processing part, 210 Question sentence generation processing part, 213 Viewpoint output part

Claims (15)

A discussion support system that includes a processor and a memory and supports discussion on an agenda comprising multiple elements,
A speech acquisition unit that acquires the content of a speech made by a speaker with respect to the agenda;
A speech analysis processing unit that decomposes the acquired speech into a plurality of words;
Based on the decomposed word, the element corresponding to the acquired utterance is specified, it is determined whether there is another element having a causal relationship with the specified element, and it is determined that there is the other element. If so, a question generator for generating a question about the agenda based on the other elements;
A question output unit for outputting the generated question;
A discussion support system. If the question generation unit determines that there is no other element, the question generation unit selects at least one of the viewpoints of the plurality of viewpoints of the specified element, and based on the viewpoint of the selected element Generate questions about the agenda,
The discussion support system according to claim 1. The question generation unit determines whether there are a plurality of elements that cause the identified element as the other elements, and determines that there are a plurality of elements that cause the plurality of elements. A question on the agenda is generated based on the selected element, and when it is determined that there are not a plurality of causative elements, a question on the agenda is generated based on the element as a result of the identified element. Generate,
The discussion support system according to claim 1. The speech analysis processing unit acquires evaluation information of the acquired speech,
The question generation unit acquires evaluation information of the identified element, determines whether or not the acquired evaluation information of the utterance corresponds to the acquired evaluation information of the element, and both correspond to each other If it is determined that there is not, extract the word having evaluation information different from the evaluation information of the element among the decomposed words, and based on the word different from the extracted word and the other elements, Generating questions about the agenda;
The discussion support system according to claim 1. The question generation unit determines whether or not the element corresponding to the utterance is an initial element in the discussion, and the evaluation information of the acquired utterance is determined only when the element is not an initial element. And whether the specified element evaluation information corresponds to,
The discussion support system according to claim 4. The question generation unit determines whether or not the other element having a causal relationship with the element corresponding to the statement is important based on information indicating the importance in the discussion set for each of the elements. If it is determined that the other element is important, a question about the agenda is generated based on the other element, and if it is determined that the other element is not important, the decomposition is performed. Generate questions about the agenda based on the words
The discussion support system according to claim 1. The question generation unit determines whether or not the identified element is an element within the authority of the person who is performing the discussion, and only when the element is determined to be an element within the authority. Generate questions about the agenda,
The discussion support system according to claim 1. The speech analysis processing unit acquires a viewpoint of the decomposed word,
The discussion support system further includes a viewpoint output unit that outputs a viewpoint of the element used for generating the question and a viewpoint of the acquired word.
The discussion support system according to claim 2. If the question generation unit determines that there is no other element, the question generation unit selects at least one of the viewpoints of the plurality of viewpoints of the specified element, and based on the viewpoint of the selected element Generate questions about the agenda,
The question generation unit determines whether there are a plurality of elements that cause the identified element as the other elements, and determines that there are a plurality of elements that cause the plurality of elements. A question on the agenda is generated based on the selected element, and when it is determined that there are not a plurality of causative elements, a question on the agenda is generated based on the element as a result of the identified element. Generate
The speech analysis processing unit acquires evaluation information of the acquired speech,
The question generation unit acquires evaluation information of the identified element, determines whether or not the acquired evaluation information of the utterance corresponds to the acquired evaluation information of the element, and both correspond to each other If it is determined that there is not, extract the word having evaluation information different from the evaluation information of the element among the decomposed words, and based on the word different from the extracted word and the other elements, Generate questions about the agenda,
The question generation unit determines whether or not the element corresponding to the utterance is an initial element in the discussion, and the evaluation information of the acquired utterance is determined only when the element is not an initial element. And whether or not the evaluation information of the identified element corresponds,
The question generation unit determines whether or not the other element having a causal relationship with the element corresponding to the statement is important based on information indicating the importance in the discussion set for each of the elements. If it is determined that the other element is important, a question about the agenda is generated based on the other element, and if it is determined that the other element is not important, the decomposition is performed. Generate questions about the agenda based on
The question generation unit determines whether or not the identified element is an element within the authority of the person who is performing the discussion, and only when the element is determined to be an element within the authority. Generate questions about the agenda,
The speech analysis processing unit acquires a viewpoint of the decomposed word,
The discussion support system further includes a viewpoint output unit that outputs a viewpoint of the element used for generating the question and a viewpoint of the acquired word.
The discussion support system according to claim 1. A discussion support method for supporting a discussion on an agenda comprising a plurality of elements by a computer having a processor and a memory,
The computer
Acquiring the content of a statement made by a speaker on the agenda item,
Decomposing the acquired utterance into a plurality of words,
Based on the decomposed word, the element corresponding to the acquired utterance is specified, it is determined whether there is another element having a causal relationship with the specified element, and it is determined that there is the other element. If so, generate a question about the agenda based on the other factors,
Output the generated question,
Discussion support method. If the computer determines that there is no other element, the computer selects at least one of the viewpoints of the identified element, and based on the viewpoint of the selected element, Generate questions about the agenda,
The discussion support method according to claim 10. The computer
Obtain evaluation information of the acquired remarks,
Obtain evaluation information of the identified element,
It is determined whether or not the acquired evaluation information of the utterance and the acquired evaluation information of the element correspond to each other. If it is determined that both do not correspond, among the decomposed words, the Extracting the word having evaluation information different from the evaluation information of the element, and generating a question on the agenda based on the word different from the extracted word and the other element;
The discussion support method according to claim 10. A discussion support program for supporting a discussion on an agenda comprising a plurality of elements, comprising: a computer having a processor and a memory;
A speech acquisition process for acquiring the content of a speech made by a speaker for the agenda;
A speech analysis process for decomposing the acquired speech into a plurality of words;
Based on the decomposed word, the element corresponding to the acquired utterance is specified, it is determined whether there is another element having a causal relationship with the specified element, and it is determined that there is the other element. If so, a question generation process for generating a question about the agenda based on the other elements;
A question output process for outputting the generated question;
A support program for discussion. When the question generation process determines that there is no other element, the question generation process selects at least one of the viewpoints of the plurality of viewpoints of the specified element, and based on the viewpoint of the selected element , A process for generating a question about the agenda,
The discussion support program according to claim 13. The speech analysis process is a process of acquiring evaluation information of the acquired speech,
The question generation function acquires evaluation information of the identified element, determines whether or not the acquired evaluation information of the utterance and the acquired evaluation information of the element correspond to each other, and both correspond to each other If it is determined that there is not, extract the word having evaluation information different from the evaluation information of the element among the decomposed words, and based on the word different from the extracted word and the other elements, A process for generating a question about the agenda;
The discussion support program according to claim 13.

Images