Meeting support system

Abstract

To improve efficiency of a conference.SOLUTION: A conference support system for supporting a conference receives an input of utterance details being the details of utterance from participants in the conference, determines the type of corresponding utterance on the basis of the utterance details input to the input part, and outputs at least either one of the utterance details, evaluation on the conference, or evaluation on the participants on the basis of a result of determination performed by the determination part.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a conference support system, a conference support device, a conference support method, and a program. [Background technology]

[0002] Conventionally, a method is known in which a conference support system reduces the burden of cumbersome tasks such as time adjustment and manages the progress of a conference so that the chairperson can concentrate on the agenda and proceed with the proceedings appropriately. In this method, the conference support system displays the progress of the conference and controls the display according to the progress of the conference. Specifically, the conference support system first sets the planned proceeding time required for each of a plurality of conference agendas in advance and creates a timetable for the entire conference. Next, the conference support system displays an indicator showing the elapsed time of the conference agenda at any time. Furthermore, the conference support system detects the end of the conference agenda and shifts the indicator to the next conference agenda. Then, when the conference agenda is ended, the conference support system reallocates the remaining planned proceeding time to each remaining conference agenda. In this way, the conference support system automatically adjusts the time so that the conference ends within the planned proceeding time, and the burden of the time adjustment work is reduced. Therefore, the chairperson can concentrate on the agenda and manage the progress of the proceedings appropriately (for example, Patent Document 1, etc.). Summary of the Invention [Problem to be solved by the invention]

[0003] However, the conventional method only manages the time for each agenda item in a meeting. On the other hand, there is a demand for visualizing the contents of a meeting to make the meeting more efficient.

[0004] An object of one aspect of the present invention is to provide a conference support system that can make a conference more efficient. [Means for solving the problem]

[0005] In one embodiment, a conference support system for supporting a conference, having one or more information processing devices, comprises an input unit that inputs statement content, which is the content of statements made by participants in the conference, a judgment unit that determines a corresponding type of statement based on the statement content input by the input unit, and an output unit that outputs at least one of the statement content, an evaluation of the conference, or an evaluation of the participants based on a judgment result by the judgment unit. Effect of the Invention

[0006] Meetings can be made more efficient. [Brief description of the drawings]

[0007] [Figure 1] 1 is a conceptual diagram illustrating an example of an overall configuration of a conference support system according to a first embodiment of the present invention. [Diagram 2] 1 is a block diagram showing an example of a hardware configuration of an information processing device according to an embodiment of the present invention. [Diagram 3] 1 is a functional block diagram illustrating an example of a functional configuration of a conference support system according to a first embodiment of the present invention. [Figure 4] FIG. 2 is a diagram showing an example of a screen displayed by the conference support system according to the first embodiment of the present invention. [Diagram 5] FIG. 2 is a diagram showing an example of a setting screen for setting an agenda, a goal, and the like in the conference support system according to the first embodiment of the present invention. [Figure 6] FIG. 2 is a diagram showing an example in which an alert and remaining time are displayed on a screen displayed by the conference support system according to the first embodiment of the present invention. [Figure 7] 10 is a flowchart showing an example of a process for determining a type of statement content by the conference support system according to the first embodiment of the present invention. [Figure 8] FIG. 2 is a diagram showing an example of a screen for switching between a summary and a full text in the conference support system according to the first embodiment of the present invention. [Figure 9]11 is a flowchart showing an example of a process for switching between a summary and a full text by the conference support system according to the first embodiment of the present invention. [Figure 10] FIG. 2 is a diagram showing an example of a screen showing a summary displayed by the conference support system according to the first embodiment of the present invention. [Figure 11] FIG. 11 is a functional block diagram illustrating an example of a functional configuration of a conference support system according to a second embodiment of the present invention. [Figure 12] FIG. 11 is a diagram showing an example of a summary screen displayed by the conference support system according to the second embodiment of the present invention. [Figure 13] FIG. 11 is a diagram showing an example of an evaluation screen displayed by the conference support system according to the second embodiment of the present invention. [Figure 14] FIG. 11 is a diagram showing an example of a screen for inputting a subjective evaluation displayed by the conference support system according to the second embodiment of the present invention. [Figure 15] 13 is a flowchart showing an example of a process of performing an evaluation based on a time during which a conference is held by the conference support system according to the second embodiment of the present invention. [Figure 16] FIG. 13 is a conceptual diagram illustrating an example of natural language processing using machine learning by a conference support system according to a third embodiment of the present invention. [Figure 17] FIG. 13 is a diagram illustrating an example of a GUI displayed by a conference support system according to a third embodiment of the present invention. [Figure 18] FIG. 13 is a diagram showing an example of a display of a topic generated by the conference support system according to an embodiment of the fourth embodiment of the present invention. [Figure 19] FIG. 13 is a diagram showing an example of display of an extraction result by the conference support system according to the fourth embodiment of the present invention. [Figure 20] 13 is a flowchart showing an example of topic generation by the conference support system according to an embodiment of the fourth embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] Hereinafter, an embodiment of the present invention will be described.

[0009] (First embodiment) (Overall configuration example) 1 is a conceptual diagram for explaining an example of the overall configuration of a conference support system according to a first embodiment of the present invention. For example, as shown in the figure, the conference support system 1 has a server 10, which is an example of a conference support device, and one or more terminals 11. The server 10 and each terminal 11 are connected via a network 12. The server 10 and the terminals 11 are information processing devices such as PCs (Personal Computers), and have, for example, the following hardware configuration.

[0010] (Hardware configuration example) FIG. 2 is a block diagram showing an example of a hardware configuration of an information processing device according to an embodiment of the present invention. For example, the server 10 and the terminal 11 have a CPU (Central Processing Unit) 10H1, an I / F (interface) 10H2, and an output device 10H3. The server 10 and the terminal 11 also have an input device 10H4, an HD (Hard Disk) 10H5, and a storage device 10H6. Furthermore, each piece of hardware is connected via a bus 10H7. The server 10 and the multiple terminals 11 may have the same hardware configuration or different hardware configurations. Below, an example in which the server 10 and the multiple terminals 11 have the same hardware configuration will be described using the server 10 as an example.

[0011] The CPU 10H1 is a calculation device that performs calculations to realize various processes and processing of various data, and a control device that controls each piece of hardware, etc. The server 10 may have a plurality of calculation devices or control devices.

[0012] The I / F 10H2 is, for example, a connector and a processing IC (Integrated Circuit), etc. For example, the I / F 10H2 transmits and receives data to and from an external device via a network, etc.

[0013] The output device 10H3 is a display or the like that displays a display screen, a GUI (Graphical User Interface), and the like.

[0014] The input device 10H4 is a touch panel, a keyboard, a mouse, a microphone, or a combination of these, for inputting user operations, etc. It is more preferable that the microphone is a so-called directional microphone that can determine the direction of a sound source.

[0015] The HD 10H5 is an example of an auxiliary storage device, which stores programs, files, and data such as various settings used in processing.

[0016] The storage device 10H6 is a so-called memory, etc. That is, programs or data stored in the HD 10H5 are read out to the storage device 10H6.

[0017] (Example of functional configuration) 3 is a functional block diagram illustrating an example of a functional configuration of a conference support system according to an embodiment of the present invention. For example, the conference support system 1 includes an input unit 1F1, a control unit 1F2, a storage unit 1F3, and an output unit 1F4.

[0018] For example, in the conference support system 1, the input unit 1F1 and the output unit 1F4 are included in the terminal 11 (FIG. 1). The control unit 1F2 and the storage unit 1F3 are included in the server 10 (FIG. 1). Note that each unit shown in the figure may be included in either the terminal 11 or the server 10.

[0019] The input unit 1F1 inputs the contents of statements made in a conference. For example, the input unit 1F1 is realized by the input device 10H4 (FIG. 2) or the like.

[0020] The control unit 1F2 performs control to realize each process. Specifically, the control unit 1F2 has a judgment unit 1F21. As shown in the figure, the control unit 1F2 may also have a decoration unit 1F22, a time management unit 1F23, and a speaker judgment unit 1F24. For example, the control unit 1F2 is realized by the CPU 10H1 (FIG. 2) or the like.

[0021] The storage unit 1F3 stores data that is input in advance. The storage unit 1F3 realizes a database, etc. The storage unit 1F3 also stores the contents of comments. For example, the storage unit 1F3 is realized by the HD 10H5 (FIG. 2) and the storage device 10H6 (FIG. 2), etc.

[0022] The output unit 1F4 displays or prints the contents of the meeting, and outputs the contents of the meeting, such as minutes, evaluation results, summaries, remarks, judgment results, or a combination of these, to the participants of the meeting in a preset format. For example, the output unit 1F4 is realized by the output device 10H3 (FIG. 2) or the like.

[0023] First, in the conference support system 1, the content of a statement made during a conference by a person who made the statement (hereinafter referred to as the "speaker") or a minutes keeper or the like is input by the input unit 1F1. Specifically, the content of a statement is input in text format, for example, by a keyboard or the like. The content of a statement may be the content that a participant verbally states and then input by a keyboard or the like later. In addition, in an online conference between remote locations using chat or the like, when a participant inputs the content he or she wishes to say by a keyboard, the content of a statement may be the content input in the chat or the like. In other words, the content of a statement is not limited to the content that is verbally stated. In other words, the content of a statement may be the content that is not verbally stated but is input by a keyboard or the like.

[0024] Alternatively, the utterance may be input by voice using a microphone or the like. In other words, when the utterance is input by voice, the conference support system 1 may perform voice recognition on the input voice to generate text data or the like. The utterance may also be input by handwriting on an electronic whiteboard, a tablet terminal, or the like. In other words, when the utterance is input by handwritten characters, the conference support system 1 may perform OCR recognition on the input handwritten characters to generate text data or the like.

[0025] In addition, when the speaker determination unit 1F24 identifies who the speaker is from among the participants, the speaker determination unit 1F24 may use a plurality of directional microphones or the like to link the speaker with the direction in which the source of the sound can be determined to be located, and identify the speaker from among the participants. Alternatively, the speaker determination unit 1F24 may perform voice authentication on the input voice data and identify the speaker based on the authentication result. Alternatively, the speaker determination unit 1F24 may identify the speaker by photographing the speaker with a camera or the like connected to the network 12 (FIG. 1) and performing face authentication using the photographed image.

[0026] Furthermore, information that can identify who made the statement may be input by the input unit 1F1 or the like. Specifically, for example, speaker information is first assigned in advance to each of a plurality of function keys ("F1" to "F12") on the keyboard. Then, when the statement is to be input, the conference support system 1 may add the speaker information to the statement when the function key is pressed. In this way, the statement may be linked to the speaker.

[0027] Next, the conference support system 1 judges the type of the inputted utterance by the judgment unit 1F21. Specifically, the types of utterances are, for example, "proposal," "question," "answer," "positive opinion," "negative opinion," "neutral opinion," "information," "request," "issue," "action item" (hereinafter sometimes referred to as "AI"), and "decision item." Below, an example of classifying the contents of utterances into such utterance types will be described.

[0028] First, in the conference support system 1, expressions defined in advance for each type of utterance are stored in the storage unit 1F3 in order to determine the type of utterance. Moreover, the expression is a word, phrase, or turn of phrase, and the storage unit 1F3 stores the type of utterance in association with an expression specific to that type of utterance. Then, when the input unit 1F1 inputs the content of the utterance, the determination unit 1F21 determines whether the content of the utterance includes an expression stored in the storage unit 1F3. Next, when the content of the utterance includes an expression stored in the storage unit 1F3, the determination unit 1F21 identifies the corresponding type from among the types of utterance. In this way, the conference support system 1 can determine the type of utterance of the content of the utterance made in the conference.

[0029] The output unit 1F4 also outputs the conference contents. For example, the output unit 1F4 displays the comment contents in an embellished manner based on the type of comment determined by the determination unit 1F21.

[0030] (Input and output examples) In the conference support system, for example, each terminal displays a screen such as the one shown below and performs each process.

[0031] 4 is a diagram showing an example of a screen displayed by the conference support system according to the first embodiment of the present invention. For example, the screen displayed on each screen of each terminal has a display and GUI as shown in the figure. The following description will be given taking the screen shown in the figure as an example.

[0032] In this example, an output unit displays an agenda 301 and the purpose of the meeting, i.e., a goal 302, on the screen. When the agenda 301 and the goal 302 are displayed during the meeting, it is possible to align the participants' recognition and the vector of the discussion. Therefore, the meeting support system can reduce deviations in the meeting. The agenda 301 and the goal 302 are set in advance, for example, before the meeting is held or at the beginning of the meeting, as follows.

[0033] 5 is a diagram showing an example of a setting screen for setting an agenda, goals, etc. in the conference support system according to the first embodiment of the present invention. For example, a setting screen as shown in the figure is displayed by the conference support system at the beginning of a conference. After the settings shown below are made on the setting screen, the settings are completed when start button 505 is pressed. In other words, after start button 505 is pressed, the screen shown in FIG. 4 and the like are displayed by the conference support system.

[0034] As shown in the figure, a participant inputs an agenda into an agenda setting text box 501. Furthermore, the participant sets a goal for each agenda input into the agenda setting text box 501. The goal is input, for example, into a goal setting text box 502. In this way, the agenda and goal input into the agenda setting text box 501 and goal setting text box 502 are displayed by the conference support system as agenda 301 and goal 302, as shown in FIG.

[0035] Also, there are cases where a meeting notice or the like is created before a meeting is held. In such cases, the agenda and goal written in the meeting notice may be displayed as agenda 301 and goal 302 by the meeting support system as shown in FIG. 4. When the meeting notice or the like is notified to the participants in advance in this way, the participants can make preparations in advance, making the meeting more efficient. Furthermore, when the meeting notice or the like is notified to the participants in advance, the participants can determine whether or not they need to participate, and can avoid participating in unnecessary meetings.

[0036] Returning to Fig. 4, the current time 306, time allocation 307, progress status 308, etc. are displayed on the screen by the conference support system. The current time 306 is displayed based on a preset time or data indicating the time acquired via a network or the like. Furthermore, the time allocation 307 is displayed based on the scheduled time for each agenda item that is preset. For example, the scheduled time for each agenda item is input in the scheduled time setting text box 503 shown in Fig. 5. Furthermore, the progress status 308 is displayed as a so-called progress bar or the like that indicates the proportion of the time that has elapsed based on the time that has elapsed from the conference start time to the current time 306, as shown in the figure.

[0037] The time management unit 1F23 (FIG. 3) and the output unit 1F4 (FIG. 3) may also display the following. For example, when the scheduled end time of the entire meeting or each agenda item approaches, the meeting support system may display the remaining time or an alert to encourage wrapping up the meeting. For example, in an example using the screen shown in FIG. 4, the meeting support system displays the following:

[0038] FIG. 6 is a diagram showing an example of an alert and remaining time displayed on a screen displayed by the conference support system according to an embodiment of the first embodiment of the present invention. As shown in the figure, the remaining time 401 and the alert 402 may be displayed by the conference support system. In the example shown in the figure, the remaining time 401 is an example of a display indicating that the time from the current time to the scheduled end time of the entire conference is "less than 10 minutes". In addition, the alert 402 is an example of a display notifying the participants that the remaining time of the conference is getting short, and therefore that they should conclude the conference, for example. When such a display is made, the participants can easily secure time to conclude the discussion. Therefore, the conference support system can reduce the occurrence of participants becoming too engrossed in the discussion, running out of time, and making the conclusion or action items of the conference unclear.

[0039] The alert and the remaining time are not limited to be output in the format shown in the figure, and may be output, for example, by voice or the like.

[0040] Returning to FIG. 4, the comment content 303 is displayed on the screen. The comment content 303 is decorated based on the judgment result by the judgment unit 1F21 (FIG. 3). For example, as shown in the figure, each comment content is displayed with a display indicating the type of comment, such as "Q:", "A:", "Proposal:", "Opinion (P):", "Opinion (N):", "Information:", "Agenda:" and "AI candidate:", added to the beginning or end of the sentence. The decoration may be any decoration that can distinguish the type of each comment. For example, the decoration may be color coding, character decoration such as italics and bold, font type, addition of characters to the beginning or end of the sentence, or a combination of these. The way of decoration can be set in advance in the conference support system. Specifically, for example, a comment content determined to be a "positive opinion" is set to be displayed in blue and the display of "Opinion (P)" is added to the beginning of the sentence. Additionally, comments that are judged to be "negative opinions" and "issues" are displayed in red, and the indications "Opinion (N)" and "Issues" are added to the beginning of each sentence. In this way, when the type of comment is indicated by decoration, participants can intuitively grasp important points in the meeting, the contents of the discussion, and the trends of the meeting. Note that decoration may be processed by the server 10 (FIG. 1) or by a browser installed on the terminal 11 (FIG. 1).

[0041] Furthermore, when a speaker is specified, the speaker may be displayed by the conference support system as shown in the figure. In the example shown in the figure, the speaker's name is displayed in parentheses. Note that the speaker may be displayed in a format other than that shown in the figure. Also, a person who is a candidate for a speaker, that is, a participant, is set in advance in a participant setting text box 504 on the setting screen shown in FIG. 5, for example. Also, as shown in FIG. 4, the participant set in the participant setting text box 504 is displayed by the conference support system as in participant display 305.

[0042] In the illustrated example, a function key is set to be associated with each participant displayed in participant display 305. For example, in the illustrated example, when one of the participants, "Sato", speaks, if the function key "F2" is pressed when inputting the content of the speech made by "Sato" in the speech content input text box 304, the inputted content of the speech is associated with "Sato", which is the speaker information. That is, in the illustrated example, when the function key "F2" is pressed when inputting the content of the speech in the speech content input text box 304, the conference support system displays the inputted content of the speech with "(Sato)" added.

[0043] The type of comment is determined, for example, by the process shown below. In the following, an example will be described in which the types of comments are "question", "answer", "suggestion", "issue", and "opinion".

[0044] FIG. 7 is a flowchart showing an example of a process for determining the type of statement content by the conference support system according to the first embodiment of the present invention.

[0045] In step S01, the conference support system initializes the question flag. Specifically, the conference support system sets the question flag to "FALSE." The question flag is an example of data indicating whether the type of the previously input utterance is a "question." Hereinafter, when the question flag is "FALSE," this indicates that the type of the previously input utterance is a type other than a "question." On the other hand, when the question flag is "TRUE," this indicates that the type of the previously input utterance is a "question."

[0046] In step S02, the conference support system inputs the content of the statement. Specifically, the conference support system inputs the content of the statement by the input unit 1F1 (FIG. 3). That is, in the example shown in FIG. 4, the content of the statement is input as text or the like into the content input text box 304.

[0047] In step S03, the conference support system determines whether the question flag is "TRUE." If the question flag is "TRUE" (YES in step S03), the conference support system proceeds to step S04. On the other hand, if the question flag is not "TRUE," that is, if it is "FALSE" (NO in step S03), the conference support system proceeds to step S05.

[0048] In step S04, the conference support system determines that the type of the utterance content is an "answer". The "answer" is a utterance content that is returned to the utterance content that corresponds to the utterance type of the "question". Therefore, the "answer" is allowed to be expressed freely. In other words, there are cases where an expression that can be determined as an "answer" cannot be set in advance for the "answer". Therefore, as shown in step S03, the conference support system determines whether the immediately preceding utterance content is a "question", that is, whether the type of the utterance content is an "answer" based on the question flag. In other words, if the immediately preceding utterance content is a "question", the conference support system determines that the type of the utterance content of the next utterance content is an "answer". On the other hand, if the immediately preceding utterance content is other than a "question", the conference support system determines that the type of the utterance content of the next utterance content is other than an "answer". In this way, the conference support system classifies the utterance content of the "answer". In addition, in step S04, the conference support system sets the question flag to "FALSE".

[0049] In step S05, the conference support system judges whether the utterance content includes an expression of "question". For example, the conference support system judges the type of the utterance content based on the auxiliary verb portion of the utterance content. Specifically, first, expressions such as "?", "Is it ~?", or "Is it not ~?" are stored in advance in storage unit 1F3 (FIG. 3). Next, the conference support system judges by judgment unit 1F21 (FIG. 3) whether all or part of the utterance content input by input unit 1F1 (FIG. 3) matches the stored expression. That is, the judgment is realized by so-called pattern matching or the like.

[0050] Next, if it is determined in step S05 that the utterance content includes the expression "question" (YES in step S05), the conference support system proceeds to step S06. On the other hand, if it is determined in step S05 that the utterance content does not include the expression "question" (NO in step S05), the conference support system proceeds to step S08.

[0051] In step S06, the conference support system determines that the type of the comment content is a "question."

[0052] In step S07, the conference support system sets the question flag to "TRUE."

[0053] In step S08, it is determined whether the utterance content includes the expression "suggestion." For example, the conference support system determines the type of the utterance content based on the auxiliary verb portion of the utterance content, similar to "question." Specifically, first, expressions such as "should" or "it would be better to do" are stored in advance in storage unit 1F3 (FIG. 3). Next, using determination unit 1F21 (FIG. 3), the conference support system determines whether all or part of the utterance content input by input unit 1F1 (FIG. 3) matches the stored expression.

[0054] Next, if it is determined in step S08 that the content of the statement includes the expression "proposal" (YES in step S08), the conference support system proceeds to step S09. On the other hand, if it is determined in step S08 that the content of the statement does not include the expression "proposal" (NO in step S08), the conference support system proceeds to step S10.

[0055] In step S09, the conference support system determines that the type of the comment content is a "proposal."

[0056] In step S10, it is determined whether the utterance content includes an expression of "issue." For example, the conference support system determines the type of utterance content based on the auxiliary verb portion of the utterance content, similar to "question." Specifically, first, expressions such as "need to do" or "have to do" are stored in advance in storage unit 1F3 (FIG. 3). Next, using determination unit 1F21 (FIG. 3), the conference support system determines whether all or part of the utterance content input by input unit 1F1 (FIG. 3) matches the stored expression.

[0057] Next, if it is determined in step S10 that the utterance content includes the expression "issue" (YES in step S10), the conference support system proceeds to step S11. On the other hand, if it is determined in step S10 that the utterance content does not include the expression "issue" (NO in step S10), the conference support system proceeds to step S12.

[0058] In step S11, the conference support system determines that the type of the comment content is "topic."

[0059] In step S12, the conference support system determines that the type of the comment content is an "opinion."

[0060] As described above, steps S02 to S12 are repeatedly performed for each comment content. By performing such processing, the conference support system can determine the type of comment content.

[0061] The conference support system may also display a summary. For example, when the following operations are performed, the conference support system switches between the summary and the full text.

[0062] 8 is a diagram showing an example of a screen for switching between a summary and a full text by the conference support system according to the first embodiment of the present invention. The illustrated screen is an example of a screen displayed during a conference. When the summary instruction button BN1, which is shown as "Display extracted important sentences", is pressed, the conference support system switches the screen to display the summary.

[0063] 9 is a flowchart showing an example of a process for switching between a summary and a full text by the conference support system according to the first embodiment of the present invention. The conference support system performs the process shown in the figure to generate a summary. The types of comments are as follows: "suggestion", "question", "answer", "positive opinion", "negative opinion", "neutral opinion", "information", "request", "issue", "action item", "decision", etc.

[0064] In step S21, the conference support system initializes a summary flag. Specifically, the conference support system sets the summary flag to "FALSE." The summary flag is a flag that switches between extracting important parts of the input remarks and displaying a summary, or displaying the full text showing all remarks. Hereinafter, when the summary flag is "FALSE," the conference support system is set to display the full text. On the other hand, when the summary flag is "TRUE," the conference support system is set to display a summary.

[0065] In step S22, the conference support system accepts pressing of the summary instruction button BN1 (FIG. 8). Each time the summary instruction button BN1 is pressed, the display switches between the full text and the summary.

[0066] In step S23, the conference support system determines whether the summary flag is "TRUE." If it is determined that the summary flag is "TRUE" (YES in step S23), the conference support system proceeds to step S24. On the other hand, if it is determined that the summary flag is not "TRUE," that is, is "FALSE" (NO in step S23), the conference support system proceeds to step S26.

[0067] In step S24, the conference support system displays the entire text. Specifically, for example, a screen shown in Fig. 8 is displayed by the conference support system. That is, in the entire text, the conference support system displays all types of utterance contents among the utterance contents inputted.

[0068] In step S25, the conference support system sets the summary flag to "FALSE."

[0069] In step S26, the conference support system displays the summary. Specifically, for example, the following screen is displayed by the conference support system.

[0070] Fig. 10 is a diagram showing an example of a screen showing a summary displayed by the conference support system according to the first embodiment of the present invention. Compared with the screen shown in Fig. 8, the screen shown in Fig. 10 is different in that a summary display ABS is displayed.

[0071] The summary is a result of extracting important utterance contents from among the utterance contents input. Whether or not the utterance contents are important is determined based on whether or not the utterance contents are of a type of utterance contents that is set in advance as important. For example, the summary is a result of extracting utterance contents of utterance types determined as "proposal," "issue," and "action item" from all types of utterance contents input. That is, the summary is a meeting minutes generated by the conference support system extracting utterance contents of a predetermined utterance type from the utterance contents input. Note that the illustrated example is an example in which the utterance contents of the utterance types determined as "issue" and "action item" are extracted by the conference support system from the full text shown in FIG. 8, and a summary is generated. As illustrated, the conference support system displays the utterance contents of the target utterance type and hides the utterance contents of the other utterance types. That is, when the summary shown in FIG. 10 is displayed, the participants can easily create action items. In this way, when the summary is displayed, the participants can review the key points of the conference. Therefore, the conference support system can make the participants organize the issues, action items, or decisions without missing anything. In this way, the conference support system can make the conference more efficient.

[0072] Returning to FIG. 9, in step S27, the conference support system sets the summary flag to "TRUE."

[0073] As described above, steps S22 to S27 are repeatedly performed by the conference support system. By performing such processing, the conference support system can switch between displaying the full text and the summary.

[0074] Second embodiment The second embodiment is an embodiment that is realized, for example, by the same overall configuration and hardware configuration as the first embodiment. Therefore, the description of the overall configuration and hardware configuration will be omitted, and the following description will focus on the differences. The second embodiment has a different functional configuration from the first embodiment.

[0075] Fig. 11 is a functional block diagram illustrating an example of the functional configuration of a conference support system according to a second embodiment of the present invention. Compared with Fig. 3, the functional configuration illustrated in Fig. 11 differs in that the control unit 1F2 includes a cost calculation unit 1F31 and an evaluation unit 1F32.

[0076] As shown in the figure, the conference support system 1 preferably further includes a translation processing unit 1F5.

[0077] The cost calculation unit 1F31 calculates the cost incurred by holding the conference based on the personnel expenses of the participants.

[0078] The evaluation unit 1F32 evaluates the conference based on the duration of the conference, the number of remarks, the types of remarks, and the ratio of the types of remarks.

[0079] The cost calculation unit 1F31 calculates the cost and the evaluation unit 1F32 evaluates the cost, for example, when a summary button 309, which is a button for "to summary / save screen" shown in Fig. 4, is pressed or when an alert 402 is displayed as shown in Fig. 6. For example, the conference system displays a summary screen as shown below.

[0080] 12 is a diagram showing an example of a summary screen displayed by the conference support system according to an embodiment of the second embodiment of the present invention. In the example of the screen shown in the figure, an agenda display 801 displays an agenda set by the conference support system in the agenda setting text box 501 (FIG. 5) in the opening notice or at the beginning of the conference. Furthermore, in the example of the screen shown in the figure, a goal display 802 displays a goal set by the conference support system in the goal setting text box 502 (FIG. 5) in the opening notice or at the beginning of the conference.

[0081] Based on the contents displayed on the agenda display 801 and the goal display 802, a conclusion to the discussion is input in a conclusion input text box 803. Based on the contents displayed on the agenda display 801 and the goal display 802, an action item decided in the conference is input in an action item input text box 804. A date, i.e., a deadline, may be set for the action item. In addition, a person in charge may be set for the action item. In the action item input text box 804, the conference support system may display, as a candidate action item, a statement determined to be an "action item," or may display, as a candidate action item, a sentence that has been compressed to be described later or a sentence that has been simplified by formatting the expression at the end of the sentence.

[0082] Furthermore, the conference support system displays the cost calculation by the cost calculation unit 1F31 (FIG. 11) and the evaluation by the evaluation unit 1F32 (FIG. 11) on an evaluation screen such as the one shown below.

[0083] 13 is a diagram showing an example of an evaluation screen displayed by the conference support system according to the second embodiment of the present invention. For example, the conference support system displays the cost and evaluation result of the conference as a score, amount, etc. as shown in the figure.

[0084] The meeting score 901 is a comprehensive score obtained by adding up the scores of the achievement evaluation score 902 , the meeting time evaluation score 903 , the activity evaluation score 904 , and the positive evaluation score 905 .

[0085] The achievement evaluation score 902 is an evaluation result of whether or not a conclusion has been input for each of the previously set goals for each agenda set in the agenda setting text box 501 (FIG. 5) etc. For example, when four agendas are set in the agenda setting text box 501 (FIG. 5) etc. and two conclusions are input for the four agendas, the conference support system sets the achievement evaluation score 902 to half of the full score.

[0086] The meeting time evaluation score 903 is an evaluation result obtained by comparing the scheduled time set in the meeting notification or the scheduled time setting text box 503 (FIG. 5) with the time the meeting took place. For example, if the time the meeting took place is shorter than the scheduled time, the meeting time evaluation score 903 will be high. On the other hand, if the time the meeting took place is longer than the scheduled time, the meeting time evaluation score 903 will be low.

[0087] The activity evaluation score 904 is an evaluation result based on the number of remarks or the number of remarks per unit time. For example, if the number of remarks or the number of remarks per unit time is large, the activity evaluation score 904 is high. On the other hand, if the number of remarks or the number of remarks per unit time is small, the activity evaluation score 904 is low. For this reason, if the activity evaluation score 904 is low, the meeting may be a place for unnecessary information sharing, or people unrelated to the agenda may be gathered and so-called side work may be taking place. Therefore, the meeting support system can encourage improvements in the management of the meeting by feeding back these possibilities.

[0088] The positivity evaluation score 905 is an evaluation result based on the types of utterances and the proportion of the types of utterances judged by the judgment unit 1F21 (FIG. 11). For example, if the proportion of types of utterances evaluated as "suggestions" is high, the positivity evaluation score 905 will be high. On the other hand, if the proportion of types of utterances evaluated as "suggestions" is low, the positivity evaluation score 905 will be low. For this reason, if the positivity evaluation score 905 is low, there is a possibility that the meeting is one with a lot of criticism and comments, and does not move closer to a conclusion. Therefore, the meeting support system can suggest this possibility and encourage participants to offer constructive opinions toward the goal.

[0089] The conference support system performs the above evaluation using the evaluation unit 1F32 (FIG. 11). Next, the conference support system can provide participants with objective feedback of the evaluation results for the conference by outputting the evaluation results to a screen or the like using the output unit 1F4 (FIG. 11). When feedback is provided in this way, the conference support system can suggest to participants which parts of the conference are likely to be problematic, or make participants aware of the time the conference took place, etc. This allows the conference support system to have participants reconsider which members to include in the conference or how the conference will proceed.

[0090] The conference support system may display each evaluation result, the number of utterances, the type of utterances, and the like during the conference. For example, in the example shown in Fig. 4, the evaluation results and the like may be displayed as in evaluation display 310. That is, the evaluation results and the like may be displayed in real time during the conference. In this way, the conference support system can, for example, encourage people who do not speak much to speak more.

[0091] Furthermore, the conference support system may weight utterances that are determined to be "issues" and "suggestions" in the evaluation. In this way, the conference support system can encourage participants to make more utterances that are types of "issues" and "suggestions". Therefore, the conference support system can encourage more useful discussions.

[0092] Additionally, the conference support system may display the cost calculation results, etc., on the evaluation screen. For example, the cost is calculated by the cost calculation unit 1F31 (FIG. 11) based on the conference time 906 and the labor costs of each participant, etc. The calculated cost is displayed, for example, as a cost display 907, etc.

[0093] To calculate the cost, first, the labor cost per unit time of the participants is input to the conference support system. The labor cost may be a fixed amount for all participants, or may be set for each position. Alternatively, the labor cost may be set based on actual salary, etc.

[0094] Next, the conference support system measures the time that has elapsed from the conference start time to the conference end time, and displays the measurement result in conference time 906. The conference support system then multiplies the labor cost per unit time of each participant by conference time 906 to calculate the cost of each participant. Next, the conference support system sums up the costs of each participant, and displays the calculation result in cost display 907.

[0095] Furthermore, the evaluation result for each participant may be displayed on the evaluation screen by the conference support system. For example, the evaluation result for each participant may be displayed as a contribution level display 908 or the like.

[0096] The contribution display 908 shows the evaluation result for each participant based on, for example, the type of comment and the number of comment contents judged by the judgment unit 1F21 (FIG. 11). Note that the comment contents are linked to the participant who made each comment by the speaker judgment unit 1F24 shown in FIG. 3, etc.

[0097] The contribution level display 908 is calculated by the conference support system, for example, as follows. First, the participants set weights for the types of comments that are considered important in accordance with the purpose of the conference, such as assigning "1 point" to a "neutral opinion" and "5 points" to "issues" and "suggestions." It is preferable that the conference support system be set so that constructive opinions such as "issues" and "suggestions" are given greater weight. When set in this way, the conference support system can encourage useful discussions toward problem solving.

[0098] Next, the conference support system evaluates each participant by multiplying the result of the judgment by the judgment unit 1F21 (FIG. 11) and the number of comments. The values ​​calculated by this evaluation are the "degree of contribution" and "earned points" shown in the contribution display 908. In this way, when the "degree of contribution" etc. is fed back to the participants by the conference support system, the participants can use the "degree of contribution" etc. to reconsider which members to allow to participate in the conference. Therefore, the conference support system can encourage more useful discussions. Furthermore, the conference support system can reduce the cost of the conference by reducing the number of unnecessary members participating.

[0099] The above-mentioned screens may be shared in real time by the terminals 11 connected to the network 12 shown in Fig. 1. In this way, even if a conference is held in a place without a display, if the participants bring their own terminals 11 such as notebook PCs, they can view the above-mentioned screens and hold the conference.

[0100] Furthermore, the comments and the like may be edited collaboratively from each terminal 11. For example, the server 10 converts the content to be displayed on the screen and the like into, for example, an HTML (Hyper Text Markup Language) file and transmits it to each terminal 11. Next, when the comments and the like are input and data indicating the comments and the like is transmitted to the server 10, the server 10 transmits an HTML file reflecting the new comments and the like to each terminal 11.

[0101] Furthermore, for real-time sharing, a protocol for two-way communication such as WebSocket is used. WebSocket is a protocol disclosed in "https: / / ja.wikipedia.org / wiki / WebSocket" and the like.

[0102] The words displayed on the screen may be subjected to a translation process. For example, when a statement or the like is input and data indicating the statement or the like is transmitted to the server 10, the server 10 translates the statement and generates an HTML file indicating the statement before translation and the statement after translation. The server 10 then transmits the HTML file indicating the statement before translation and the statement after translation to each terminal 11. In this way, the conference support system can reduce miscommunication even when participants speak different languages.

[0103] The translation processing is realized by the translation processing unit 1F5 (FIG. 11). The translation processing unit 1F5 has a functional configuration including, for example, a translation unit, a translation language setting unit, and a translation result storage processing unit, which are shown below. Specifically, the translation processing is realized by the method disclosed in, for example, JP2015-153408A. The translation processing unit 1F5 is realized by, for example, the CPU 10H1 (FIG. 2), etc.

[0104] First, the translation language setting unit sets the language in which the participants will speak (hereinafter referred to as the "input language") and the language into which the input language is translated (hereinafter referred to as the "translation language") based on a user operation.

[0105] The translation unit then performs part-of-speech decomposition and syntactic analysis on the utterance content input in the input language by referring to the language model and the word dictionary. In this way, the translation unit translates the utterance content input in the input language into the translation language.

[0106] Next, the translation result storage processor stores the characters, etc., of the translation language generated by the translation unit through translation in a text file, etc. Then, the conference support system may output the characters, etc., of the translation language stored in the translation result storage processor.

[0107] Furthermore, each terminal 11 may perform user authentication. For example, user authentication is performed by inputting a user name into a user name input text box displayed on the screen. Alternatively, user authentication may be performed using an IC card, RFID, or the like. In this manner, similar to the function key assignment shown in FIG. 4, the speaker and the content of the speech may be linked based on the authentication result of the above user authentication and the terminal into which the content of the speech was input. Note that information related to the speaker may be added by coloring, displaying the beginning or end of a sentence, or the like, similar to the first embodiment.

[0108] Furthermore, a subjective evaluation may be input for the content of the speech, and the conference support system may perform an evaluation based on the input subjective evaluation. For example, the subjective evaluation may be input as follows:

[0109] Fig. 14 is a diagram showing an example of a screen for inputting a subjective evaluation displayed by the conference support system according to the second embodiment of the present invention. Compared to Fig. 6, the example of the screen shown in the figure is different in that a subjective evaluation button 1001 is added for each of the displayed remarks.

[0110] As shown in the figure, a subjective evaluation button 1001 is displayed for each utterance content. Then, when each participant judges that the utterance content is one that he or she supports, the participant presses the subjective evaluation button 1001 displayed for the utterance content that he or she supports. The subjective evaluation button 1001 may be pressed by the participant in charge of inputting the subjective evaluation, or each participant may press the button based on their own judgment. The subjective evaluation button 1001 may also be a button that is pressed when the participant does not support the utterance. In this way, when the subjective evaluation is input, the conference support system can evaluate whether or not each utterance content is a good utterance based on the subjective opinion of each participant. Then, the conference support system can display the evaluation result based on the subjective evaluation, and encourage the participants to make good utterances that will help achieve the purpose of the conference.

[0111] Furthermore, the conference support system may display the speech content with decoration based on the input from the subjective evaluation button 1001. For example, in the illustrated screen, the conference support system displays the speech content for which the subjective evaluation button 1001 is pressed as a large display 1002. As illustrated, the large display 1002 is an example of decoration in which the font size of the characters indicating the speech content is larger than that of other speech contents. When the speech content is output with decoration in this way, the participants can intuitively know whether that speech content is supported among the speech contents. Furthermore, the conference support system may display the status of approval or disapproval of the speech content, such as who is in favor and who is against a particular opinion, by using user information.

[0112] The decoration is realized by, for example, using support and non-support information and modifying the description of CSS (Cascading Style Sheets) using a language such as Javascript (registered trademark).

[0113] Furthermore, when displaying summaries, the conference support system may display utterances that are determined to be important based on the input subjective evaluation, in addition to displaying based on the type of utterance. In this way, participants can better understand the process of the discussion leading up to the conclusion. In other words, this allows the conference support system to generate minutes that are effective for reviewing the conference and sharing information.

[0114] Furthermore, the evaluation unit 1F32 (FIG. 11) may evaluate a conference based on the time the conference took place in. For example, the evaluation unit 1F32 (FIG. 11) performs the evaluation as follows.

[0115] FIG. 15 is a flowchart showing an example of a process of performing an evaluation based on the duration of a conference by the conference support system according to the second embodiment of the present invention.

[0116] In step S31, the conference support system accepts pressing of an evaluation button. For example, when an evaluation button such as "Rate this conference" shown in FIG. 10 is pressed, the conference support system starts the evaluation. The duration of the conference is assumed to be the time from when the conference started to when the conference ended (hereinafter referred to as "elapsed time"). Furthermore, the scheduled duration of the conference (hereinafter simply referred to as "scheduled duration") is assumed to be input to the conference support system in advance, such as before the evaluation button is pressed.

[0117] In step S32, the conference support system determines whether the elapsed time exceeds the scheduled time by 10% or more. If the conference support system determines that the elapsed time exceeds the scheduled time by 10% or more (YES in step S32), the conference support system proceeds to step S33. On the other hand, if the conference support system determines that the elapsed time does not exceed the scheduled time by 10% or more (NO in step S32), the conference support system proceeds to step S34.

[0118] In step S33, the conference support system sets the evaluation point to "0" points.

[0119] In step S34, the conference support system determines whether the elapsed time exceeds the scheduled time by less than 10%. If the conference support system determines that the elapsed time exceeds the scheduled time by less than 10% (YES in step S34), the conference support system proceeds to step S35. On the other hand, if the conference support system determines that the elapsed time does not exceed the scheduled time by less than 10% (NO in step S34), the conference support system proceeds to step S36.

[0120] In step S35, the conference support system assigns the evaluation score of "5" points.

[0121] In step S36, the conference support system determines whether the elapsed time exceeds the scheduled time by less than 5%. If the conference support system determines that the elapsed time exceeds the scheduled time by less than 5% (YES in step S36), the conference support system proceeds to step S37. On the other hand, if the conference support system determines that the elapsed time does not exceed the scheduled time by less than 5% (NO in step S36), the conference support system proceeds to step S38.

[0122] In step S37, the conference support system assigns the evaluation score of "10" points.

[0123] In step S38, the conference support system determines whether the elapsed time is 10% or more shorter than the scheduled time. If the conference support system determines that the elapsed time is 10% or more shorter than the scheduled time (YES in step S38), the conference support system proceeds to step S39. On the other hand, if the conference support system determines that the elapsed time is not 10% or more shorter than the scheduled time (NO in step S38), the conference support system proceeds to step S40.

[0124] In step S39, the conference support system sets the evaluation point to "25" points.

[0125] In step S40, the conference support system determines whether the elapsed time is 5% or more shorter than the scheduled time. If the conference support system determines that the elapsed time is 5% or more shorter than the scheduled time (YES in step S40), the conference support system proceeds to step S41. On the other hand, if the conference support system determines that the elapsed time is not 5% or more shorter than the scheduled time (NO in step S40), the conference support system proceeds to step S42.

[0126] In step S41, the conference support system sets the evaluation point to "20" points.

[0127] In step S42, the conference support system sets the evaluation point to "15" points.

[0128] For example, as shown in the figure, the conference support system evaluates a case where the elapsed time is shorter than the scheduled time so as to give a higher evaluation point. On the other hand, the conference support system evaluates a case where the elapsed time exceeds the scheduled time so as to give a lower evaluation point. Note that the numerical values ​​of the evaluation point and the judgment criteria are not limited to the values ​​and judgments shown in the figure, and may be changeable by settings, etc. As shown in the figure, the evaluated evaluation point is displayed by the conference support system.

[0129] The conference support system may also display the evaluation results using stars or the like, as shown in FIG. 13. Furthermore, the conference support system may also display comments or the like encouraging behavioral change, as shown in FIG. 13. Furthermore, the conference support system may display the evaluation results relating to time together with other evaluation results, as shown in FIG. 13. When the evaluation results evaluating a conference based on the time the conference was held are displayed in this way, the conference support system can make the participants conscious of the time involved in the conference and encourage them to achieve their goals within the time. In this way, the conference support system can make the conference more efficient.

[0130] Third embodiment The third embodiment is an embodiment that is realized, for example, by the same overall configuration, hardware configuration, and functional configuration as the first embodiment. The following mainly describes the differences. The third embodiment is different from the first embodiment in the determination method used by the determination unit 1F21 (FIG. 3).

[0131] In the third embodiment, the conference support system uses a tagged corpus as so-called training data. The conference support system then learns a classifier that performs natural language processing by performing so-called supervised learning, which is a machine learning method. Below, an example will be described in which the tags, i.e., types of comments, are "suggestions," "questions," "answers," "positive opinions," "negative opinions," "neutral opinions," "information," "requests," "issues," "action items," and "decisions."

[0132] The machine learning method is, for example, a support vector machine (SVM), a naive Bayes classifier, decision tree learning, or a conditional random field (CRF). The machine learning method is preferably a support vector machine. A support vector machine can make decisions sequentially, improving real-time performance.

[0133] The classifier determines the type of utterance based on the characteristics of the utterance content (hereinafter referred to as "features"). Specifically, the features are adverbs, adjectives, nouns, or auxiliary verbs contained in the utterance content. For example, adjectives may indicate positive or negative expressions, etc. Furthermore, nouns may be expressions indicating "action items" and "issues" such as "needs consideration" and "needs a response" in addition to positive or negative expressions, etc. Furthermore, auxiliary verbs, etc., may be expressions indicating hope, obligation, necessity, request, proposal, assertion, conjecture, hearsay, evaluation, etc. at the end of a sentence, etc. Therefore, when the classifier determines the type of utterance based on features, the conference support system can accurately determine the type of utterance.

[0134] Furthermore, the conference support system may allow the user to set the type of conference. For example, even if the content of a statement is the same, the nature may be different between a conference between on-site staff and a review conference. Specifically, in a conference between on-site staff, the nature of the content of a statement may be "request," whereas in a review conference, the nature of a statement by a decision maker may be "action item" or "decision." Such labels are closely related to the type of conference. Therefore, by setting the type of conference, the conference support system can accurately determine the type of statement.

[0135] Furthermore, the meeting support system may infer the role of a speaker based on the content of the remarks and use the inferred role as an attribute. For example, in a meeting such as a review meeting or a report session, a speaker who makes many comments that are "questions" regarding the report content is inferred to be a decision maker. In addition, a speaker who makes many comments that are "issues" is inferred to be an expert on the topic of the meeting.

[0136] If the results of such estimation are used as features for learning and judgment of the classifier, the conference support system can accurately determine the type of speech even if a decision maker or the like is not set in the conference support system.

[0137] The conference support system performs natural language processing using machine learning, for example, according to the method described in "The Fundamentals of Natural Language Processing," by Manabu Okumura, published by Corona Publishing.

[0138] FIG. 16 is a conceptual diagram showing an example of natural language processing using machine learning by a conference support system according to an embodiment of the third embodiment of the present invention. As shown in the figure, the conference support system learns a classifier CLF by supervised learning. First, data expressed as a set of feature-value pairs is input to the classifier CLF, and the classifier outputs a class to which the input test data TSD belongs. The class is indicated by a class label CLB. In other words, the class label CLB indicates the result of judging the type of utterance of the utterance content indicated by the test data TSD.

[0139] To train the classifier CLF using a machine learning algorithm, as shown in the figure, the conference support system inputs data and a set of pairs of classes to which the data belongs as training data TRD. Next, when the training data TRD is input, the conference support system trains the classifier CLF using a machine learning algorithm. After training the classifier CLF in this way, when test data TSD is input, the conference support system can output the class to which the test data TSD belongs using the classifier CLF. In this configuration, when information on words surrounding a word in the speech content to be judged is input, the conference support system can train a classifier that can output the meaning of the word.

[0140] The method of determining the type of speech using the classifier CLF allows the conference support system to determine the type of speech with greater accuracy than methods that use pattern matching, etc. Also, the method of determining the type of speech using the classifier CLF requires less effort, such as updating data, than methods that use pattern matching, etc. In other words, when the method of determining the type of speech using the classifier CLF is used, so-called maintainability can be improved.

[0141] Also, the system may be configured so that the participant can change the class label. In this way, if the class label CLB resulting from the determination by the classifier CLF indicates a class different from the class intended by the participant, the participant can change the class. For example, even if the content of a statement is determined to be an "opinion" by the conference support system, if the participant thinks that it is a "question," the participant performs an operation to change the class label from "opinion" to "question."

[0142] In this way, when a change is made and feedback is provided, the meeting support system can learn the classifier CLF in the same way as when training data TRD is input in machine learning. This makes it possible to train the classifier CLF in the meeting support system with less effort, such as updating data. Therefore, the meeting support system can accurately determine the type of speech.

[0143] Additionally, the conference support system may determine the type of statement based on the person who uttered the statement. For example, assume that in a conference, there is a person who always talks about "issues" and a person who always talks about "negative opinions". In such a case, the conference support system may identify the person who uttered the statement and determine the type of statement, such as "issues" or "negative opinions". Furthermore, information about the person who uttered the identified statement may be used as the background. In this way, the conference support system can determine the type of statement taking into account the tendencies of the person who uttered the statement. Therefore, the conference support system can accurately determine the type of statement.

[0144] Furthermore, the conference assistance system may determine the type of utterance based on the role in the conference of the person who uttered the utterance. For example, in a conference such as a report session, if the person who uttered the utterance is a "decision maker," the type of utterance may be determined based on the fact that the utterance was made by a "decision maker." In this way, the conference assistance system can determine the type of utterance taking into account the tendencies of the person who uttered the utterance. Therefore, the conference assistance system can accurately determine the type of utterance. In particular, the conference assistance system can accurately determine the types of utterances such as "action items" and "decisions."

[0145] Furthermore, the accuracy of the system often decreases as the number of types of utterances, i.e., the number of labels, increases. Therefore, the meeting support system may classify the types of utterances into groups. For example, a description will be given of an example in which the types of utterances are "suggestions," "questions," "answers," "positive opinions," "negative opinions," "neutral opinions," "information," "requests," "issues," "action items," and "decisions."

[0146] For example, the groups are referred to as "first group" and "second group" as shown below.

[0147] First group: "Suggestions", "Questions", "Answers", "Positive opinions", "Negative opinions", "Neutral opinions", "Information" and "Requests". Group 2: "Action items" and "Decisions" Then, the conference support system determines which of the utterance types in the "first group" the utterance content belongs to. Next, the conference support system determines which of the utterance types in the "second group" the utterance content belongs to. In the configuration shown in Fig. 16, a classifier CLF is prepared for each group.

[0148] For example, if someone says in a meeting, "I want something to be done between the people in charge," the meeting support system will determine that this comment is a type of "answer" in the first group and a type of "action item" in the second group.

[0149] For example, assume that the contents of utterances of the utterance types "action items" and "decisions" are important in a meeting. In such a case, if important utterance types are selected and grouped, as in the above-mentioned second group, the meeting support system can accurately determine the contents of important utterance types. Furthermore, since the number of utterance types in each group is smaller than when they are not grouped, the meeting support system can accurately determine the utterance types.

[0150] The conference support system may also display candidates for conclusions or decisions for the remarks. For example, the conference support system displays the remarks that are candidates for conclusions or decisions as "conclusion," "decision," "issue," or the like in the action item input text box 804 shown in FIG.

[0151] Furthermore, the conference support system may display a trash can icon or the like next to the action item input text box 804. If the user determines that one of the candidates such as "Conclusion", "Decision", or "Task" displayed by the conference support system in the action item input text box 804 is inappropriate, the user uses the trash can icon to delete the inappropriate candidate. In this way, the conference support system may delete the displayed candidates based on the user's operation. When deleting, the conference support system also modifies the labels of the training data. The conference support system can retrain the classifier using the modified data, i.e., the remarks and the labels, thereby enabling the classifier to learn better.

[0152] Furthermore, the user may perform an operation to add candidates such as "Conclusion," "Decision," or "Task" to be displayed in the action item input text box 804. In other words, the user may be able to add candidates that the conference support system has not listed to the action item input text box 804. For example, the conference support system displays a pull-down menu. Then, the user performs an operation to select a label to be added from the pull-down menu. The user may also perform an operation to modify the candidates displayed by the conference support system. Based on the results of these modifications, the labels of the training data are also modified. In this manner, the conference support system may retrain the classifier.

[0153] To perform the above operations, for example, a GUI such as the one shown below is used.

[0154] 17 is a diagram showing an example of a GUI displayed by the conference support system according to the third embodiment of the present invention. Hereinafter, an example will be described in which the conference support system displays candidates as shown in the figure in the action item input text box 804.

[0155] In this example, the conference support system displays a trash can icon INTR next to each candidate, as shown in the figure, and when the trash can icon INTR is clicked, the conference support system deletes the corresponding candidate.

[0156] In this example, the conference support system displays a pull-down menu PDMN as shown in the figure. When the pull-down menu PDMN is operated, the conference support system modifies or adds a label.

[0157] Using the above-described GUI or the like, the conference support system modifies, adds, or deletes candidates.

[0158] (Fourth embodiment) The fourth embodiment is an embodiment that is realized, for example, by the same overall configuration, hardware configuration, and functional configuration as the first embodiment. The following mainly describes the differences. The fourth embodiment is different from the first embodiment in the output by the output unit 1F4 (FIG. 3). A topic is a word that indicates the title, subject, or theme of a plurality of statements. For example, the conference support system displays topics generated as follows.

[0159] 18 is a diagram showing an example of display of topics generated by a conference support system according to an embodiment of the fourth embodiment of the present invention. In the example shown, the conference support system generates topics 1101, and arranges and displays corresponding comment contents 1102 for each topic 1101.

[0160] In many cases, the contents of statements made in a conference are made up of multiple topics 1101. Note that the contents of statements made in a conference may be made up of multiple topics 1101 or may be the same as the topic 1101.

[0161] As shown in the figure, the comment contents 1102 are grouped by topic 1101. The comment contents 1102 grouped by topic are then decorated according to the type of each comment content 1102. The example shown in the figure is an example in which, of the comment contents 1102, the comment contents 1102 of "positive opinions" are arranged on the left side, while the comment contents 1102 of "negative opinions" and "issues" are arranged on the right side. In this way, when the comment contents are arranged according to the type of comment content, the conference support system can display the content of the discussion in a way that makes it easy for the user to intuitively grasp it.

[0162] Furthermore, if the utterance content 1102 is long, the conference support system may shorten the utterance content 1102. For example, the conference support system may perform so-called sentence compression or the like. That is, the conference support system may convert the utterance content of each individual utterance into a simple format by using sentence compression or the like.

[0163] Specifically, the conference support system first performs a syntactic analysis of the utterance content using a syntactic analysis tool or the like. The syntactic analysis then divides the utterance content into multiple elements. The conference support system then assigns an importance level to each element. The conference support system then deletes elements with low importance based on the importance level. In this way, the conference support system can reduce the number of elements in a sentence and shorten the utterance content without destroying the sentence structure.

[0164] The importance is determined, for example, as follows:

[0165] · Nouns, proper nouns, subjects and objects have high importance.

[0166] Elements, adverbs, or attributive clauses that include "etc." or the particle "ya" to indicate examples are of low importance.

[0167] Therefore, the conference support system compresses sentences by deleting attributive clauses, embedded sentences, subordinate clauses, etc., to shorten the content of speech. Sentence compression is, for example, a method disclosed in "Fundamentals of Natural Language Processing" by Manabu Okumura, published by Corona Publishing.

[0168] In this way, when the topics are displayed by the conference support system, the user can easily review the conference. In other words, the conference support system displays the "tasks," "action items," and "decisions" associated with each topic, allowing the user to efficiently summarize the conference.

[0169] Also, for example, a user may add candidates for "issues," "action items," and "decisions" using the GUI shown in Fig. 17. In such a case, the meeting support system may extract utterances related to the added candidates. For example, the extraction results are displayed on a screen such as the one shown below.

[0170] 19 is a diagram showing an example of display of extraction results by the conference support system according to the fourth embodiment of the present invention. For example, in the illustrated example, the conference support system displays results extracted from "action items" and "decisions" in text boxes INBX, etc.

[0171] Also, in the text box INBX, the full sentence format may be displayed. Note that the figure shows an example in which the text box INBX displays the contents of comments, including comments that are candidates for "action items," grouped according to boundaries estimated by a method for estimating topic boundaries, which will be described later. That is, the illustrated text box INBX displays a summary. On the other hand, the full text AL may be displayed in the text box INBX. That is, the text box INBX may be switched to display the full text AL. Also, the summary is generated, for example, from the full text AL.

[0172] For example, the extraction result is generated by a fourth estimation method described later and a calculation of similarity. Specifically, the conference support system first expresses the utterance contents and each candidate grouped according to the boundary estimated by the method for estimating topic boundaries described later as a document vector. Next, the conference support system calculates the compatibility based on the cosine similarity or the like. Then, the conference support system extracts or links the utterance contents and candidates having high similarity. These methods are realized by, for example, the methods disclosed in "Basics and Technology of Natural Language Processing", by Yo Okuno et al., Shoeisha Publishing, and "http: / / www.cse.kyoto-su.ac.jp / ~g0846020 / keywords / cosinSimilarity.html".

[0173] As a result, as shown in the figure, a part of the minutes extracted from all the minutes is displayed in the text box INBX. Also, the text box INBX displays minutes consisting of multiple comments divided by topic.

[0174] Specifically, when there is an item in a discussion about a certain topic that is determined to be an "issue," "a decision," or an "action item," a statement indicating the content is extracted into the text box INBX.

[0175] As shown in the figure, the text box INBX may display comments indicating "issues," "decisions," or "action items" related to the comment indicating the contents of the extracted "action item."

[0176] The conference system is not limited to being used at the end of a conference. For example, the conference system may be used after the conference is over by inputting minutes and the like. In addition, it is preferable that the distance between topics is generated so as to reduce gaps and overlaps, for example, by the methods disclosed in "http: / / ryotakatoh.hatenablog.com / entry / 2015 / 10 / 29 / 015502" and "http: / / tech-sketch.jp / 2015 / 09 / topic-model.html".

[0177] The topic 1101 is generated, for example, as follows.

[0178] 20 is a flowchart showing an example of topic generation by the conference support system according to the fourth embodiment of the present invention. For example, when the process shown in the figure is performed, the conference support system can display a screen as shown in FIG.

[0179] In step S31, the conference support system inputs the contents of comments. For example, if a conference has already been held using the conference support system, the contents of comments made in the conference have been input to the conference support system.

[0180] In step S32, the conference support system estimates topic boundaries. Specifically, since a topic is generated for each group of comment contents, the conference support system first divides the comment contents into groups. Therefore, the conference support system estimates topic boundaries that serve as delimiters for dividing the comment contents into groups. For example, the conference support system estimates topic boundaries so that comment contents with common content or comment contents that can be estimated to be highly related to each other are grouped together.

[0181] Topic boundaries are estimated, for example, in the following four ways:

[0182] First, the first estimation method is an estimation method in which a topic boundary is estimated when a keyword indicating an explicit topic transition is detected. Specifically, keywords that can be estimated to indicate a topic transition are "By the way," "By the way," etc. Also, such keywords are words that are set in the meeting support system by a user or the like before the meeting, for example.

[0183] The second estimation method is an estimation method that estimates a topic boundary when "a word indicating a change of dialogue initiative is detected." Specifically, the second estimation method is an estimation method that estimates a topic boundary when words such as "just now," "the other day," or "long ago" appear. Furthermore, such words are, for example, words that are set in advance in the conference support system by a user or the like before the conference.

[0184] The first and second estimation methods are, for example, the methods disclosed in "Detecting Topic Transition in Information Seeking Chat," http: / / www.cl.cs.titech.ac.jp / _media / publication / 632.pdf.

[0185] Furthermore, the third estimation method is a method of estimating that a topic boundary exists when a statement of the type "question" appears. In other words, when a statement that is determined to be a "question" appears, the meeting support system estimates that the statement exists as a topic boundary.

[0186] Furthermore, the fourth estimation method is a method of estimating topic boundaries using "lexical cohesion." In the fourth estimation method, for example, "lexical cohesion" is calculated using a thesaurus or the like. Specifically, a chain consisting of a set of utterance contents that have a "lexical cohesion" relationship, that is, a so-called "lexical chain," is first detected. Then, in the fourth estimation method, "lexical cohesion" is considered to indicate a meaningful unity of utterance contents, and the starting point and the end point of the "lexical chain" are estimated to be topic boundaries. For example, the fourth estimation method is a method disclosed in "Text Segmentation Based on Lexical Cohesion," "https: / / ipsj.ixsq.nii.ac.jp / ej / index.php?action=pages_view_main&active_action=repository_action_common_download&item_id=49265&item_no=1&attribute_id=1&file_no=1&page_id=13&block_id=8," and "Fundamentals of Natural Language Processing," by Manabu Okumura, Corona Publishing Co., Ltd.

[0187] In step S33, the conference support system generates topics. That is, in step S33, the conference support system generates each topic for each topic boundary estimated in step S32. Therefore, step S33 is repeated the number of times corresponding to the number of groups separated by the topic boundaries.

[0188] Topics are inferred, for example, by the following two methods.

[0189] The first generation method is a method of extracting proper nouns or utterances not registered in a dictionary, so-called unknown words, from the utterances grouped by topic boundaries, and setting them as topics.

[0190] In the second generation method, first, nouns, verbs, or adjectives are extracted from the utterances grouped by topic boundaries. Next, in the second generation method, the conjugation of the extracted utterances is restored to its original form. Then, the utterances that appear frequently are set as topics. In this way, the conference support system can display topics, etc.

[0191] All or part of each process according to the present invention may be realized by a program written in a programming language such as assembler, C, C++, C#, Java (registered trademark), Javascript, Python (registered trademark), Ruby, and PHP, or an object-oriented programming language, for causing a computer to execute each procedure related to the conference support method. In other words, the program is a computer program for causing a computer such as an information processing device or a conference support system having one or more information processing devices to execute each procedure.

[0192] The program can be distributed by storing it in a computer-readable recording medium such as a ROM or an EEPROM (Electrically Erasable Programmable ROM). The recording medium may be an EPROM (Erasable Programmable ROM), a flash memory, a flexible disk, a CD-ROM, a CD-RW, a DVD-ROM, a DVD-RAM, a DVD-RW, a Blu-ray disk, an SD (registered trademark) card, or an MO. The program can be distributed through an electric communication line.

[0193] Furthermore, the conference support system may have two or more information processing devices connected to each other via a network, etc. In other words, the conference support system may perform all or part of various processes in a distributed manner, in parallel as in a duplex system, or in a redundant manner as in a dual system.

[0194] Although the preferred embodiments of the present invention have been described in detail above, the present invention is not limited to such specific embodiments, and various modifications and changes are possible within the scope of the gist of the present invention described in the claims. [Explanation of symbols]

[0195] 1. Meeting support system 10 Server 11 Terminal 1F1 Input section 1F2 Control section 1F21 Judgment section 1F22 Decoration Department 1F23 Time Management Department 1F24 Speaker Judgment Section 1F31 Cost Calculation Department 1F32 Evaluation Department 1F3 Storage section 1F4 Output section 1F5 Translation processing section CLF classifier [Prior art documents] [Patent documents]

[0196] [Patent Document 1] JP 2007-43493 A

Claims

1. A system having one or more information processing devices, an input unit for inputting the contents of statements made in a conference; a determination unit that determines whether the comment content input by the input unit is a candidate for an action item; an output unit that displays a screen including the comment content that is determined by the determination unit to be a candidate for an action item, the output unit displaying the screen allowing a person in charge of an action item to be set for the comment content; A conference support system comprising:

2. Estimating topic boundaries that separate groups of the content of the comments; generating a topic for each of the groups separated by the topic boundaries; 2. The conference support system according to claim 1, wherein the output unit displays the generated topics and the comments included in the groups corresponding to each topic.

3. The conference support system of claim 2, wherein the topic indicates either the title, topic, subject or compressed sentence of a plurality of remarks belonging to the group.

4. The output unit: The conference support system according to claim 1 , further comprising a display that allows a time limit to be set for the content of the speech.

5. The output unit: The conference support system according to claim 1 , further comprising a display that indicates that the content of the comment has been determined to be a candidate for the action item.

6. Further comprising a speaker determination unit for identifying a person who made the statement from among the participants of the conference, The conference support system according to claim 1 , wherein the output unit displays the content of the comment in association with the person who made the comment.

7. The output unit:

7. The conference support system according to claim 1, wherein only comment contents determined to be candidates for the action item are displayed from among the input comment contents.

8. A conference support system as described in any one of claims 1 to 7, further comprising an evaluation unit that evaluates the conference based on the content of the remarks made in the conference, and the output unit outputs the evaluation result by the evaluation unit.

9. The conference support system of claim 8, wherein the evaluation unit evaluates the conference based on the time the conference was held and the planned duration of the conference.

10. A conference support system as described in any one of claims 1 to 9, further comprising a cost calculation unit that calculates costs based on the labor costs of the participants in the conference, and the output unit further outputs the costs.

11. The output unit: outputting a communication diagram showing a communication intensity indicating the intensity of communication by the conference or email; The communication diagram shows communication by lines, The conference support system according to claim 1 , wherein pressing the line indicates a log of the communication.

12. A conference support system as described in any one of claims 1 to 11, which selects the sentence to be output based on the affiliation of the person who sent the sentence input by the input unit, the format in which the sentence was input, or a combination of these.

13. The determination unit determines the type of the comment content, The conference support system of claim 1 , wherein the output unit displays the comment content in a manner that allows a person in charge to be set for the comment content when the comment content is determined to be a candidate for an action item to be set among the types.

Images