Group activity device, group activity system, training device, training system and program

Abstract

To provide a group activity device, a group activity system, a training device, a training system, and a program capable of achieving diverse group activities.SOLUTION: A training server 1 to be a group activity device and a training device that performs role-playing type training by setting different roles to respective participants includes: a role setting processing part 13 for associating a training participant with a role; an utterance acquisition part 15 which acquires utterances of the participant from the terminal 5 the participant uses; an utterance output part 16 outputting the utterances acquired by the utterance acquisition part 15 to the terminal 5 of another participant; and a reply utterance output part 18 which sets an unset role to a virtual avatar, acquires a reply utterance based on the unset role with respect to the utterance by using a knowledge server 3 having knowledge corresponding to respective roles in the training, and outputs it to the terminal 5 of each participant when the object of the utterance acquired by the utterance acquisition part 15 is an unset role with which no participant is associated.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a group activity device, a group activity system, a training device, a training system, and a program. [Background technology]

[0002] Traditionally, there are group trainings where participants gather together, and group trainings include lecture-style training where the instructor speaks one-sidedly, and interactive training where participants do role-playing and group discussions. In such group trainings, the training content, date and time, location, and other training conditions are disclosed, and participants apply after confirming the training conditions. In the above-mentioned training, a training mediation device capable of holding training under suitable conditions for trainees and instructors has been disclosed (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7325786 Summary of the Invention [Problem to be solved by the invention]

[0004] The training mediation device described in Patent Document 1 allows multiple applicants to take one training course at the same time, collects the desired conditions for taking the training course from the multiple applicants, and determines the training content, location, date and time, etc. of the training by referring to the desired conditions of the multiple applicants and the available training conditions of the instructor. Therefore, the training is conducted when the conditions of the applicants and the instructor are met. If this training mediation device is applied to role-playing type training in which different roles are set for each participant, the training is conducted only when the number of applicants is reached.

[0005] On the other hand, nowadays, even group training where participants gather together can be conducted in an online environment without actually meeting in person. Unlike face-to-face training, online training has the advantage that it can be conducted anywhere, making it easier for people who want to participate. However, with role-playing-style training in which each participant is assigned a different role, there is still the problem that meaningful training cannot be conducted unless there is a sufficient number of people who want to take the course. Furthermore, in any group conversation activity, not just training sessions, it is possible to give participants a variety of insights by making efforts to add diversity to the group.

[0006] SUMMARY OF THE PRESENT EMBODIMENTS Accordingly, an object of the present invention is to provide a group activity device, a group activity system, a training device, a training system, and a program that enable diverse group activities to be carried out. [Means for solving the problem]

[0007] The present invention solves the above problems by the following solving means. The first invention is a group activity device for supporting group activities between one or more participants and a virtual avatar, comprising: a profile setting means for setting a profile for the virtual avatar; a comment acquisition means for acquiring comments from the participant from a terminal used by the participant; a comment output means for outputting the comments acquired by the comment acquisition means to the terminals of other participants; and, when the comment acquired by the comment acquisition means is directed to the virtual avatar, a reply comment to the comment based on the profile set by the profile setting means, and outputting it to the terminal of each participant. A second invention is a group activity device according to the first invention, wherein the group activity is a role-playing type training in which a different role is assigned to each participant, and the group activity device is equipped with a knowledge unit having knowledge corresponding to each role in the training, and a role setting means for matching the participants of the training with the roles, the profile setting means sets an unset role to which the participant is not associated as the profile to the virtual avatar, and the reply comment output means, when the target of the comment acquired by the comment acquisition means is directed to the role to which the virtual avatar is associated, acquires a reply comment based on the role in response to the comment using the knowledge unit and outputs the reply comment to the terminal of each participant. A third invention is a group activity device according to the second invention, wherein the knowledge unit is a language model that has learned a plurality of documents, and comprises a document memory unit that stores documents having information corresponding to each role, and a related document acquisition means that acquires related text from the document related to the comment from the document memory unit when the comment acquired by the comment acquisition means is made to the role associated with the virtual avatar by the role setting means, and the reply comment output means outputs the reply comment output by the knowledge unit by inputting the comment into the knowledge unit together with the text acquired by the related document acquisition means to the terminal of each participant. A fourth invention is a group activity device according to the second invention, wherein the knowledge unit is a character model that has learned a plurality of documents relating to the character, and has a character model corresponding to each character, and the reply statement output means outputs to the terminal of each participant the reply statement output by the character model in the knowledge unit that corresponds to the character associated with the virtual avatar in response to inputting the statement into the character model. A fifth invention is a group activity device according to the second invention, wherein the knowledge unit is a dialogue information storage unit that stores a plurality of pieces of dialogue information consisting of questions and replies related to the role, and the reply statement output means searches the knowledge unit to obtain a question similar to the statement, and outputs the reply of the dialogue information corresponding to the obtained question as the reply statement to the terminal of each participant. A sixth invention is a group activity device according to any one of the second to fifth inventions, comprising an image output means for arranging avatar images of the participants, including avatar images of the virtual avatar, in a virtual space and outputting to the terminal of each participant a training space image in which role names indicating the roles of each participant are arranged near the avatar images. A seventh invention is a group activity device according to the sixth invention, wherein the statement acquisition means accepts the statement as voice data, and the reply statement output means acquires the reply statement to the statement using the knowledge unit after converting the voice data into text, and converts the acquired reply statement into voice data and outputs it to the terminal of each participant. An eighth invention is a group activity device according to any one of the second to seventh inventions, comprising a training information storage unit which stores each role in correspondence with a training content, and a participation acceptance means which accepts a participation request for one of the training sessions from the terminal of the participant, wherein the role setting means randomly determines the role for the participant corresponding to the participation request accepted by the participation acceptance means by referring to the training information storage unit, associates the participant with the determined role, and comprises a role output means which outputs information on each role set by the role setting means and the corresponding participant to the terminal of each participant. A ninth invention is a group activity device according to any one of the second to seventh inventions, comprising a training information storage unit which stores each role and a priority in correspondence with a training content, and a participation acceptance means which accepts a participation request for one of the training sessions from the terminal of the participant, wherein the role setting means refers to the training information storage unit to determine the role for the participant corresponding to the participation request accepted by the participation acceptance means in order of highest priority, associates the participant with the determined role, and comprises a role output means which outputs information on the participant corresponding to each role set by the role setting means and each role set by the role setting means to the terminal of each participant. A tenth invention is a group activity device according to the first invention, further comprising: a conversation storage means for storing conversation information including the utterance acquired by the utterance acquisition means and the reply utterance output by the reply utterance output means; and a reply instruction generation means for generating a reply creation instruction including the utterance, the conversation information accumulated by the conversation storage means, and the profile set by the profile setting means when the target of the utterance acquired by the utterance acquisition means is directed to the virtual avatar, wherein the reply utterance output means inputs the reply creation instruction generated by the reply instruction generation means to a sentence generation AI which generates a sentence based on the creation instruction, thereby outputting the reply output by the sentence generation AI as the reply utterance to the terminal of each participant. An eleventh invention is a group activity device according to the tenth invention, further comprising a creation instruction receiving means for receiving a profile creation instruction including attribute items related to the personality of the virtual avatar, and a profile generation means for obtaining a profile output by the sentence generation AI by inputting the profile creation instruction received by the creation instruction receiving means to the sentence generation AI, and the profile setting means sets the profile generated by the profile generation means for the virtual avatar. A twelfth invention is a group activity device according to the eleventh invention, wherein the creation instruction receiving means acquires external data corresponding to the purpose of the group activity from an external server communicatively connected to the group activity device, and receives the profile creation instruction to which the acquired external data is further added. A thirteenth invention is a group activity device according to the eleventh or twelfth invention, further comprising a profile output means for outputting the profile generated by the profile generation means, and the creation instruction receiving means receives an instruction to create the profile edited based on the profile output by the profile output means. A fourteenth invention is a group activity device according to any one of the eleventh to thirteenth inventions, further comprising a template output means for outputting a template for creating the profile creation instruction, and the creation instruction receiving means receives the profile creation instruction created using the template output by the template output means. A fifteenth invention is a group activity device according to the tenth invention, further comprising an attribute item information storage unit which stores attribute item information in which a plurality of item values ​​are associated with attribute items relating to the personality of the virtual avatar, and a profile receiving means which outputs the attribute item information stored in the attribute item information storage unit so as to be referenceable and receives combinations of the attribute items and the item values, and the profile setting means sets the combination of the attribute items and the item values ​​received by the profile receiving means as the profile for the virtual avatar. A sixteenth aspect of the present invention is a group activity system including the group activity device according to any one of the tenth to fifteenth aspects of the present invention and the sentence generation AI. A seventeenth invention is a training device for carrying out role-playing type training in which a different role is assigned to each participant, the training device comprising: a role setting means for matching participants of the training with the roles; a statement acquisition means for acquiring utterances of the participant from a terminal used by the participant; a statement output means for outputting the utterance acquired by the statement acquisition means to the terminals of other participants; and a reply statement output means for, when the target of the utterance acquired by the statement acquisition means is an unset role to which the participant has not been associated by the role setting means, acquiring a reply utterance based on the unset role in response to the utterance using a knowledge unit communicatively connected to the training device and having knowledge corresponding to each role in the training, and outputting the reply utterance to the terminal of each participant. An eighteenth aspect of the present invention is a training system including the training device and the knowledge section of the seventeenth aspect of the present invention. The 19th invention is a program for causing a computer to function as a profile setting means for setting a profile for a virtual avatar that engages in group activities with one or more participants, a comment acquisition means for acquiring comments from the participant from a terminal used by the participant, a comment output means for outputting the comments acquired by the comment acquisition means to the terminals of other participants, and a reply comment output means for, when the comment acquired by the comment acquisition means is directed at the virtual avatar, acquiring a reply comment to the comment based on the profile set by the profile setting means and outputting it to the terminal of each participant. Effect of the Invention

[0008] According to the present invention, it is possible to provide a group activity device, a group activity system, a training device, a training system, and a program that enable diverse group activities to be carried out. [Brief description of the drawings]

[0009] [Figure 1] 1 is an overall configuration diagram of a training system according to a first embodiment and a functional block diagram of a training server. [Diagram 2]4 is a diagram illustrating an example of a training information storage unit of a training server according to the first embodiment; FIG. [Diagram 3] 10 is a flowchart showing a training reception process of the training server according to the first embodiment. [Figure 4] FIG. 11 is a diagram showing an example of determining a participant corresponding to a winning role in the training server according to the first embodiment. [Diagram 5] 5 is a diagram showing an example of a screen displayed on a terminal 5 according to the first embodiment. FIG. [Figure 6] 5 is a flowchart showing training processing of the training server according to the first embodiment. [Figure 7] FIG. 11 is a diagram for explaining a model statement process of the training server according to the first embodiment. [Figure 8] FIG. 11 is an overall configuration diagram of a training system according to a second embodiment and a functional block diagram of a training server. [Figure 9] FIG. 11 is a diagram for explaining a model statement process of the training server according to the second embodiment. [Figure 10] 13 shows an example of participants taking part in a discussion held in a group activity system according to the third embodiment. [Figure 11] 13 is an overall configuration diagram of a group activity system according to a third embodiment and a functional block diagram of a group activity server. FIG. [Figure 12] 13 is a flowchart showing a member reception process of a group activity server according to the third embodiment. [Figure 13] FIG. 13 is a diagram showing an example of a profile creation instruction according to the third embodiment. [Figure 14] FIG. 13 is a diagram showing an example of a generated profile according to the third embodiment. [Figure 15] 13 is a flowchart showing a discussion process of a group activity server according to the third embodiment. [Figure 16] 13 is a flowchart showing a virtual avatar statement process of the group activity server according to the third embodiment. [Figure 17] 13A to 13C are diagrams showing examples of reply writing instructions and reply examples according to the third embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Note that this is merely an example, and the technical scope of the present invention is not limited to this example. (First embodiment) <Overall configuration of training system 100> FIG. 1 is an overall configuration diagram of a training system 100 according to the first embodiment and a functional block diagram of a training server 1. As shown in FIG. FIG. 2 is a diagram showing an example of the training information storage unit 23 of the training server 1 according to the first embodiment.

[0011] 1, the training system 100 includes a training server 1 (training device, group activity device), a knowledge server 3 (knowledge section), and a terminal 5 of each participant participating in the training. The training server 1, the knowledge server 3, and the terminal 5 of each participant are connected via a communication network N. Training system 100 is a system for conducting role-playing type training in which training server 1 assigns different characters to each participant, and in the event that there are not enough participants to assign roles, the unassigned characters are assigned to virtual avatars and training server 1 performs the training instead. In addition, in training system 100, participants join from terminals 5 and conduct training. In this way, training system 100 is capable of conducting training appropriately in an online environment even when the number of participants does not reach the required number.

[0012] <Training Server 1> The training server 1 is a device that accepts participation requests from participants, assigns different roles to each participant, and conducts role-playing type training. The training server 1 is, for example, a server, and performs processing with the terminal 5 through communication. The training server 1 may be, for example, a general personal computer (PC) or the like. The training server 1 includes a control unit 10, a storage unit 20, and a communication interface unit 29. The control unit 10 is a CPU (Central Processing Unit) that controls the entire training server 1. The control unit 10 appropriately reads and executes an OS (Operating System) and various application programs stored in the storage unit 20, thereby cooperating with the above-mentioned hardware and executing various functions.

[0013] Before describing the control unit 10, the storage unit 20 will be described. The storage unit 20 is a storage device such as a hard disk or a semiconductor memory device for storing programs, data, etc. required for the operation of the training server 1. The storage unit 20 includes a program storage unit 21, a document storage unit 22, and a training information storage unit .

[0014] The program storage unit 21 is a storage area for storing programs. The program storage unit 21 stores a training program 21a (program). The training program 21a is, for example, a program for executing various functions of the control unit 10 described below. Note that the various functions of the control unit 10 may be executed by one training program 21a, or may be executed by multiple programs for each function. The document storage unit 22 is a storage area for storing a plurality of documents having information corresponding to the training roles. The documents stored in the document storage unit 22 are, for example, documents describing background knowledge of the roles. The document storage unit 22 stores documents for each training role, for example.

[0015] The training information storage unit 23 is a storage area for storing training information. FIG. 2 shows an example of the training information storage unit 23. The training information storage unit 23 shown in FIG. 2 is an example in which the training outline shown in FIG. 2(A) and the scenario information shown in FIG. 2(B) are stored. The training summary in FIG. 2(A) stores training ID (IDentification), training content, and role in association with each other. The training ID is identification information that uniquely identifies the training. The training content is a document that briefly describes the training content and also includes information on the number of people required to conduct the training. A role is a role that a participant plays in the training, and one role corresponds to one participant (including a computer NPC (Non Player Character).) Here, an NPC refers to a character that is not operated by a human, and is controlled by a computer (CPU) on behalf of the participant. 2B stores a scenario to be presented to each participant. The scenario is a topic for the participants to discuss, and may include hints to help the discussion. The training content is merely an example, and other content may be used as long as it involves role-playing training. The number of people required is two or more.

[0016] Next, the control unit 10 will be described. The control unit 10 shown in Figure 1 includes a training overview presentation unit 11, a participation acceptance unit 12 (participation acceptance means), a character setting processing unit 13 (character setting means, profile setting means), an image output unit 14 (image output means, character output means), a comment acquisition unit 15 (comment acquisition means), a comment output unit 16 (comment output means), a related document acquisition unit 17 (related document acquisition means), and a reply comment output unit 18 (reply comment output means). When a connection is made from the terminal 5, the training outline presenting unit 11 outputs to the terminal 5 a screen (not shown) that presents an outline of the training and enables the terminal 5 to express its intention to participate.

[0017] The participation reception unit 12 receives an intention to participate in a training course (participation request) from the terminal 5. The participation request includes information about the participant, including the participant's name (or nickname, etc.). The character setting processing unit 13 randomly determines characters for participants accepted by the participation accepting unit 12 by referring to the training information storage unit 23. Then, the character setting processing unit 13 associates the participants with the determined characters and stores them in the storage unit 20. Note that, if the number of participants is less than the required number of people for the training, the character setting processing unit 13 associates an NPC (virtual avatar) with a character not determined by a participant. The image output unit 14, for example, places avatar images of participants, including avatar images corresponding to unset roles to which virtual avatars have been set, in a virtual space simulating a training venue, and outputs to each participant's terminal 5 a training space image in which the role names of each participant are placed near the avatar images.

[0018] The comment acquisition unit 15 acquires comments from participants from the terminal 5. The comment acquisition unit 15 may acquire the comments from participants as text data or as voice data. The utterance output unit 16 outputs the utterances acquired by the utterance acquisition unit 15 to the terminals 5 of the other participants attending the training. If the utterances were acquired as text data, the utterance output unit 16 can output the text data as is to the terminals 5 of the other participants. Furthermore, if the utterances were acquired as voice data, the utterance output unit 16 may output the voice data as is, or may convert the voice data into text data and then output it.

[0019] When the target of a utterance acquired by the utterance acquisition unit 15 is an unset role to which no participant is associated, i.e., a role to which a virtual avatar is associated, in other words, when the target of the utterance is an NPC (computer), the related document acquisition unit 17 acquires related text, which is a related sentence, from the document related to the utterance, from the document memory unit 22. The reply comment output unit 18 outputs the reply comment received from the knowledge server 3 to each participant's terminal 5 by transmitting the comment to the knowledge server 3 together with the related text acquired by the related document acquisition unit 17. If the comment acquired by the comment acquisition unit 15 is voice data, the reply comment output unit 18 converts the voice data into text and then transmits the text data of the comment together with the related text to the knowledge server 3, and converts the reply comment that is text data received from the knowledge server 3 into voice data and outputs it to each participant's terminal 5. Each of these functions will be described in detail later.

[0020] The communication interface unit 29 is an interface for communicating between the knowledge server 3 and the terminals 5 of the participants via the communication network N. It should be noted that a computer refers to an information processing device equipped with a control unit, a storage device, etc., and the training server 1 is an information processing device equipped with a control unit 10, a storage unit 20, etc., and is included in the concept of a computer.

[0021] Furthermore, there is no limit to the number of pieces of hardware that make up the training server 1. It may be configured with one or more pieces as necessary. Furthermore, the hardware of the training server 1 may be configured to include various servers such as a web server, a database server, and an application server as necessary, and may be configured with one server or with separate servers. Furthermore, the training server 1 may be, for example, a cloud.

[0022] <Knowledge Server 3> Returning to FIG. 1, the knowledge server 3 is a server equipped with a language model (knowledge unit) that has learned multiple documents. The language model is a deep learning model that is a huge neural network that learns language patterns from a huge amount of text data and enables tasks related to text, such as text generation and summarization, to be performed with high accuracy, and is, for example, a large-scale language model (LLM). The language model may also be a generative AI. The knowledge server 3 may also be, for example, a cloud.

[0023] <Terminal 5> The terminal 5 is a computer used by participants in the training. The terminal 5 can be configured, for example, as a PC or a tablet terminal. Although not shown, the terminal 5 includes a control unit, a storage unit, a display unit, an input unit, a communication interface unit, etc. The communication network N is a network for communication between the training server 1, the knowledge server 3, and the terminal 5. The communication network N is, for example, a communication network such as the Internet, and may be wired or wireless.

[0024] <Processing Description> Next, the process performed by the training server 1 will be described. FIG. 3 is a flowchart showing the training reception process of the training server 1 according to the first embodiment. FIG. 4 is a diagram showing an example of determining participants corresponding to roles in the training server 1 according to the first embodiment. FIG. 5 is a diagram showing an example of a screen displayed on the terminal 5 according to the first embodiment.

[0025] 3, the control unit 10 of the training server 1 starts accepting applications to participate in the training. More specifically, in response to a connection from the terminal 5, the training outline presentation unit 11 outputs to the terminal 5 a screen (not shown) that presents an outline of the training and enables the participant to declare their intention to participate. The user of the terminal 5 checks the overview of the training by looking at the screen, and if he / she wants to participate, he / she selects a button for participation, etc. By doing so, the control unit of the terminal 5 transmits a participation request to the training server 1, and the participation reception unit 12 receives the participation request from the terminal 5 in S12.

[0026] In S13, for example, after a certain period of time (for example, an application period of 10 days or the like) has elapsed since the start of the acceptance of applications for participation in the training, the control unit 10 closes the acceptance of applications for participation in the training. In S14, the character setting processing unit 13 refers to the training information storage unit 23 and sets each character for each participant. The character setting processing unit 13 randomly determines characters for the participants accepted by the process of S12, for example. Then, the character setting processing unit 13 stores the participants and the determined characters in the storage unit 20 in association with each other.

[0027] In S15, the role setting processing unit 13 judges whether or not there is a role to which the participant has not been set. If there is a role to which the participant has not been set (S15: YES), the control unit 10 shifts the process to S16. On the other hand, if there is no role to which the participant has not been set (S15: NO), the control unit 10 shifts the process to S17. In S16, the role setting processing unit 13 sets an NPC (computer) to a role that has not been set.

[0028] Here, a specific example of the role setting will be described with reference to FIG. FIG. 4(A) is an explanatory diagram 41 of role setting when there is one vacancy for the required number of people. In this example, the roles required for the training are four different roles: security officer, business department manager, information system officer, and public relations officer. When three participants, A, B, and C, are accepted, the role setting processing unit 13 randomly determines the role of each participant. As a result, for example, A is set as the security officer, B as the business department manager, and C as the information system officer. At this point, no participant has been set for the public relations officer role. Therefore, the role setting processing unit 13 sets an NPC for the public relations officer role.

[0029] FIG. 4(B) is an explanatory diagram 42 of role setting when there are two vacancies compared to the required number of people. The roles required for the training in this example are the same as those in FIG. 4(A) above. When two participants, D and E, are accepted, the role setting processing unit 13 randomly determines the role of each participant. As a result, for example, D is set as the security officer and E is set as the information system officer. At this point, no participants have been set for the roles of business division manager and public relations officer. Therefore, the role setting processing unit 13 sets NPCs for the roles of business division manager and public relations officer.

[0030] In S17 of FIG. 3, image output unit 14 outputs an avatar image together with the set role to terminal 5 of each participant. FIG. 5 shows an example of a training space image 51 output to the terminal 5. The training space image 51 is an image in which the avatar images of the participants (including NPCs) are arranged in a state where they can hold a discussion. The training space image 51 includes a scenario output area 52 and avatar images 53 to 56. The scenario output area 52 is an area for outputting a training scenario. The scenario output area 52 appropriately displays the scenario information in the training information storage unit 23.

[0031] Avatar images 53 to 56 are avatar images of participants playing various roles. Image output unit 14 displays the avatar images together with the role names near the avatar images. This makes it easier for participants to confirm which avatar image represents which role. In this example, avatar image 53 is for the security officer role, avatar image 54 is for the business division manager role, avatar image 55 is for the information system officer role, and avatar image 56 is for the public relations officer role.

[0032] When requesting participation, the participant may select a preferred avatar image from an avatar image selection screen (not shown), and the image output unit 14 may output the avatar image selected by the participant. The image output unit 14 may add an indication that the participant is the participant to the avatar image of the role of each participant and output the avatar image to the terminal 5 of the participant. In this case, the image output unit 14 outputs a training space image 51 with a different display for each participant to the terminal 5 of each participant. This allows the participants to participate in the training while always being aware of which role they are playing.

[0033] Thereafter, the control unit 10 ends this process and starts the training process. FIG. 6 is a flowchart showing the training process of the training server 1 according to the first embodiment. FIG. 7 is a diagram for explaining the model statement process of the training server according to the first embodiment. 6, the control unit 10 outputs the title of the training scenario. More specifically, the control unit 10 outputs the title to the scenario output area 52 of the training space image 51. This allows each participant to check the title and progress through the training by making comments in accordance with the title.

[0034] In S22, the comment acquisition unit 15 judges whether or not a comment from a participant has been acquired from the terminal 5. If a comment from a participant has been acquired from the terminal 5 (S22: YES), the control unit 10 shifts the process to S23. On the other hand, if a comment from a participant has not been acquired from the terminal 5 (S22: NO), the control unit 10 shifts the process to S26. In S23, the comment output unit 16 outputs the acquired comment to the other terminals 5. In this way, the other participants can understand the content of the comment made by the participant.

[0035] In S24, the control unit 10 judges whether the acquired comment is directed to an NPC (computer). If the acquired comment is directed to an NPC (S24: YES), the control unit 10 shifts the process to S25. On the other hand, if the acquired comment is not directed to an NPC (S24: NO), the control unit 10 shifts the process to S26. In S25, the control unit 10 performs a model statement process.

[0036] Here, the model utterance processing will be explained with reference to FIG. For example, in the case where the public relations officer is an NPC and the statement from person A acquired by the statement acquisition unit 15 is something like "What do you think, public relations officer?", when the control unit 10 acquires the statement in S31, it determines that the acquired statement is directed to the NPC. The control unit 10 interprets the content of the acquired statement using, for example, known natural language processing, and then determines whether the acquired statement is directed to the NPC or not.

[0037] If it is determined that the target of the utterance is an NPC, then in S32 the related document acquisition unit 17 searches for related text related to the utterance using the document storage unit 22, and in S33 the related document acquisition unit 17 obtains the related text as a search result. The related document acquisition unit 17 searches for related text related to the utterance from documents of the relevant character (in this example, the public relations officer) in the document storage unit 22, using words including related words of the words included in the utterance, obtained by referring to a related word storage unit (not shown).

[0038] In S34, the reply comment output unit 18 transmits the comment together with the acquired related text to the knowledge server 3. In this way, the related text and the comment are input to the knowledge unit 30, which is a language model, in the knowledge server 3. Then, the knowledge server 3 performs output in response to the input. In S35, the reply comment output unit 18 receives the answer output from the knowledge server 3, and in S36, the reply comment output unit 18 outputs the received answer as a reply comment to the terminals 5 of each participant including Mr. A.

[0039] In S26 of Fig. 6, the control unit 10 judges whether or not to output the answer to the topic. The control unit 10 may judge whether or not to output the answer to the topic when a predetermined time (for example, the end time of the discussion, such as 10 minutes) has elapsed, or may judge in response to receiving data indicating the end from the participant's terminal 5. If the answer to the topic is to be output (S26: YES), the control unit 10 transfers the process to S27. On the other hand, if the answer to the topic is not to be output (S26: NO), the control unit 10 transfers the process to S22. In S27, the control unit 10 outputs the answer to the question to each terminal 5, and ends this process. Note that outputting the answer to the question at the end of the training is just one example, and the training may be ended when another end condition is met.

[0040] Thus, according to the first embodiment, the training server 1 has the following advantages. (1) Participants in the training are associated with roles, and the participant's utterances are acquired from the terminal 5 used by the participant. The acquired utterances are output to the terminals 5 of the other participants. If the acquired utterance is directed to an unset role to which the participant is not associated, i.e., a role to which a virtual avatar is associated, a reply utterance based on the unset role in response to the utterance is acquired using a knowledge server 3 having knowledge corresponding to each role in the training, and output to the terminal 5 of each participant. Therefore, even if a participant's comment is directed to an unassigned role that is not associated with a participant in the training, a reply comment to the comment can be obtained using the knowledge server 3 and output to each participant's terminal 5. As a result, training can be held even if the number of participants does not reach the required number. Moreover, since a reply comment based on an unassigned role to a comment is obtained using the knowledge server 3 that has knowledge corresponding to each role in the training, training can be appropriately conducted in an online environment.

[0041] (2) The knowledge server 3 is a language model that has learned multiple documents, and when the acquired utterance is made to an unset role with no participant associated therewith, the knowledge server 3 acquires relevant text from a document related to the utterance from the document memory unit 22, which stores documents having information corresponding to each role, and outputs a reply utterance output by the knowledge server 3 in response to the utterance being input to the knowledge server 3 together with the acquired text to the terminal 5 of each participant. Therefore, a reply message corresponding to a message can be obtained by using the language model. Moreover, a related text from a document related to a message is obtained from the document storage unit 22 that stores documents having information corresponding to each role, and is input to the language model, so that a reply message corresponding to the role can be obtained.

[0042] (3) Avatar images of the participants, including avatar images corresponding to roles that have not yet been set, are placed in the virtual space, and a training space image in which the role names indicating the roles of each participant are placed near the avatar images is output to each participant's terminal 5. Therefore, training can be held in a virtual space, and even if no participants are set and roles are played by NPCs instead, they will be represented by avatar images in the same way as roles for which participants are set, making it possible for each participant to appear as if they were actually present.

[0043] (4) The participant's utterances are accepted as voice data from the terminal 5 used by the participant, the voice data is converted into text, and then a reply utterance to the utterance is obtained using the knowledge server 3. The obtained reply utterances are converted into voice data and output to the terminal 5 of each participant. Therefore, even if the character is an NPC, they can respond to a statement with voice data.

[0044] (5) A request to participate in a training session is accepted from the participant's terminal 5, and a role is randomly determined for the participant corresponding to the accepted participation request by referring to the training information memory unit 23 which stores each role in correspondence with the training content, and the participant is associated with the determined role. Information on each role and the corresponding participant is output to each participant's terminal 5. Therefore, when a user requests to participate in a training session, the roles of the participants are randomly determined, so that the roles can be determined fairly. In addition, the determined roles are output to the terminal 5 of the participant, so that the user can know the role he or she has been assigned and the roles of the other participants.

[0045] Second embodiment In the second embodiment, a model (knowledge section) of each role is created in advance, and when a participant is not associated with an unset role (virtual avatar), the model of the role is used. Also, a training program is performed by setting a role with a high priority for the participant. In the following description, the same reference numerals or the same reference numerals at the end are used for parts that perform the same functions as those in the first embodiment, and duplicated descriptions are omitted as appropriate.

[0046] <Overall configuration of training system 200> FIG. 8 is an overall configuration diagram of a training system 200 and a functional block diagram of a training server 201 according to the second embodiment. The training system 200 shown in FIG. 8 includes a training server 201 and a terminal 5 of each participant.

[0047] <Training Server 201> The training server 201 includes a control unit 210 , a storage unit 220 , and a communication interface unit 29 . Before describing the control unit 210, the storage unit 220 will be described. The storage unit 220 includes a program storage unit 221 , a document storage unit 22 , a training information storage unit 223 , and a knowledge unit 230 . A training program 221a (program) is stored in the program storage unit 221. The training program 221a is, for example, a program for executing various functions of the control unit 210, which will be described next.

[0048] The training information storage unit 223 is a storage area for storing training information. The training information storage unit 223 stores a training summary and scenario information. The training summary stored in the training information storage unit 223 stores the training content, the role, and the priority of the role in association with the training ID. The priority of roles is the priority of each role according to the content of the training, with higher priorities being set for important roles. The knowledge section 230 is a storage area for storing a role model that has learned knowledge for each role.

[0049] Next, the control unit 210 will be described. The control unit 210 includes a training overview presentation unit 11, a participation acceptance unit 12, a character setting processing unit 213 (character setting means), an image output unit 14, a comment acquisition unit 15, a comment output unit 16, a reply comment output unit 218 (reply comment output means), and a model learning unit 219. The character setting processing unit 213 determines characters in the order of priority for participants accepted by the participation accepting unit 12 by referring to the training information storage unit 223. Then, the character setting processing unit 213 associates the participants with the determined characters and stores them in the storage unit 220. Note that, when the number of participants is less than the required number of participants for the training, the character setting processing unit 213 associates an NPC (computer) with a character not determined by a participant.

[0050] When the target of the utterance acquired by the utterance acquisition unit 15 is made to an unset character with which no participant is associated, that is, when the target of the utterance is an NPC, the reply utterance output unit 218 outputs a reply utterance output by the character model in the knowledge unit 230 corresponding to the NPC character by inputting the utterance to the character model, to each participant's terminal 5. When the utterance acquired by the utterance acquisition unit 15 is voice data, the reply utterance output unit 218 converts the voice data into text, inputs the utterance to the character model in the knowledge unit 230 corresponding to the NPC character, converts the reply utterance output by the character model into voice data, and outputs it to each participant's terminal 5. The model learning unit 219 uses a plurality of documents related to each role stored in the document storage unit 22 for each training ID stored in the training information storage unit 223 to learn a role model.

[0051] <Processing Description> Next, the processing performed by the training server 201 will be described. The processing flow for the training reception process is the same as that of the first embodiment (FIG. 3), except for the process corresponding to S14 in FIG. 3, the role setting processing unit 213 determines roles for the participants accepted in the process of S12 in order of priority by referring to the training information storage unit 223. Then, the role setting processing unit 213 stores the determined roles in the storage unit 220 in association with the participants.

[0052] Next, the processing flow for the training process is the same as that of the first embodiment (FIG. 6), except for the process corresponding to S25 in FIG. Here, the model utterance process corresponding to S25 in FIG. 6 will be described with reference to FIG. First, the character model to be used in the model utterance process is created in advance. FIG. 9(A) is a diagram for explaining the learning of the role model. 9(A), the model learning unit 219 learns a role model for each role using a plurality of documents related to the role stored in the document storage unit 22. Then, the control unit 210 stores the learned role model in the knowledge unit 230.

[0053] Next, the model comment process will be described. For example, a case where the business division manager and the public relations officer are NPCs will be described with reference to FIG. 9(B). In S231, the comment acquisition unit 15 acquires a comment from D, and if the target of the acquired comment is directed to the business division manager, in S232, the control unit 210 determines that the target of the comment is directed to an NPC, and the reply comment output unit 218 queries the character model of the business division manager stored in the knowledge unit 230 about the acquired comment. In this way, the comment is input to the character model of the business division manager. Then, the character model performs output in response to the input. In S233, the reply statement output unit 218 obtains an answer output from the role model of the business division manager, and in S234, the reply statement output unit 218 outputs the obtained answer as a reply statement to the terminal 5 of each participant including Mr. D.

[0054] When D, who has received the reply comment from the business division manager, makes a further comment, the comment acquisition unit 15 acquires a comment from D in S235, and if the acquired comment is directed at the public relations officer, the control unit 210 determines in S236 that the comment is directed at an NPC, and the reply comment output unit 218 queries the character model of the public relations officer stored in the knowledge unit 230 about the acquired comment. In this way, the comment is input to the character model of the public relations officer. The character model then performs output in response to the input. In S237, the reply statement output unit 218 obtains an answer output from the character model of the public relations officer, and in S238, the reply statement output unit 218 outputs the obtained answer as a reply statement to the terminals 5 of each participant including Mr. D.

[0055] As described above, according to the second embodiment, the training server 201 has the following advantages. (1) The knowledge section 230 is a character model that has learned multiple documents related to characters, and has a character model corresponding to each character. When a comment is input to a character model corresponding to an unset character stored in the knowledge section 230, the character model for that character outputs a reply comment to the terminal 5 of each participant. Therefore, a character model is trained in advance based on multiple documents related to characters, and when a comment is made to an NPC, a reply comment to the comment can be obtained using the character model. As a result, a reply comment corresponding to an unspecified character can be obtained for a comment, allowing training to be conducted appropriately in an online environment.

[0056] (2) A request to participate in a training session is accepted from the participant's terminal 5, and for the participant corresponding to the accepted participation request, the training information memory unit 223, which stores each role and priority in association with the training content, determines the roles in order of highest priority, associates the participant with the determined roles, and outputs information on each set role and the corresponding participant to the terminal 5 of each participant. Therefore, when a user requests to participate in a training session, the roles of the participants are determined in order of priority, so that important roles in the training session can be set for the participants. In addition, the determined roles are output to the terminal 5 of the participant, so that the user can know which role he or she has been assigned and the roles of the other participants.

[0057] Third embodiment In the first and second embodiments, a training session in which participants and NPCs (virtual avatars) make statements according to their predetermined roles is described. In the third embodiment, a session in which participants hold a discussion based on their own opinions in a diverse group is described. In the third embodiment, a session in which participants hold a discussion on a certain theme is described as an example of group activity.

[0058] <Overall Configuration of Group Activity System 300> FIG. 10 shows an example of participants taking part in a discussion held in a group activity system 300 according to the third embodiment. FIG. 11 is an overall configuration diagram of a group activity system 300 according to the third embodiment and a functional block diagram of a group activity server 301. As shown in FIG. In the example shown in FIG. 10, participants in the discussion include two real people, Mr. E and Mr. F, as well as two virtual avatars, NPC-A and NPC-B. NPC-A and NPC-B have designated profiles 361A and 361B, respectively. The profiles 361A and 361B are set for NPC-A and NPC-B before a discussion is held. In the group activity system 300, the profiles 361A and 361B are set as desired according to the purpose using background knowledge data 362. The background knowledge data 362 is external data and general knowledge possessed by the generating AI. Here, the number of participants in a discussion does not need to be particularly limited as long as it is multiple people, and in group activity system 300, a group activity is established as long as there is one or more participants who are real people.

[0059] A group activity system 300 shown in FIG. 11 includes a group activity server 301, a terminal 5 for each participant, a sentence generation AI server 304, and a terminal 307 used by the organizer of the group activity. <Group Activity Server 301> The group activity server 301 includes a control unit 310 , a storage unit 320 , and a communication interface unit 29 . Before describing the control unit 310, the storage unit 320 will be described. The storage unit 320 includes a program storage unit 321, a template storage unit 324a, a theme storage unit 324b, an external data storage unit 324c, and a conversation storage unit 324d. A group activity program 321b (program) is stored in the program storage unit 321. The group activity program 321b is, for example, a program for executing various functions of the control unit 310, which will be described next.

[0060] The template storage unit 324a is a storage area for storing templates of profile creation instructions for creating a profile for a virtual avatar. The profile creation instructions are called prompts, and are instructions input by a user to an interactive system such as a generation AI. The theme storage unit 324b is a storage area that stores at least the theme and summary of the discussion (group activity). The theme storage unit 324b may also store conditions for virtual avatars to participate in the discussion. The conditions for the virtual avatars include, for example, information on the purpose of the virtual avatar to be generated, the nationality of the virtual avatar to be generated, and information on the number of virtual avatars to be generated.

[0061] The external data storage unit 324c is a storage area for storing external data used to create a profile for a virtual avatar. The external data used here refers to a large amount of statistical data, such as questionnaire response data, for making the responses of the virtual avatar more realistic. The external data storage unit 324c may store the external data itself, or may store the address of a web page having the external data. The conversation storage unit 324d is a storage area for storing conversation information, which is utterances made by participants in a group activity, including replies made by virtual avatars, in chronological order.

[0062] Next, the control unit 310 will be described. The control unit 310 includes a discussion summary presentation unit 311, a participation acceptance unit 12, a profile setting processing unit 325 (template output means, creation instruction acceptance means, profile generation means, profile setting means, profile output means), an image output unit 14, a comment acquisition unit 315 (comment acquisition means, conversation storage means), a comment output unit 16, a reply instruction generation unit 326 (reply instruction generation means), and a reply comment output unit 318 (reply comment output means, conversation storage means). When a connection is made from a terminal 5, the discussion summary presentation unit 311 outputs to the terminal 5, for example, a screen (not shown) that presents an overview of the discussion and enables participation to be confirmed.

[0063] The profile setting processing unit 325 sets a profile of a virtual avatar that will hold a discussion with the participants accepted by the participation accepting unit 12. As an example, the profile setting processing unit 325 accepts a profile creation instruction including attribute items related to the personality of the virtual avatar. Here, the profile setting processing unit 325 may output a template for creating a profile creation instruction stored in the template storage unit 324a to the terminal 307, and accept a profile creation instruction created by the organizer using the template from the terminal 307. The profile setting processing unit 325 may also accept a profile creation instruction to which external data corresponding to the purpose of the group activity stored in the external data storage unit 324c has been further added.

[0064] Next, the profile setting processing unit 325 receives and acquires the profile from the sentence generation AI server 304 by sending a profile creation instruction to the sentence generation AI server 304. Then, the profile setting processing unit 325 sets the acquired profile for the virtual avatar. The profile setting processor 325 may also provide a step of outputting the profile received from the writing generation AI server 304 to the terminal 307, and the organizer may check the profile. The profile setting processor 325 may then re-accept a profile creation instruction edited by the organizer using the terminal 307 based on the output profile. In this case, the profile setting processor 325 may transmit the edited profile creation instruction to the writing generation AI server 304, and obtain the profile by receiving it again from the writing generation AI server 304.

[0065] The comment acquisition unit 315 acquires comments from participants from the terminal 5. The comment acquisition unit 315 may acquire the comments from participants as text data or as voice data. Then, the comment acquisition unit 315 registers the acquired comments in the conversation storage unit 324d. When the target of the comment acquired by the comment acquisition unit 315 is directed to a virtual avatar, the reply instruction generation unit 326 generates a reply creation instruction including the comment, the conversation information stored in the conversation memory unit 324d, and the set profile. The reply comment output unit 318 sends the reply creation instruction generated by the reply instruction generation unit 326 to the sentence generation AI server 304, and outputs the reply received from the sentence generation AI server 304 as a reply comment to each participant's terminal 5. In addition, the reply comment output unit 318 registers conversation information including the output reply comment in the conversation storage unit 324d.

[0066] <Text Generation AI Server 304> The sentence generation AI server 304 is a server equipped with a generation AI as a language model that has learned multiple documents. The sentence generation AI server 304 includes a control unit, a storage unit, a communication interface unit, and the like, which are not shown. The sentence generation AI server 304 may be, for example, a cloud.

[0067] <Terminal 307> The terminal 307 is a computer used by a person who organizes a group activity. The terminal 307 can be configured, for example, as a PC or a tablet terminal. Although not shown, the terminal 307 includes a control unit, a storage unit, a display unit, an input unit, a communication interface unit, and the like.

[0068] <Processing Description> Next, the processing performed by group activity server 301 will be described. FIG. 12 is a flowchart showing the member reception process of the group activity server 301 according to the third embodiment. FIG. 13 is a diagram showing an example of a profile creation instruction according to the third embodiment. FIG. 14 is a diagram showing an example of a generated profile according to the third embodiment.

[0069] 12, the control unit 310 of the group activity server 301 presents the content of the discussion and starts accepting participation in the discussion. More specifically, in response to a connection from the terminal 5, the discussion summary presentation unit 311 presents an outline of each discussion stored in the theme storage unit 324b, and outputs to the terminal 5 a screen (not shown) that enables participants to declare their participation.

[0070] The user of terminal 5 checks the outline of the discussion by looking at the screen, and if he / she wishes to participate, selects a button or the like for participation. In response, the control unit of terminal 5 transmits a participation request to group activity server 301, and in S312, participation reception unit 12 accepts the participation request from terminal 5. At that time, control unit 310 may, for example, transmit an avatar image stored in advance in storage unit 320 to terminal 5 of the participant, and accept the avatar image selected by the participant.

[0071] In S313, the profile setting processing unit 325 accepts a profile creation instruction from the host's terminal 307. Here, when the control unit 310 accepts a participation request, it may transmit to the host's terminal 307 a notice that a participation request has been received. This allows the host to create a profile creation instruction. Furthermore, in response to a request from the host's terminal 307, the profile setting processing unit 325 may transmit to the terminal 307 a template for creating a profile creation instruction.

[0072] An example of a profile creation instruction will now be described with reference to FIG. FIG. 13 shows a profile creation instruction prompt 370. The example prompt 370 is comprised of a description area 371 , a profile area 372 , a data area 373 , and a condition area 374 . The explanation area 371 is an area for specifically describing instructions to the generating AI. Profile area 372 is an area for entering profile items to be created. In this example, profile area 372 includes categories of basic information 372a, work 372b, personality 372c, and lifestyle 372d. The categories in profile area 372 and the items within the categories are examples. Here, the item values ​​for the items of each category in the profile area 372 may be entered in advance if they are essential for the profile to be generated.

[0073] The data area 373 is an area for writing external data. As shown in the text 373a, the data area 373 writes data that can be the background of the profile to be generated. The data area 373 can write external data stored in the external data storage unit 324c, for example. Here, the external data stored in the external data storage unit 324c may be various for different purposes, but as an example, statistical data issued by the Statistics Bureau of the Ministry of Internal Affairs and Communications may be used. For example, if you want to reflect age-related differences in behavior in your profile, you can use statistical information on behavior by age. One example is the "Results of the Basic Survey on Social Life" (results on time use and daily activities). For example, if you want to reflect information about your work or occupation in your profile, you can use statistical information by industry or the results of a survey on work-related attitudes. One example is information such as "A community-based society with connections and support" in the Ministry of Health, Labor and Welfare White Paper. Although the above examples show the results of surveys conducted by the government and local governments, data from surveys conducted by marketing companies may also be used. In addition, multiple external data may also be used.

[0074] Condition area 374 is an area for writing the conditions of the virtual avatar to be generated. As shown by sentence 374a, condition area 374 writes the conditions of the virtual avatar to be generated. For example, the conditions of the virtual avatar stored in theme storage unit 324b can be written in condition area 374. It should be noted that the prompt 370 described here is just one example. Furthermore, the template stored in the template storage unit 324a may be a prompt 370 in which, for example, an explanation area 371 and a profile area 372 are filled in, and a data area 373 and a condition area 374 are left blank.

[0075] In S314 of FIG. 12, the profile setting processing unit 325 transmits an instruction to create a profile to the sentence generation AI server 304. In the sentence generation AI server 304, a command to create a profile is input to the sentence generation AI, and the profile is output. Then, the sentence generation AI server 304 transmits the output profile to the group activity server 301, and in S315, the profile setting processing unit 325 receives the profile.

[0076] An example of a generated profile will now be described. FIG. 14(B) shows a profile 381 output by the generation AI as a result of inputting the prompt 380 in FIG. 14(A) into the generation AI. In this example, the profile area of ​​the prompt 380 lists only the item names of each category, as shown in Fig. 13. In contrast, item values ​​are set for each item name in the profile 381. Here, for ease of understanding, the item values ​​set by the sentence generation AI server 304 are shown in bold and underlined.

[0077] Here, the profile setting processing unit 325 may transmit the profile received from the sentence generation AI server 304 to the host's terminal 307, so that the host can edit the profile. In S316 of FIG. 12, the profile setting processing unit 325 sets the received profile in the virtual avatar. In S317, the image output unit 14 outputs the avatar image of each participant, including the virtual avatar, to the terminal 5 of each participant. Thereafter, the control unit 310 ends this process and starts the next discussion process. In the above example, the process of creating a profile for a virtual avatar is performed during the member reception process, but the present invention is not limited to this. Of course, a profile may be created before the member reception process.

[0078] Next, the discussion process will be described. FIG. 15 is a flowchart showing the discussion process of the group activity server 301 according to the third embodiment. FIG. 16 is a flowchart showing a virtual avatar statement process of the group activity server 301 according to the third embodiment. FIG. 17 shows an example of a reply creating instruction and an example of a reply according to the third embodiment. 15, the control unit 310 of the group activity server 301 outputs the topic of the discussion. The control unit 310 may output the topic of the discussion stored in the theme storage unit 324b. This allows each participant to check the title and progress the discussion by making comments according to the title.

[0079] In S322, the comment acquisition unit 315 judges whether or not a comment from a participant has been acquired from the terminal 5. If a comment from a participant has been acquired from the terminal 5 (S322: YES), the control unit 310 shifts the process to S323. On the other hand, if a comment from a participant has not been acquired from the terminal 5 (S322: NO), the control unit 310 shifts the process to S327. In S323, the comment output unit 316 outputs the acquired comment to the other terminals 5. In this way, the other participants can grasp the content of the comment made by the participant. In S324, the comment acquiring unit 315 stores the acquired comment in the conversation storage unit 324d, thereby accumulating conversation information.

[0080] In S325, the control unit 310 judges whether the acquired comment is directed to a virtual avatar. If the acquired comment is directed to a virtual avatar (S325: YES), the control unit 310 shifts the process to S326. On the other hand, if the acquired comment is not directed to a virtual avatar (S325: NO), the control unit 310 shifts the process to S327. In S326, control unit 310 performs virtual avatar utterance processing.

[0081] Here, the virtual avatar statement processing will be described with reference to FIG. In S331 of FIG. 16, the reply instruction generator 326 generates a reply creation instruction that includes the comment acquired in S322 of FIG. 15, the conversation information stored in the conversation storage unit 324d, and the set profile. As an example of a reply creation instruction, a reply creation prompt 390 is shown in FIG. 17(A). A prompt 390 shown in FIG. 17(A) is made up of an explanation area 391, a conversation history area 392, a question area 393, and a profile area 394. The explanation area 391 is an area for specifically describing instructions to the generating AI. The conversation history area 392 is an area in which conversation information stored in the conversation storage unit 324d is written in chronological order. The question area 393 is an area for writing comments to the virtual avatar. Profile area 394 is an area where the profile of the virtual avatar is described.

[0082] In S332 of FIG. 16, the reply comment output unit 318 sends the generated reply creation instruction to the text generation AI server 304. In S333, the reply comment output unit 318 receives the reply from the sentence generation AI server 304. Here, FIG. 17(B) shows a reply 395 output in response to the prompt 390 in FIG. 17(A). A response 395 shown in FIG. 17(B) is generated in response to a question in a question area 393 of a prompt 390 based on a profile set in a profile area 394, an explanation area 391, and a conversation history area 392.

[0083] In S334 of FIG. 16, the reply output module 318 outputs the received reply as a reply to the terminal 5 of the participant. In S335, the reply comment output section 318 stores the reply comment in the conversation storage section 324d. After that, the control section 310 advances the process to S327 in FIG.

[0084] In S327 of Fig. 15, the control unit 310 judges whether or not to end the process. For example, when a participant inputs that the discussion has ended or a predetermined time has elapsed, the control unit 310 judges to end the process. If the process is to be ended (S327: YES), the control unit 310 ends this process. On the other hand, if the process is not to be ended (S327: NO), the control unit 310 moves the process to S322.

[0085] As the group activities described above, for example, the following activities are envisaged: (1) Discussions by diverse members It is possible to hold discussions with members from diverse occupations, nationalities, origins, age groups, etc., which would be difficult to achieve in reality. For example, in the case of a discussion among members of various nationalities, it is difficult in reality to create a group in which people of different nationalities can converse smoothly in the same language. However, by using the group activity system 300 of this embodiment, this can be easily realized by specifying the nationality and language of the virtual avatar. In addition, in-house discussions are limited in the number of members, but if there are members with attributes that are not available within your company or if your company does not have enough people to hold group activities, using the group activity system 300 of this embodiment makes it easy to implement.

[0086] (2) Discussions that take into consideration individual personalities All members except one participant may be virtual avatars. In reality, to hold a discussion among multiple participants, it is necessary to adjust the time and the date, but by discussing with a virtual avatar, it is possible to hold a discussion at a pace that suits each individual. In particular, by using the group activity system 300 of this embodiment, it is possible to hold a discussion that includes honest thoughts that are difficult to express due to concern for others, and to discuss sensitive themes. In addition, discussions can be held with consideration for people who have difficulty communicating with others due to chronic illnesses or disabilities, and the system can also be used for training purposes.

[0087] Thus, according to the third embodiment, the group activity server 301 has the following advantages. (1) Conversation information including the statement acquired by the statement acquisition unit 315 and the reply statement output by the reply statement output unit 318 is stored in the conversation memory unit 324d, and when the target of the statement acquired by the statement acquisition unit 315 is directed to a virtual avatar, a reply creation instruction including the statement, the conversation information stored in the conversation memory unit 324d, and the profile of the virtual avatar is generated, and the reply statement output unit 318 inputs the reply creation instruction to the sentence generation AI so that the reply output by the sentence generation AI is output as a reply statement to the terminal 5 of each participant. Therefore, the virtual avatar can be made to respond based on the profile, and can be made to respond in a manner similar to that of a living person.

[0088] (2) A profile creation instruction including attribute items related to the personality of the virtual avatar is accepted, and the accepted profile creation instruction is input to a text generation AI to obtain a profile output by the text generation AI, and the obtained profile is set for the virtual avatar. Therefore, the profile of the virtual avatar can be created by a text generation AI, and various virtual avatars can be created as participants.

[0089] (3) Group activity server 301 acquires external data according to the purpose of group activities from an external server communicably connected to it, and receives an instruction to create a profile that further includes the acquired external data. Therefore, by adding external data to the profile creation instructions, it is possible to construct virtual avatars as diverse participants according to the purpose of group activities. In other words, external data can be used as background knowledge for the virtual avatars.

[0090] (4) The text generation AI outputs the profile it generates and accepts profile creation instructions edited by the user based on the output profile. Therefore, by editing the profile generated by the text generation AI, a user can create a virtual avatar with a profile that suits their preferences.

[0091] (5) A template for creating a profile creation instruction is output, and profile creation instructions created using the output template are accepted. This provides an environment that makes it easy to create a profile creation instruction.

[0092] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-mentioned embodiments. Furthermore, the effects described in the embodiments are merely a list of the most preferable effects resulting from the present invention, and the effects of the present invention are not limited to those described in the embodiments. The above-mentioned embodiments and the modified forms described below can be used in appropriate combination, but detailed description will be omitted.

[0093] (Variations) (1) In the first embodiment, a language model that has learned a plurality of documents is used to obtain a reply statement from an NPC, and in the second embodiment, a character model that has learned a plurality of documents related to characters is used to obtain a reply statement from an NPC. However, the present invention is not limited to these. For example, the storage unit may have a dialogue information storage unit that stores a plurality of dialogue information consisting of questions and replies related to characters, and a reply statement output means that is a control unit may search the dialogue information storage unit to obtain a question similar to the replies, and output the replies of the dialogue information corresponding to the obtained questions as reply statements to the terminals of the participants. Even in this way, the NPC can appropriately make a reply statement to the participant's statement even in the character that it corresponds to.

[0094] (2) In the first embodiment, the roles of the participants are determined randomly, and in the second embodiment, the roles of the participants are determined in order of priority. However, this is not limiting. For example, the role desired by the participant may be determined by obtaining a role designation from the participant. In this case, if multiple participants designate the same role, a combination of multiple determination methods may be used, such as randomly determining one of the participants who designated the same role, and determining the roles of the remaining participants in order of priority. Alternatively, the roles desired by the participants may be determined in the order of application.

[0095] (3) In each embodiment, the training or discussion is carried out by displaying avatar images in a virtual space, but the present invention is not limited to this. For example, the training or discussion may be a chat-based training or discussion conducted only using text, or may be a training or discussion using a video conference system.

[0096] (4) In the first and second embodiments, the security training has been described as an example, but the present invention is not limited to this. Examples of the training include sales training in which participants are divided into sales and customer roles and demonstrate actual business negotiations, and business etiquette training in which participants demonstrate business etiquette such as answering the phone and serving customers according to their roles. The present invention can also be used in various role-playing type training.

[0097] (5) In the first and second embodiments, the document storage unit and the training information storage unit are stored in the storage unit of the training server, and in the second embodiment, the knowledge unit is further stored in the storage unit of the training server. However, the present invention is not limited to this. The document storage unit and the like may be stored in an external server, and the training server may be connected to the external server that stores the document storage unit and the like via the communication network N or directly.

[0098] (6) In the third embodiment, a profile is created using a text generation AI, but the present invention is not limited to this. For example, a profile may be created manually. Alternatively, the group activity server may include an attribute item information storage unit that stores attribute item information in which multiple item values ​​are associated with attribute items related to the personality of a virtual avatar, and the control unit (profile reception means) may output the attribute item information stored in the attribute item information storage unit so that it can be referenced, and receive combinations of attribute items and item values. Then, the control unit (profile setting means) may set the received combination of attribute items and item values ​​as a profile for the virtual avatar. [Explanation of symbols]

[0099] 1,201 Training Servers 3 Knowledge Server 5,307 devices 10,210,310 Control section 12 Participant Registration Section 13,213 Role setting processing section 14 Image output section 15,315 Speech Acquisition Department 16 Speech output section 17 Related Document Acquisition Department 18,218,318 Reply output section 20,220,320 storage section 21a,221a Training Program 22 Document storage unit 23,223 Training Information Storage Section 100,200 Training System 219 Model Learning Department 230 Knowledge Department 300 Group Activity System 301 Group Activity Server 304 Text generation AI server 311 Discussion summary presentation section 321b Group Activity Program 324a Template storage section 324b Theme Memory Section 324c External data storage unit 324d Conversation Memory Section 325 Profile setting processing section 326 Reply Instruction Generation Unit

Claims

1. 1. A group activity device for supporting a group activity between one or more participants and a virtual avatar, comprising: a profile setting means for setting a profile for the virtual avatar; a comment acquisition means for acquiring comments of the participants from terminals used by the participants; a comment output means for outputting the comment acquired by the comment acquisition means to terminals of other participants; a reply comment output means for, when the target of the comment acquired by the comment acquisition means is directed to the virtual avatar, acquiring a reply comment to the comment based on the profile set by the profile setting means, and outputting the reply comment to the terminal of each participant; A group activity device comprising:

2. 2. The group activity device according to claim 1, The group activity is a role-playing type training in which each participant is assigned a different role, a knowledge department having knowledge corresponding to each role in the training; A role setting means for associating the participants of the training with the roles; Equipped with The profile setting means sets an unset role to which the participant is not associated as the profile to the virtual avatar, The reply statement output means, when the target of the statement acquired by the statement acquisition means is directed to the role to which the virtual avatar is associated, acquires a reply statement based on the role to the statement using the knowledge unit, and outputs the reply statement to the terminal of each participant, in a group activity device.

3. 3. The group activity device according to claim 2, The knowledge unit is a language model trained on a plurality of documents, a document storage unit for storing a document having information corresponding to each winning combination; a related document acquisition means for acquiring, from the document storage unit, a related text from the document related to the utterance when the utterance acquired by the utterance acquisition means is made to the character associated with the virtual avatar by the character setting means; Equipped with A group activity device in which the reply comment output means outputs the reply comment output by the knowledge section by inputting the comment together with the text acquired by the related document acquisition means into the knowledge section, to the terminal of each participant.

4. 3. The group activity device according to claim 2, the knowledge unit is a role model that has learned a plurality of documents related to the roles, the role model corresponding to each role, The reply statement output means outputs to the terminal of each participant the reply statement output by the character model in the knowledge section corresponding to the character associated with the virtual avatar when the statement is input into the character model.

5. 3. The group activity device according to claim 2, the knowledge unit is a dialogue information storage unit that stores a plurality of dialogue pieces of information consisting of questions and responses regarding the role, The reply statement output means searches the knowledge section to obtain a question similar to the statement, and outputs the reply of the dialogue information corresponding to the obtained question as the reply statement to the terminal of each participant.

6. 3. The group activity device according to claim 2, A group activity device comprising an image output means for arranging avatar images of the participants, including an avatar image of the virtual avatar, in a virtual space and outputting to each participant's terminal a training space image in which a role name indicating the role of each participant is arranged near the avatar image.

7. 7. The group activity device according to claim 6, The statement acquisition means receives the statement as voice data, The reply comment output means converts the voice data into text, then uses the knowledge unit to acquire the reply comment to the comment, converts the acquired reply comment into voice data, and outputs it to the terminal of each participant.

8. A group activity device according to any one of claims 2 to 7, a training information storage unit that stores each role in association with the training content; A participation reception means for receiving a participation request for one of the training sessions from the terminal of the participant; Equipped with the role setting means randomly determines the role for the participant corresponding to the participation request accepted by the participation accepting means by referring to the training information storage unit, and associates the participant with the determined role; A group activity device comprising: a role output means for outputting information on each role set by the role setting means and the corresponding participant to the terminal of each participant.

9. A group activity device according to any one of claims 2 to 7, a training information storage unit that stores each role and a priority in association with the training content; A participation reception means for receiving a participation request for one of the training sessions from the terminal of the participant; Equipped with the role setting means determines the roles of the participants corresponding to the participation requests accepted by the participation accepting means in the order of priority by referring to the training information storage unit, and associates the participants with the determined roles; A group activity device comprising: a role output means for outputting information on each role set by the role setting means and the participants corresponding thereto to the terminal of each participant.

10. 2. The group activity device according to claim 1, a conversation storage means for storing conversation information including the comments acquired by the comment acquisition means and the reply comments output by the reply comment output means; a reply instruction generating means for generating a reply creation instruction including the comment, the conversation information stored by the conversation storage means, and the profile set by the profile setting means, when the comment acquired by the comment acquiring means is directed to the virtual avatar; Equipped with The reply statement output means inputs the reply creation instruction generated by the reply instruction generation means to a sentence generation AI that generates a sentence based on the creation instruction, and outputs the reply output by the sentence generation AI as the reply statement to the terminal of each participant, in this group activity device.

11. 11. The group activity device according to claim 10, a profile creation instruction receiving means for receiving an instruction to create a profile including attribute items related to the personality of the virtual avatar; a profile generation means for acquiring a profile output by the sentence generation AI by inputting the profile generation instruction received by the generation instruction receiving means to the sentence generation AI; Equipped with The profile setting means sets the profile generated by the profile generating means for the virtual avatar.

12. 12. The group activity device according to claim 11, The creation instruction receiving means of the group activity device acquires external data corresponding to the purpose of the group activity from an external server communicatively connected to the group activity device, and receives the profile creation instruction to which the acquired external data is further added.

13. 12. The group activity device according to claim 11, a profile output means for outputting the profile generated by the profile generation means, The creation instruction receiving means receives an instruction to create the profile edited based on the profile output by the profile output means.

14. 12. The group activity device according to claim 11, a template output means for outputting a template for generating the profile generation instruction, The creation instruction receiving means receives an instruction to create the profile created using the template output by the template output means.

15. 11. The group activity device according to claim 10, an attribute item information storage unit that stores attribute item information in which a plurality of item values ​​are associated with attribute items related to the personality of the virtual avatar; a profile receiving means for outputting the attribute item information stored in the attribute item information storage unit in a referable manner and receiving a combination of the attribute items and the item values; The profile setting means sets a combination of the attribute items and the item values ​​accepted by the profile accepting means as the profile for the virtual avatar.

16. A group activity system comprising the group activity device and the sentence generation AI according to any one of claims 10 to 15.

17. A training device for performing role-playing type training in which different roles are set for each participant, A role setting means for associating participants in the training with the roles; a comment acquisition means for acquiring comments of the participants from terminals used by the participants; a comment output means for outputting the comment acquired by the comment acquisition means to terminals of other participants; a reply statement output means for, when the target of the statement acquired by the statement acquisition means is for an unset role to which the participant is not associated by the role setting means, acquiring a reply statement based on the unset role in response to the statement using a knowledge section which is communicably connected to the training device and has knowledge corresponding to each role in the training, and outputting the reply statement to the terminal of each participant; A training device comprising:

18. A training system comprising the training device according to claim 17 and the knowledge unit.

19. Computer, A profile setting means for setting a profile for a virtual avatar that is engaged in a group activity with one or more participants; a comment acquisition means for acquiring comments of the participants from terminals used by the participants; a comment output means for outputting the comment acquired by the comment acquisition means to terminals of other participants; a reply comment output means for, when the target of the comment acquired by the comment acquisition means is directed to the virtual avatar, acquiring a reply comment to the comment based on the profile set by the profile setting means, and outputting the reply comment to the terminal of each participant; A program to make it function as such.

Images