JP7460839B1 - Programs and systems - Google Patents

Abstract

[Problem] To provide a program that can publish a unique profile created according to an object in association with the object without increasing the workload of the user. [Solution] The program causes a computer to function as: object generation means that generates an object according to multiple setting values selected by the user to set the visual characteristics of the object to be placed in an online space; profile generation means that inputs the multiple setting values into a trained model that has been trained in advance to generate a profile for the object; and publishing means that publishes in the online space the profile generated by the profile generation means in association with the object generated by the object generation means. [Selected Figure] Figure 13

Description

The present invention relates to a program and a system.

BACKGROUND ART Conventionally, there has been known a technique for disclosing the profile of one's avatar to other avatars in a VR (Virtual Reality) space (see, for example, Patent Documents 1 and 2). Then, in the VR space, it is possible to see the profiles of other avatars and decide whether to become friends with the avatars (so-called friend registration) or not.

Similarly, in cooperative online games in which multiple avatars play together, or competitive online games in which multiple avatars play against each other, players can view the profiles of other avatars and select an avatar to cooperate or compete with. In this way, the profile of each avatar is becoming increasingly important in VR spaces and online games.

JP 2018-170013 Publication Japanese Patent Application Publication No. 2018-106499

However, many users feel that creating a profile is a burden. Therefore, when the number of avatars that use a predetermined default profile increases, a problem arises in that the profile becomes useless when selecting an avatar to register as a friend, cooperate with, or compete against.

The present invention has been made in consideration of the above circumstances, and its purpose is to provide a program that can publish a unique profile created to match an object, in association with the object, without increasing the workload of the user.

To solve the above problem, the program of the present invention causes a computer to function as an object generation means for generating an object in accordance with multiple setting values selected by a user to set the visual characteristics of the object to be placed in an online space, a profile generation means for inputting the multiple setting values into a trained model that has been trained in advance to generate a profile for the object, and a publishing means for associating the profile generated by the profile generation means with the object generated by the object generation means and publishing the profile in the online space.

According to the present invention, a unique profile created to match an object can be made public in association with the object without increasing the user's workload.

1 is a diagram showing an overview of a system according to an embodiment of the present invention; 1 is a conceptual diagram of AI-based learning processing (A) and generation processing (B). FIG. 1 is a diagram showing a neural network that realizes AI. FIG. 2 is a hardware configuration diagram of a user terminal. FIG. 2 is a functional block diagram of a user terminal. This is an example of an avatar generation screen. 13 shows an example of the priority (A) of setting parameters and setting value histograms (B) and (C). 1 shows an example of profiles (A), (B) and learning data (C). This is another example of the avatar generation screen. This is another example of the setting parameter priority (A) and setting value histograms (B) and (C). 13 is another example of profiles (A), (B) and learning data (C). This is an example of a profile editing screen (A) and learning data (B). 13A shows an example of a user selection screen and FIG. 13B shows an example of a user details screen.

Below, the system 1 according to the embodiment will be described with reference to the drawings. Note that the embodiment of the present invention described below is an example of how the present invention can be realized, and does not limit the scope of the present invention to the scope of the described embodiment. Therefore, the present invention can be implemented by making various modifications to the embodiment.

[Overview of System 1]
FIG. 1 is a diagram showing an overview of a system 1 according to this embodiment. As shown in FIG. 1, the system 1 mainly includes a game server 10, an AI server 15, and a plurality of user terminals 20. Although three user terminals 20 are illustrated in FIG. 1, the number of user terminals 20 included in the system 1 is not limited to this. The game server 10, the AI server 15, and the user terminals 20 are connected to each other so as to be able to communicate with each other via a communication network 2. Specific examples of the communication network 2 are not particularly limited, and may be, for example, the Internet, a mobile communication system (e.g., 4G, 5G, etc.), a wireless network such as Wi-Fi (registered trademark), or a combination thereof.

As an example, the system 1 realizes a cooperative online game in which users (players) of a plurality of user terminals 20A, 20B, and 20C cooperate to play. As another example, the system 1 realizes a competitive online game in which users (players) of a plurality of user terminals 20A, 20B, and 20C compete against each other. As yet another example, a virtual space is realized in which users of a plurality of user terminals 20A, 20B, and 20C communicate via avatars. Since these systems 1 are already well known, detailed explanation will be omitted.

[Configuration of game server 10]
The game server 10 is realized, for example, by a general-purpose computer such as a workstation or a personal computer. As an example, a game server 10 that provides online games synchronizes game data of each of a plurality of user terminals 20. As another example, the game server 10 that provides a virtual space synchronizes virtual space data of each of a plurality of user terminals 20. Hereinafter, online games and virtual spaces in which multiple users can participate will be collectively referred to as "online spaces." That is, the game server 10 provides online space to the plurality of user terminals 20A, 20B, and 20C.

[Configuration of AI server 15]
The AI server 15 is realized, for example, by a workstation or a general-purpose computer such as a personal computer. The AI server 15 realizes AI (Artificial Intelligence) 16 including a trained model. As an example, the AI server 15 may be equipped with a dedicated AI optimized for the online space provided by the game server 10. At this time, the game server 10 and the AI server 15 may be realized by one piece of hardware. As another example, the AI server 15 may be equipped with general-purpose AI that can be used for various purposes. Since the configuration of the AI 16 is already well known, a detailed explanation will be omitted, but for example, the configuration is as follows.

FIG. 2 is a conceptual diagram of learning processing (A) and generation processing (B) by the AI 16. FIG. 3 is a diagram showing the neural network 17 that realizes the AI 16. The AI 16 installed in the AI server 15 is a so-called "generation AI" that processes input data and generates output data. The AI 16 then executes the learning process shown in FIG. 2(A) and the generation process shown in FIG. 2(B). Furthermore, the AI 16 uses, for example, the neural network 17 shown in FIG. 3 to generate output data from input data.

As shown in FIG. 2(A), AI 16 includes learning model 16a. Learning model 16a refers to the neural network 17 before the learning process is performed. The learning model 16a becomes trained model 16b with learning data including input data and correct answer data. Input data refers to data input to AI 16. Correct answer data refers to data that should be output when input data is input. By inputting multiple pieces of learning data into learning model 16a, the neural network 17 is optimized as described below, and becomes trained model 16b.

Also, as shown in FIG. 2(B), the trained model 16b generates and outputs output data by inputting input data to the neural network 17. The input data and output data may be in any format, such as text, image, or audio. The input data and output data may also be in different formats. On the other hand, the format of the correct answer data is the same as the format of the output data.

The learning process shown in FIG. 2(A) may be performed not only on the learning model 16a, but also on the trained model 16b. In addition, the AI 16 does not need to learn using the input data and correct answer data actually used in the present invention, and may learn using general-purpose learning data. Furthermore, the learning process performed by the AI 16 is not limited to "supervised learning" in which input data and correct answer data are input, but may be "unsupervised learning" in which correct answer data is not input, or may perform reinforcement learning, transfer learning, etc.

As shown in FIG. 3, the neural network 17 of the learning model 16a or the trained model 16b includes an input layer L1 composed of a plurality of nodes I1, I2, and I3, and a plurality of nodes H1, H2, H3, H4, and H5. and an output layer L3 including a plurality of nodes O1, O2, and O3. In the example of FIG. 3, the numbers of nodes in the input layer L1 and output layer L3 are the same, but the numbers of nodes in the input layer L1 and output layer L3 may be different. Furthermore, the neural network 17 may include a plurality of intermediate layers L2. Furthermore, although FIG. 3 shows a fully connected neural network 17 in which a plurality of nodes constituting each layer L1, L2, and L3 are connected to all nodes of adjacent layers, the structure of the neural network 17 is as follows. but not limited to.

The learning process is a process of adjusting the weighting coefficients and biases of each node so that correct data is output from the output layer L3 when input data is input to the input layer L1. The generation process is a process of generating output data and outputting it from the output layer L3 by processing the input data input to the input layer L1 using weighting coefficients and biases that have been adjusted in advance for each node.

However, the specific example of the processing by the AI 16 described above is just an example, and the AI server 15 can employ various well-known techniques. As another example, the AI 16 may have a CNN (Convolutional Neural Network) network structure. As still another example, the network structure may have a configuration such as LLM (Large Language Model), RNN (Recurrent Neural Network), or LSTM (Long Short-Term Memory). good. That is, the AI 16 may have a network structure other than deep learning.

In this embodiment, an example will be described in which input data in text format is input to the trained model 16b, and output data in text format is generated. More specifically, the input data in this embodiment is a number of setting values selected by the user to set the visual or auditory (at least visual) characteristics of an avatar (object). Also, the output data in this embodiment is a profile of the avatar (object) generated in accordance with the number of setting values.

[Configuration of user terminal 20]
The user terminal 20 is realized as, for example, an HMD set, a tablet terminal, a smartphone, a feature phone, a laptop computer, a desktop computer, etc. In this embodiment, an example of the user terminal 20 as a tablet terminal will be described as shown in FIG.

Figure 4 is a hardware configuration diagram of the user terminal 20. As shown in Figure 4, the user terminal 20 mainly includes a processor 21, a memory 22, a storage 23, a communication interface 25, a monitor 31, cameras 33 and 34, a microphone 35, a speaker 36, a motion sensor 41, and an operation device 42 (operation unit). Each component of the user terminal 20 is connected to a communication bus 29.

The processor 21 executes a series of instructions included in the terminal program 23P stored in the memory 22 or the storage 23, thereby realizing the processing described below. The processor 21 is, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an MPU (Micro Processing Unit), or an FPGA (Field-Programmable Gate A). rray) and other devices.

The memory 22 temporarily holds the terminal program 23P and data. The terminal program 23P is loaded, for example, from the storage 23. The data includes data input to the user terminal 20 and data generated by the processor 21. For example, the memory 22 is realized as a RAM (Random Access Memory) or other volatile memory.

The storage 23 permanently holds the terminal program 23P and data. The storage 23 is realized, for example, as a ROM (Read-Only Memory), a hard disk device, a flash memory, or other nonvolatile storage device. Further, the storage 23 may be realized as a removable storage device such as a memory card. As yet another example, the storage 23 may be connected to the user terminal 20 as an external storage device instead of being built into the user terminal 20. According to such a configuration, for example, in a scene where a plurality of user terminals 20 are used as in an amusement facility, it becomes possible to update the terminal program 23P and data all at once.

The communication interface 25 communicates with other devices (eg, the game server 10, the AI server 15) connected to the communication network 2. The communication interface 25 may be, for example, a wired communication interface such as a LAN (Local Area Network), or a wireless communication interface such as Wi-Fi (Wireless Fidelity), Bluetooth (registered trademark), or NFC (Near Field Communication). Realized.

For example, as shown in FIG. 1, the monitor 31 is provided on the surface of a flat case. The monitor 31 is a display device (display unit) that displays images or videos. The camera 33 is a so-called in-camera that is attached to the surface of a flat case and captures an image of the face of the user viewing the monitor 31 . The camera 34 is a so-called outside camera that is attached to the back surface of the flat case (the surface opposite to the monitor 31) and captures an image of the surroundings.

The microphone 35 converts the user's speech into an audio signal (electrical signal) and outputs it to the computer 26. The speaker 36 converts the audio signal output from the computer 26 into sound and outputs it to the user. Note that the user terminal 20 may include an earphone instead of the speaker 36.

The motion sensor 41 detects the movement of the housing (for example, rotation around three mutually orthogonal axes). The motion sensor 41 may be realized by, for example, an angular velocity sensor, a geomagnetic sensor, or an acceleration sensor.

The operation device 42 accepts input (operation) of commands by the user to the user terminal 20. The operation device 42 is, for example, a touch panel that is superimposed on the monitor 31 and accepts various touch operations by the user. In other words, the monitor 31 according to this embodiment is a touch panel type display unit. However, the specific hardware configuration of the user terminal 20 is not limited to the above example. As another example, the user terminal 20 may be provided with a controller equipped with buttons, an operation stick, etc. as the operation device 42.

[Functional block diagram of user terminal 20]
FIG. 5 is a functional block diagram of the user terminal 20. As shown in FIG. 5, the terminal program 23P loaded into the memory 22 connects the user terminal 20 (computer) to the object generation means 210, the selection means 220, the profile generation means 230, the automatic editing means 240, and the learning It functions as means 250, publishing means 260, user editing means 270, and feedback means 280.

Although FIG. 5 mainly shows functional blocks for generating and publishing a profile, the user terminal 20 also handles processing in the online space (for example, participation in a game, avatar movement). It has further functions to realize. Further, although FIG. 5 mainly shows the functional blocks of the user terminal 20A, the user terminals 20B and 20C also have similar functional blocks.

First, an example of the process of generating an avatar and profile in the user terminal 20A will be described with reference to FIGS. 6 to 8. FIG. 6 is an example of an avatar generation screen. FIG. 7 is an example of the setting parameter priority (A) and setting value histograms (B) and (C). FIG. 8 is an example of profiles (A), (B) and learning data (C).

The object generation means 210 displays the avatar generation screen shown in FIG. 6 on the monitor 31. The avatar generation screen is a screen for generating an avatar (character). The avatar generation screen is also a screen for allowing the user of the user terminal 20A to select multiple setting values in order to set the visual (appearance) or auditory (tone of voice) characteristics of the avatar. Below, the process of generating an avatar and profile will be explained, focusing on visual characteristics.

An avatar is an example of an object that represents a user and operates within an online space. An avatar may be in a 3D format, which is a shape drawn in three dimensions, or in a 2D format, which is a shape drawn in two dimensions. An avatar is specified by data representing visual characteristics (e.g., image data, 3D model data) and data representing auditory characteristics (e.g., audio data), or a combination of these.

As shown in FIG. 6, the avatar generation screen includes an avatar display area A1, a gender selection area A2, an age selection area A3, a setting parameter selection area A4, a candidate icon selection area A5, and a "Decide" icon.

Avatar display area A1 is an area that displays an image of an avatar (hereinafter, referred to as an "avatar image") generated according to the setting values selected by the user through selection areas A2 to A5. That is, the appearance of the avatar image displayed in avatar display area A1 changes as the setting values are changed through selection areas A2 to A5. In addition, the displayed avatar image may change (e.g., rotate, shrink, enlarge) in response to a touch operation (e.g., a swipe operation, a pinch-in operation, a pinch-out operation) on avatar display area A1.

Gender selection area A2 is an area for selecting the gender of the avatar. Gender selection area A2 includes a [male] icon and a [female] icon as candidates for setting values of gender parameters. The age selection area A3 is an area for selecting the age of the avatar. The age selection area A3 includes a [child] icon and an [adult] icon as candidates for the setting value of the age parameter. However, the age selection area A3 may include a text box or pull-down menu for inputting the date of birth.

The setting parameter selection area A4 is an area for selecting setting parameters (contour, hairstyle, skin colour, clothing, colour). The setting parameter selection area A4 includes an [Contour] tab, an [Hairstyle] tab, a [Skin Colour] tab, a [Clothing] tab, and a [Color] tab. The contour indicates the contour of the avatar's face. The hairstyle indicates the avatar's hairstyle. The skin colour indicates the avatar's skin colour. The clothing indicates the clothing worn by the avatar. The colour indicates the base colour of the avatar (e.g. the colour of the hair, clothing, accessories).

The candidate icon selection area A5 is an area for selecting a setting value corresponding to the selection parameter selected in the setting parameter selection area A4. As an example, when the [Hairstyle] tab is selected in the setting parameter selection area A4, the setting parameter selection area A4 shows candidates for selectable hairstyle setting values (for example, short, bun, twintails, ponytail). Multiple candidate icons are displayed. As another example, when the [Clothing] tab is selected in the setting parameter selection area A4, candidates for selectable clothing settings (for example, dress, miniskirt, denim, maid) are displayed in the setting parameter selection area A4. Multiple candidate icons are displayed.

In the example of FIG. 6, gender, age, contour, hairstyle, skin color, clothing, and color are examples of setting parameters. Also, the candidate icons (e.g., the "male" icon, the "female" icon) that can be selected in association with the setting parameters (e.g., gender) are candidates for setting values that represent the visual characteristics of the avatar (object). In other words, a setting parameter represents a group (type) of setting values of the same type.

However, specific examples of setting parameters are not limited to the above examples. Furthermore, for the above setting parameters, the selectable setting value candidates are not limited to the above examples. Furthermore, setting parameters are not limited to those that set the visual characteristics of the avatar. For example, a voice tone parameter is an example of a setting parameter that sets the auditory characteristics of the avatar. Setting values for the voice tone parameter include soprano, mezzo-soprano, alto, etc.

Please note that the avatar itself as a visual feature, the items worn by the avatar as a visual feature (e.g. costumes, accessories, etc.), and the data of the avatar's voice as an auditory feature may be collected through billing, gacha, etc. It may be possible to obtain it. For example, it may be possible to acquire avatars that look different from the default avatars, items worn by characters appearing in manga or anime, voice data of real voice actors, and the like.

Then, the object generation means 210 generates a plurality of setting values selected by the user through the selection areas A2 to A5 (for example, gender=female, age=child, outline=round face, hairstyle=pwintails, skin color=fair skin, clothes= Avatar data (image data, 3D model data) is generated according to the maid (color=white), and an avatar image indicated by the generated avatar data is displayed in the avatar display area A1. Further, the object generation means 210 finalizes the generated avatar data in response to the selection of the [Determination] icon.

The selection means 220 selects setting values to be input to the trained model 16b from among the multiple setting values selected through the avatar generation screen. The selection means 220 selects setting values to be input to the trained model 16b, for example, using the priority table shown in FIG. 7(A) and the setting value histograms shown in FIG. 7(B) and FIG. 7(C). In this embodiment, four setting values are selected from the seven setting values. However, the number of setting values to be selected is not limited to the above example.

The priority table shown in FIG. 7A holds initial priorities for each setting parameter. The initial priority indicates a predetermined priority of a plurality of setting parameters. The smaller the initial priority, the higher the priority. For example, the priority table is set by the administrator of the game server 10 and downloaded to the user terminal 20A in advance. That is, if the selection means 220 follows only the priority table shown in FIG. 7(A), it will select fixed setting values for gender, age, hairstyle, and skin color.

The setting value histograms shown in FIGS. 7(B) and 7(C) show setting values selected by other users in the past for each setting parameter (for example, hairstyle, clothes) in the online space provided by the game server 10. This is the total. The vertical axis in FIG. 7(B) and FIG. 7(C) may be the number of selections (frequency) or the selection ratio (%). These setting value histograms are compiled by the game server 10, for example. Then, the selection means 220 receives the setting value histogram from the game server 10 through the communication IF 25. Although FIG. 7 shows only the setting value histograms of the hairstyle parameters and clothing parameters, reference may be made to the setting value histograms of other setting parameters.

Then, the selection means 220 changes the priority of the setting parameters based on the setting value histogram acquired from the game server 10. As an example, as shown in FIG. 7(B), the selection means 220 does not change the priority of the hairstyle parameters because "twin tails" has been selected the second most times among the hairstyle parameters. As another example, as shown in FIG. 7C, the selection means 220 changes the priority of the clothing parameter to "1" (the highest do).

As a result, in the examples of FIGS. 6 and 7, the selection means 220 selects gender = female, age = child, clothing = maid, and skin color = fair skin as the setting values to be input to the learned model 16b. Become. In this way, the selection means 220 may select, from among the plurality of setting values, the setting value that has been selected less frequently by other users. However, the method by which the selection means 220 selects the setting value is not limited to this. As another example, the selection means 220 may randomly select a setting value to be input to the learned model 16b from among a plurality of setting values. As still another example, the selection means 220 may input setting values indicating the avatar, item, and voice obtained by paying a fee to the learned model 16b with priority over other setting values.

The profile generating means 230 inputs the settings selected by the selecting means 220 (e.g., gender = female, age = child, clothing = maid, skin color = fair) into the trained model 16b to generate an avatar profile. The profile generating means 230 may also input the name of the avatar determined (input) by the user (e.g., name = Linda) into the trained model 16b. The profile generating means 230 may also input the range of the number of characters in the profile (e.g., 70 to 100 characters) into the trained model 16b.

For example, when the AI 16 capable of natural language processing generates a profile, the profile generation means 230 inputs text-format input data such as ``Generate a profile of a woman, child, maid, and fair-skinned Linda in 70 to 100 characters.'' Please send it to the AI server 15 via the communication IF 25.

The AI server 15 inputs input data acquired from the user terminal 20A into the input layer L1, processes the data in the intermediate layer L2, and outputs a text-format profile from the output layer L3. The AI server 15 also transmits the generated profile to the user terminal 20A. The profile generation means 230 then receives the profile (output data) shown in FIG. 8(A) from the AI server 15 via the communication IF 25.

A profile is text-format data that introduces an avatar. A profile may include any text information related to an avatar, such as the avatar's preferences, what the avatar wants to do in the online space, messages to other users, a backstory (e.g., hometown, educational background, work history), and family composition. A profile is also text that is imagined from multiple settings entered as input data. However, a profile is not limited to text format, and may also be in image format, audio format, or a combination of these. Furthermore, a profile does not have to be true, and may be fiction.

The automatic editing means 240 edits the profile generated by the profile generation means 230. As one example, the automatic editing means 240 may delete predetermined keywords from the profile generated by the profile generation means. As another example, the automatic editing means 240 may replace predetermined keywords in the profile generated by the profile generation means with other words. Examples of keywords include sexual, violent, and discriminatory expressions, setting values included in the input data, and specific personal information (current address, phone number).

In the example of FIG. 8(B), the setting value itself, "fair-skinned maid", is deleted, and the sexual expression "sexy" is replaced with "various". Whether to delete a keyword or replace it with another word may be predetermined, or the decision may be made by AI 16. Furthermore, the automatic editing means 240 may skip executing the edit if a predetermined keyword is not included in the profile.

As shown in FIG. 8(C), the learning means 250 takes as input data "Please create a profile of Linda, female, child, maid, fair-skinned, in 70 to 100 characters." and inputs the learning data, in which the profile edited by the automatic editing means 240 is used as correct answer data, into the trained model 16b for further learning. That is, the learning means 250 transmits the learning data to the AI server 15 via the communication IF 25. The AI server 15 uses the learning data received from the user terminal 20A to cause the trained model 16b to execute a learning process.

The publishing unit 260 associates the profile generated by the profile generating unit 230 or edited by the automatic editing unit 240 with the avatar generated by the object generating unit 210, and publishes it in the online space. More specifically, the publishing means 260 associates the avatar data generated by the profile generation means 230 with the profile data generated by the profile generation means 230 or the automatic editing means 240, and transmits the data to the game server 10 via the communication IF 25. .

The game server 10 transmits the avatar data and profile data received from the user terminal 20A to the user terminals 20B and 20C in response to requests from the user terminals 20B and 20C. As a result, the users of the user terminals 20B and 20C can view the avatar (hereinafter referred to as "Avatar A") and profile (hereinafter referred to as "Profile A") generated by the user terminal 20A on the monitor 31 or It can be perceived through the speaker 36.

As an example, the users of user terminals 20B and 20C visually recognize avatar A that comes into view of their own avatars B and C in the online space. The users of user terminals 20B and 20C also view profile A by selecting avatar A that comes into view. Furthermore, the users of user terminals 20B and 20C view avatar A and profile A through a user selection screen or a user details screen, which will be described later with reference to FIG. 13.

Next, an example of a process for generating an avatar and profile in the user terminal 20B will be described with reference to FIGS. 9 to 11. FIG. 9 is another example of the avatar generation screen. FIG. 10 is another example of the setting parameter priority (A) and setting value histograms (B) and (C). FIG. 11 shows other examples of profiles (A), (B) and learning data (C). Note that a detailed explanation of the common points with the examples of FIGS. 6 to 8 will be omitted, and the explanation will focus on the differences. The examples shown in FIGS. 9 to 11 differ from the examples shown in FIGS. 6 to 8 in the combination of setting values selected by the user through the avatar generation screen.

More specifically, the user of the user terminal 20B selects, as setting values, gender=male, age=adult, contour=round face, hairstyle=mohawk, skin color=dark, clothes=special clothes, and color=white. It shall be assumed that Object generation means 210 generates avatar data according to these setting values, and displays an avatar image indicated by the generated avatar data in avatar display area A1.

As shown in Figs. 10(A) and 10(B), the selection means 220 changes the priority of the hairstyle parameter "mohawk" and the clothing parameter "kamikaze uniform" to "1" because they have been selected the least number of times. Then, in the example of Figs. 9 and 10, the selection means 220 selects gender = male, age = adult, hairstyle = mohawk, and clothing = kamikaze uniform as the setting values to be input to the trained model 16b.

The profile generation means 230 inputs text-format input data "Please generate a profile of Taro, male, adult, Mohawk, special attack uniform, in 70 to 100 characters." to the learned model 16b, and generates the profile in FIG. 11 ( Generate the profile shown in A). The automatic editing means 240 deletes the violent expression "fight" and replaces the violent expression "street fight" with "sports" in the profile shown in FIG. 11(A). Note that the trained model 16b may play the role of the automatic editing means 240. That is, the trained model 16b may have pre-trained a predetermined keyword (in other words, a keyword unsuitable for use in a profile) and generate a profile that does not include the keyword.

As shown in FIG. 11(C), the learning means 250 uses as input data "Please generate a profile of Taro, male, adult, Mohican, special attack uniform, in 70 to 100 characters.", and the automatic editing means 240 Learning data in which the profile edited by is correct data is input to the learned model 16b for further learning.

The publishing means 260 converts the profile generated by the profile generating means 230 or edited by the automatic editing means 240 (hereinafter referred to as "profile B") into the avatar generated by the object generating means 210 (hereinafter referred to as "avatar B"). ) and publish it in the online space.

FIG. 12 is an example of the profile editing screen (A) and learning data (B). The user editing means 270 edits the profile generated by the profile generation means 230 or edited by the automatic editing means 240 according to instructions from the user of the user terminal 20A. More specifically, the user editing means 270 causes the monitor 31 to display a profile editing screen shown in FIG. 12(A).

The profile editing screen includes a profile display area A6, a software keyboard A7, and a [Set] icon. The profile display area A6 is an area where a profile generated by the profile generation means 230 or edited by the automatic editing means 240 is displayed. The software keyboard A7 is an area that accepts user operations to instruct editing (addition, deletion, modification of characters) of the profile displayed in the profile display area A6.

The user editing means 270 edits the profile displayed in the profile display area A6 according to user operations (instructions) via the software keyboard A7. Then, the user editing means 270 finalizes the profile edits in response to the selection of the [OK] icon. Note that, when the user inputs a predetermined keyword (in other words, a keyword that is not suitable for use in the profile), the user editing means 270 may not reflect the keyword in the profile.

Then, the publishing means 260 publishes the profile edited by the user editing means 270 in association with the avatar. Further, as shown in FIG. 12(B), the learning means 250 uses the learning data including the input data generated by the profile generation means 230 and the profile (correct data) edited by the user editing means 270 to model the learned model. 16b for further learning.

Next, referring to FIG. 13, a process in which a user of user terminal 20C views avatars A and B and profiles A and B will be described. Here, a process in which a user selects users to play cooperatively in a cooperative online game will be described as an example. FIG. 13 shows example screens of a user selection screen (A) and a user details screen (B).

The user terminal 20C displays the user selection screen shown in FIG. 13(A) on the monitor 31. The user selection screen displays a list of users who can play cooperatively (e.g., users who are logged in to the online game). The user selection screen displays, for example, the avatar, name, level, and [Details] icon for each user who can play cooperatively. In other words, the user terminal 20C receives the avatars, names, and levels of users who can play cooperatively from the game server 10, and displays them in a list on the user selection screen.

In response to the selection of the [Details] icon corresponding to the name Linda, for example, the user terminal 20C displays the user details screen shown in FIG. 13(B) on the monitor 31. The user details screen is a screen that displays detailed information about the user selected through the user selection screen. The user details screen includes, for example, the selected user's name, level, avatar image, profile, rating, and [Request for Cooperation] icon.

The evaluation items on the user details screen include a [GOOD] icon, a [BAD] icon, the number of times the [GOOD] icon was previously selected (91), and the number of times the [BAD] icon was previously selected and the number of times the [BAD] icon was previously selected and the number of times the [BAD] icon was previously selected and the number of times the [BAD] icon was previously selected and the number of times the [GOOD] icon was previously selected and the number of times the [BAD ...

The feedback means 280 receives the updated number of likes and dislikes from the game server 10 through the communication IF 25. The number of likes and dislikes is an example of an evaluation of the profile. The feedback means 280 then feeds back the evaluation of the profile to the trained model 16b. The feedback means 280 transmits, for example, the number of likes as a score for the profile to the AI server 15 through the communication IF 25. The trained model 16b relearns or additionally learns using the number of likes fed back from the feedback means 280. That is, the trained model 16b becomes more likely to generate a profile that is similar to a profile with a large number of likes and less likely to generate a profile that is similar to a profile with a small number of likes.

The timing and conditions under which the feedback means 280 provides feedback on the evaluation may be, for example, as follows. As one example, the feedback means 280 performs feedback when a predetermined time (e.g., 10 hours) has elapsed since the profile was made public. As another example, the feedback means 280 performs feedback on the condition that the number of likes, the number of dislikes, or the total number of likes and dislikes has reached a predetermined number (e.g., 100). However, the specific method of other users' evaluation of a profile is not limited to the above examples.

[Effects of the embodiment]
According to the above embodiment, a profile of an avatar is generated in the trained model 16b using multiple setting values selected when generating an avatar as input data. This makes it possible to publish a unique profile generated according to an avatar in association with the avatar without increasing the workload of the user.

Furthermore, according to the above embodiment, a unique profile can be generated by selecting a setting value to be input to the trained model 16b from among a plurality of setting values. In particular, a more unique profile can be generated by preferentially inputting setting values that are selected less frequently by other users. However, when all setting values are input to the trained model 16b, the selection means 220 can be omitted.

In addition, according to the above embodiment, by editing the profile generated by the trained model 16b, it is possible to prevent a profile including undesirable expressions from being published in the online space. However, if the trained model 16b is trained to exclude undesirable expressions, the automatic editing means 240 can be omitted.

In addition, according to the above embodiment, by enabling the user to edit the profile, the user's preferred profile can be made public in the online space. Also, compared to having the user create the profile from scratch, the user's workload can be reduced. However, the user editing means 270 can be omitted.

In addition, according to the above embodiment, the profile edited by the automatic editing means 240 and the user editing means 270 can be retrained as correct answer data in the trained model 16b to generate a profile that is more unique, does not include undesirable expressions, and matches the user's preferences. However, when using a general-purpose AI 16 that does not disclose the input interface for training data, the training means 250 can be omitted.

Furthermore, according to the embodiment described above, the evaluations of other users who viewed the profile published in the online space are aggregated and fed back to the learned model 16b, thereby generating a profile that is more suitable for the online space. be able to. However, when using a general-purpose AI 16 whose feedback input interface is not made public, the feedback means 280 can be omitted.

[Other embodiments]
Note that specific examples of objects are not limited to avatars, and may include characters such as animals, monsters, and robots, equipment (e.g., weapons, armor, and accessories), vehicles (e.g., automobiles, ships, aircraft, and spaceships), and buildings (e.g., houses and castles). Specific examples of setting parameters vary depending on the type of object. For example, setting parameters corresponding to a house object include the shape of the roof, the material of the walls, the shape of the windows, and the layout of the house.

Furthermore, the program according to the present invention is not limited to a single program, but may be a collection of multiple programs. Further, the program according to the present invention is not limited to being executed by a single device, but may be shared and executed by a plurality of devices. Furthermore, the role sharing among the game server 10, AI server 15, and user terminal 20 is not limited to the above example. That is, a part of the processing of the game server 10 may be executed by the AI server 15 or the user terminal 20, a part of the processing of the AI server 15 may be executed by the game server 10 or the user terminal 20, A part of the processing of the user terminal 20 may be executed by the game server 10 or the AI server 15.

Further, part or all of the means implemented by the program can also be implemented by hardware such as an integrated circuit. Furthermore, the program may be provided by being recorded on a computer-readable non-transitory recording medium. The recording medium refers to, for example, a hard disk, an SD card, a DVD, and a server on the Internet.

DESCRIPTION OF SYMBOLS 1... System, 2... Communication network, 10... Game server, 15... AI server, 16... Learning model, 16a... Learning model, 16b... Learned model, 17... Neural network, 20... User terminal, 21... Processor, 22 ...Memory, 23...Storage, 23P...Terminal program, 25...Communication interface, 26...Computer, 29...Communication bus, 31...Monitor, 33...Camera, 34...Camera, 35...Microphone, 36...Speaker, 41...Motion sensor , 42... Operating device, 210... Object generation means, 220... Selection means, 230... Profile generation means, 240... Automatic editing means, 250... Learning means, 260... Publishing means, 270... User editing means, 280... Feedback means

Claims (9)

computer,
Object generation means for generating the object according to a plurality of setting values selected by the user to set visual characteristics of the object placed in the online space;
profile generation means for generating a profile of the object by inputting the plurality of setting values into a trained model trained in advance;
A program that functions as a publishing unit that associates the profile generated by the profile generating unit with the object generated by the object generating unit and publishes the profile in the online space. The program according to claim 1,
causing the computer to function as automatic editing means for editing the profile generated by the profile generation means,
The publishing means is a program for publishing the profile edited by the automatic editing means. The program according to claim 2,
The automatic editing means is a program for deleting a predetermined keyword from the profile generated by the profile generation means. The program according to claim 2,
A program that causes the computer to function as a learning means for further learning by inputting learning data in which the plurality of setting values are input data and the profile edited by the automatic editing means is correct data to the trained model. . The program according to claim 1,
A program that causes the computer to function as a selection means for selecting the setting value to be input to the trained model from among the multiple setting values. The program according to claim 5,
The selection means is a program for preferentially selecting, from among the plurality of setting values, the setting value that has been selected less frequently by other users. The program according to claim 1,
The computer,
a user editing means for editing the profile generated by the profile generating means in accordance with an instruction from a user;
The program functions as a learning means for inputting learning data, which has a plurality of the setting values as input data and the profile edited by the user editing means as correct answer data, into the trained model for further training. The program according to claim 1,
A program that causes the computer to function as a feedback means for obtaining evaluations of the profile from other users who have viewed the profile published by the publishing means and providing feedback to the trained model. Object generation means for generating the object according to a plurality of setting values selected by the user to set visual characteristics of the object placed in the online space;
profile generation means for generating a profile of the object by inputting the plurality of setting values into a trained model trained in advance;
A system comprising: publishing means for publishing the profile generated by the profile generating means in the online space in association with the object generated by the object generating means.

Images