JP7493664B1 - Program and system - Google Patents

Abstract

[Problem] To reduce the load of setting up in-game elements such as items handled in a game. [Solution] The program causes a computer to function as: a first generation means for generating unknown story data relating to the story of the game including in-game elements when unknown keywords related to the game and unknown log data related to play of the game are input; an input means for inputting unknown data including the unknown story data and the unknown log data to a first trained model trained using first training data including story data and log data; a second generation means for generating output data relating to the in-game elements based on the unknown data when the unknown data is input; and a progress means for progressing the game in which the in-game elements appear based on the output data. [Selected Figure] Figure 11

Description

The present invention relates to a program and a system.

Conventionally, in computer-based games (hereinafter simply referred to as "games"), a technique is known that provides a user with items to be used in the game depending on how the user plays the game.

For example, when a player wins a casino game within a game, i.e., in a casino in a virtual space, the player acquires virtual currency. When the amount of virtual currency exceeds a predetermined amount, the player is notified. The virtual currency can then be exchanged at an exchange office within the game for items that can be used within the game. In this way, technology is known that motivates players to acquire virtual currency (for example, see Patent Document 1, etc.).

JP 2016-29957 A

However, with conventional technology, when in-game elements such as items are introduced, exchanged, used, granted, bought and sold, improved, or acquired, it is necessary to define settings in advance for each of these situations during the game development stage. In other words, for each in-game element, it is necessary to set up behaviors that correspond to each situation one by one, and the more situations there are, the greater the workload of setting up in-game elements in advance.

The present invention aims to reduce the burden of configuration work related to in-game elements such as items used in the game.

To solve the above problems, the present invention provides a program that causes a computer to function as: a first generation means that generates unknown story data related to the story of the game including in-game elements when unknown keywords related to the game and unknown log data related to the game play are input; an input means that inputs unknown data including the unknown story data and the unknown log data to a first trained model trained using first training data including story data and log data; a second generation means that generates output data related to the in-game elements based on the unknown data when the unknown data is input; and a progression means that progresses the game in which the in-game elements appear based on the output data.

This invention can reduce the burden of configuration work related to in-game elements such as items used in the game.

FIG. 1 is a diagram illustrating an example of a system configuration according to an embodiment of the present invention. FIG. 13 is a diagram illustrating an example of pre-processing. FIG. 11 illustrates an example of an execution process. A diagram showing an example of the overall process of AI learning and execution. FIG. 2 is a diagram illustrating an example of a hardware configuration of an information processing device. FIG. 1 is a network diagram showing an example of an AI configuration. A figure showing an example of a learning process of a first learning model in an RPG. A figure showing an example of a learning process of a second learning model in an RPG. FIG. 11 is a diagram showing an example of an execution process in an RPG. FIG. 13 is a diagram illustrating an example of the entire process. FIG. 2 is a diagram illustrating an example of a functional configuration. FIG. 13 is a diagram showing a configuration example in which an auxiliary device is used.

The following describes the embodiment with reference to the drawings.

[System configuration example]
Fig. 1 is a diagram showing an example of a system configuration according to the present embodiment. For example, as shown in Fig. 1, the system 1 mainly includes user terminals 20A, 20B, and 20C (hereinafter, these may be collectively referred to as "user terminals 20") and a server 11.

Hereinafter, the person who manages the server 11 will be referred to as the "administrator 5." In addition, the people who operate the user terminals 20A, 20B, and 20C will be referred to as the "user 4A," "user 4B," and "user 4C" (hereinafter, these may be collectively referred to as "user 4").

The administrator 5 is a person who operates the information processing service provided by the system 1. On the other hand, the user 4 is a person who uses the information processing service provided by the system 1. The administrator 5 and the user 4 differ in which information processing device they operate: the server 11, which is an example of a management device, or the user terminal 20. Hereinafter, the user 4 will be the player of the game, and the administrator 5 will manage the game (for example, generating, modifying, and distributing the game) and the server 11.

Note that the example shown in FIG. 1 has three user terminals 20 and one server 11, but the number of servers 11, the number of user terminals 20, the number of administrators 5, and the number of users 4 are not important.

The server 11 and the user terminal 20 are connected to each other so that they can communicate with each other via a communication network 2. For example, the communication network 2 is the Internet, a mobile communication system (for example, a public line such as 4G (4th Generation, 4th generation mobile communication standard) or 5G (5th Generation, 5th generation mobile communication standard)), a wireless network such as Wi-Fi (registered trademark), or a combination of these.

The user terminal 20 downloads a program for playing the game (hereinafter referred to as the "game program") from the server 11, or accesses the server 11, which then provides a game service. Note that communication with the server 11 is not required to play the game. In other words, the user terminal 20 may download the program or install it from a medium to create an environment for playing the game. The game may also be played on an information processing device other than the user terminal 20.

[Examples of AI (Artificial Intelligence) Learning and Execution]
Hereinafter, the AI learns through "pre-processing" (sometimes referred to as "learning processing"). The AI in the learning stage, i.e., "pre-processing", is referred to as the "learning model A1". Hereinafter, there are two learning models A1, which are referred to as the "first learning model A11" and the "second learning model A12". Furthermore, the "first learning model A11" and the "second learning model A12" may be collectively referred to as the "learning model A1".

Then, once learning has progressed to a certain extent, the learning model A1 becomes the "trained model A2." Hereinafter, the execution stage in which output processing is executed using the trained model A2 will be referred to as the "execution process." Furthermore, once the "first learning model A11" has trained, it becomes the "first trained model A21." Similarly, once the "second learning model A12" has trained, it becomes the "second trained model A22." Hereinafter, the "first trained model A21" and the "second trained model A22" may be collectively referred to as the "trained model A2."

"Pre-processing" is performed before "execution processing". However, "pre-processing", i.e., trained model A2 may continue to train, before "execution processing" is performed.

[Pre-processing example]
2 is a diagram showing an example of the pre-processing. For example, the pre-processing is performed by the server 11.

The first learning model A11 learns by inputting the first learning data DL1. Similarly, the second learning model A12 learns by inputting the second learning data DL2. In other words, the learning model A1 learns in a so-called "supervised" manner.

The first learning data DL1 includes log data DL12, story data DL11, and first correct answer data DL13.

The first learning data DL1 is data in which the log data DL12 and the story data DL11 are associated with the first correct answer data DL13, which is the "correct answer." In this way, the log data DL12 and the story data DL11 are data in which the correct answer is already known.

Details of the first learning data DL1, story data DL11, log data DL12, and first correct answer data DL13 will be described later.

Through the above-described learning process, when the first learning data DL1 is input, the first learning model A11 learns the correlation between the combination of the log data DL12 and the story data DL11 and the "correct answer" indicated by the first correct answer data DL13.

The second learning data DL2, like the first learning data DL1, includes log data DL12. Therefore, the log data DL12 is used in both the first learning data DL1 and the second learning data DL2 by duplicating or reusing the data.

In addition, the second learning data DL2 includes keywords DL21 and second correct answer data DL22.

Details of the second learning data DL2, the keywords DL21, and the second answer data DL22 will be described later.

The second learning data DL2 is data in which the log data DL12 and the keyword DL21 are associated with the second correct answer data DL22, which is the "correct answer." In this way, the log data DL12 and the keyword DL21 are data for which the correct answer is already known.

The second correct answer data DL22 is data showing the same content as the story data DL11. Therefore, the second correct answer data DL22 is used in both the first learning data DL1 and the second learning data DL2 by being duplicated or reused. However, while the second correct answer data DL22 is correct answer data showing the "correct answer" in the second learning data DL2, in the first learning data DL1, the story data DL11 is input data. In this way, even though the data has the same content, the second correct answer data DL22 and the story data DL11 have different roles in the learning process.

Furthermore, it is preferable that the learning model A1 learns with big data D4. For example, the big data D4 is data on the Internet. However, the big data D4 may also be data input by an administrator 5 or the like. In this way, by learning with the big data D4, it is possible to generate an AI that generates natural sentences and data that resemble those written by a human.

[Example of execution process]
3 is a diagram showing an example of the execution process. For example, the execution process is performed by a plurality of information processing devices, such as the user terminal 20, the server 11, or the user terminal 20 and the server 11, in cooperation with each other.

The trained model A2 is in a state where the trained model A1 has been trained through pre-processing. In other words, when the pre-processing shown in FIG. 2 is executed, the trained model A2 is generated.

When the first unknown data DU1 is input, the first trained model A21 generates output data D3.

The first unknown data DU1 is a combination of log data (hereinafter referred to as "unknown log data DU11") and story data (hereinafter referred to as "unknown story data DU12"), where the "correct answer" is unknown.

Details of the first unknown data DU1, the unknown log data DU11, the unknown story data DU12, and the output data D3 will be described later.

When the second unknown data DU2 is input, the second trained model A22 generates unknown story data DU12.

The second unknown data DU2 is data that is a combination of unknown log data DU11 and a keyword (hereinafter referred to as "unknown keyword DU21"), where the "correct answer" is unknown. Note that the unknown log data DU11 is copied or reused as the first unknown data DU1 and the second unknown data DU2, similar to the log data DL12 in the learning process, to use the same data.

The unknown story data DU12 generated by the second trained model A22 is input to the first trained model A21 as the first unknown data DU1.

The above configuration is a configuration in which the story generator 12 performs pre-processing to generate unknown story data DU12. Note that in the above configuration, the AI is divided into two AIs, one for pre-processing and one for post-processing, but the AI may be divided into one AI or three or more AIs.

When the output data D3 is generated, the server 11 distributes the game to the user terminal 20 based on the output data D3. Note that the game may be transmitted as output data D3 to the user terminal 20. As long as the content of the output data D3 is reflected in the game, the configuration for distributing the game and the data configuration for holding the output data D3 are not important.

Figure 4 shows an example of the overall process of AI learning and execution. The relationship between the pre-processing shown in Figure 2 and the execution process shown in Figure 3 is as shown in Figure 4.

Note that the pre-processing and execution processing do not have to be performed in the consecutive order as illustrated in the figure. Therefore, it is not necessary to consecutively perform the period in which preparation is performed by the pre-processing and the period in which the execution processing is performed thereafter. Therefore, once the trained model A2 has been created, the execution processing may be performed after a period of time has elapsed since the pre-processing. Furthermore, once the trained model A2 has been generated, the execution processing may be performed by repurposing the trained model A2.

The first learning data DL1 and the first unknown data DU1 are different in the learning process and the execution process. Furthermore, the second learning data DL2 and the second unknown data DU2 are different in the learning process and the execution process. Also, in the learning stage, the AI is a learning model A1, but once learning has progressed to a certain extent, it becomes a trained model A2. In this way, the trained model A2 that has been trained using big data D4 and the like as training data is what is known as a "generative AI."

The "correct answer" of each piece of learning data is known, whereas the "correct answer" of each piece of unknown data is unknown. Specifically, since the first learning data DL1 includes the first correct answer data DL13, the first correct answer data DL13 is associated with the combination of the log data DL12 and the story data DL11, whereas the first unknown data DU1 does not include the first correct answer data DL13. Therefore, in the learning process, the relationship between the combination of the log data DL12 and the story data DL11 and the first correct answer data DL13 is known. Therefore, when the first learning model A11 inputs the first learning data DL1, it can learn the correlation between the combination of the log data DL12 and the story data DL11 and the "correct answer."

Similarly, the second learning data DL2 includes the second correct answer data DL22, and therefore the second correct answer data DL22 is associated with the combination of the log data DL12 and the keyword DL21, whereas the second unknown data DU2 does not include the second correct answer data DL22. Therefore, in the learning process, the relationship between the combination of the log data DL12 and the keyword DL21 and the second correct answer data DL22 is known. Therefore, when the second learning model A12 receives the second learning data DL2, it can learn the correlation between the combination of the log data DL12 and the keyword DL21 and the "correct answer."

On the other hand, the first unknown data DU1 does not include the first correct answer data DL13, and the "correct answer" for the first unknown data DU1 is unknown. The first trained model A21 then generates output data D3 for the first unknown data DU1 based on the correlation between the combination of the log data DL12 and story data DL11 trained in pre-processing and the first correct answer data DL13.

Similarly, the second unknown data DU2 does not include the second correct answer data DL22, and the "correct answer" for the second unknown data DU2 is unknown. The second trained model A22 generates unknown story data DU12 for the second unknown data DU2 based on the correlation between the combination of the log data DL12 and the keyword DL21 trained in pre-processing and the second correct answer data DL22.

The execution process may be a process in which a part of the execution uses a table or the like. In this way, in a configuration that uses a table, that is, a so-called rule-based configuration, the pre-processing is a process that prepares for inputting a table (also called a look-up table (LUT) or a formula, etc.).

[Example of hardware configuration of information processing device]
5 is a hardware configuration diagram of an information processing device. The information processing device is a server 11, a user terminal 20, etc. Hereinafter, it is assumed that the information processing device has the same hardware configuration as the server 11. For example, the information processing device is a workstation or a general-purpose computer such as a personal computer. However, each information processing device may have a different hardware configuration.

The server 11 mainly comprises a processor 111, a memory 112, a storage 113, an input/output interface 114, and a communication interface 115. In addition, each component of the server 11 is connected to a communication bus 116.

The processor 111 performs processing and control by executing a series of instructions contained in the server program 11P stored in the memory 112 or storage 113.

The processor 111 is, for example, a calculation device and a control device such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an MPU (Micro Processing Unit), an FPGA (Field-Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), or a combination of these.

The memory 112 is a main storage device that stores the server program 11P and data, etc. For example, the server program 11P is loaded from the storage 113. The data includes data input to the server 11 and data generated by the processor 111. For example, the memory 112 is a RAM (Random Access Memory) or other volatile memory.

Storage 113 is an auxiliary storage device that stores server program 11P and data, etc. Storage 113 is, for example, a ROM (Read-Only Memory), a hard disk drive, a flash memory, or other non-volatile storage device. Storage 113 may also be a removable storage device such as a memory card. As another example, storage 113 may be an external storage device. With this configuration, for example, in a situation where multiple user terminals 20 are used, such as an amusement facility, it becomes possible to collectively update server program 11P or data.

The input/output interface 114 is an interface that connects external devices such as a monitor, an input device (e.g., a keyboard or a pointing device), an external storage device, a speaker, a camera, a microphone, and a sensor to the server 11.

The processor 111 also communicates with external devices through the input/output interface 114. The input/output interface 114 is, for example, a Universal Serial Bus (USB), a Digital Visual Interface (DVI), a High-Definition Multimedia Interface (HDMI (registered trademark)), wireless, and other terminals.

The communication interface 115 communicates with other devices (such as the user terminal 20) connected to the communication network 2. For example, the communication interface 115 is a wired communication interface such as a LAN (Local Area Network), or a wireless communication interface such as Wi-Fi (registered trademark) (Wireless Fidelity), Bluetooth (registered trademark), or NFC (Near Field Communication).

However, the information processing device is not limited to the above hardware configuration. For example, the user terminal 20 may further include a sensor such as a camera. Various data acquired by the user terminal 20 through the sensor may be transmitted to the server 11.

[Examples of the configuration of learning models and trained models]
6 is a network diagram showing an example of an AI configuration. The learning model A1 and the trained model A2 are, for example, an AI having a configuration shown in the following network.

In the following, the learning model A1 and the trained model A2 are described as being implemented on the server 11, i.e., on the cloud. However, part or all of the learning model A1 and the trained model A2 may be implemented on the user terminal 20, etc.

The network 300 has, for example, an input layer L1, an intermediate layer L2 (also called a "hidden layer"), and an output layer L3.

The input layer L1 is the layer that inputs data.

The intermediate layer L2 converts the data input in the input layer L1 based on weights, biases, etc. The results of processing in this way in the intermediate layer L2 are transmitted to the output layer L3.

The output layer L3 is the layer that outputs the output contents, etc.

Then, the weighting coefficients and the parameters to be changed by learning are optimized through learning. Note that the network 300 is not limited to the network structure shown in the figure. In other words, the AI may be realized by other machine learning methods.

For example, the AI may be configured to perform preprocessing such as dimensionality reduction (for example, a process that converts a relationship of three or more dimensions into a relationship that can be determined by simple calculations of three dimensions or less) using "unsupervised" machine learning. It is desirable to process the relationship between input and output using simple calculations such as linear expressions. Such calculations can reduce computational costs.

The AI may also undergo processing to reduce overfitting (also called "overfitting" or "over-fitting"), such as dropout. Other preprocessing such as dimensionality reduction and normalization may also be performed.

The AI may have a network structure such as a CNN (Convolution Neural Network). In addition, for example, the network structure may have a configuration such as an LLM (Large Language Model), an RNN (Recurrent Neural Network), or an LSTM (Long Short-Term Memory). In other words, the AI may have a network structure other than deep learning.

Also, the AI may be configured to have hyperparameters. In other words, the AI may be configured such that some of the settings are performed by a user or the like. Furthermore, the AI may specify the features to be learned, or the user may set some or all of the features to be learned.

Furthermore, the learning model A1 and the trained model A2 may use other machine learning. For example, the learning model A1 and the trained model A2 may be unsupervised models, and normalization or the like may be performed in preprocessing. Furthermore, learning may be reinforcement learning (a learning method in which an AI is made to make a choice and an evaluation (reward) is given for the choice, so that the evaluation becomes large), etc.

In learning, data expansion, etc. may be performed. That is, in order to increase the amount of learning data used in learning the learning model A1, pre-processing may be performed to expand a single piece of experimental data, etc., into multiple pieces of learning data. Increasing the amount of learning data in this way allows the learning of the learning model A1 to progress further.

The learning model A1 and the trained model A2 may be configured to perform transfer learning or fine tuning. That is, since the user terminal 20 often has a different execution environment for each device, the settings may be different for each device according to the execution environment. For example, the basic configuration of the AI is trained on another information processing device. After that, each information processing device may be trained or configured additionally to optimize it for its respective execution environment.

[Example of learning process in RPG (Role-Playing Game)]
The following will be described using an example of an RPG. In an RPG, as the game progresses, events occur and various elements appear. Hereinafter, elements that appear within the game will be referred to as "in-game elements."

Specifically, in-game elements include items, characters, skills, magic, etc. that appear in the story of the game.Specifically, in-game elements include various elements that make up the game, such as items, characters, skills, magic, etc. that appear in the story of the game.In-game elements may be objects that can be operated during play, such as items or skills, or they may not be objects that can be operated by the user 4.

For example, items are one of the in-game elements, and include event items necessary for events that progress through the game, attack items that can attack enemies, recovery items that restore a character's physical strength or abnormal conditions, and auxiliary items that improve various parameters such as a character's attack power, defense power, and speed (providing a so-called buff effect) or that lower various enemy parameters (providing a so-called debuff effect, etc.).

Note that items are not limited to the above, and may include equipment such as weapons, armor, and accessories, as well as cards that make up a deck, and various material items used for item synthesis or monster synthesis. Items may also include items that are not necessarily required for game progress and cannot be used within the game (for example, medals or badges that recognize the achievement of a certain condition within the game, such as defeating a certain enemy (meaning defeating a certain number of enemies, etc.)).

Characters are people, animals, robots, etc. in the game, and are elements that trigger events to occur as the game progresses, join the protagonist and form a party, or make up the story. Characters include playable characters that the player can control, and non-playable characters (so-called NPCs) that the player cannot control. Characters are not necessarily limited to real characters, and can be objects that are set up to fit the worldview of the game, such as demi-humans, beastmen, and androids.

Skills (techniques) or magic (the names vary depending on the game) are elements that include commands and the like that are activated based on the operation of the user 4 and allow the user to gain an advantage in battles and the like within the game. Note that skills and magic are not necessarily limited to those that provide an advantage, and include those that strengthen some parameters while weakening others, such as fire resistance +50 but ice resistance -50. They also include so-called "passive skills" that are activated without the user being prompted to perform an operation.

The following description will be given using an example in which the in-game element is an "item." In the following example, output data D3 indicates what the item is (for example, the specifications of what effect will be produced when it is used), and with output data D3 it is possible to make the item appear in the game being played on user terminal 20.

[Examples of first learning data, story data, log data, and first correct answer data]
7 is a diagram showing an example of a learning process of a first learning model in an RPG. For example, the following data is used for learning the first learning model A11.

For example, when a game is played, the log data DL12 is generated in accordance with the progress of the game. Note that the log data DL12 may be settings and plans for progressing through the game before the game is started. On the other hand, when the game has progressed partway through, the log data DL12 is a record of the game from the start of the game to the progress of the game.

Log data DL12 is a data configuration that includes, for example, image data DL121 and profile data DL122. However, each piece of data does not have to be in a separate file, and a single piece of data may include multiple contents. Log data DL12 also indicates various settings in the game.

The image data DL121 is data showing, for example, an output screen (including a screen not yet disclosed to the user 4), the appearance of a character, or the appearance of various items. Note that the image data DL121 may be subjected to image analysis to extract parameters.

The profile data DL122 indicates the profile of a character appearing in the game or of user 4. Which character or user 4's profile the profile data DL122 indicates varies depending on the game settings, input items, etc.

For example, the profile data DL122 is input by the user 4 when starting the game. However, the profile data DL122 may be set in advance rather than set by the user 4, or may be obtained from the user terminal 20 or another database, etc.

Specifically, the profile data DL122 includes items such as name, place of origin, sex, age, occupation, and friendships. The types of profile items are set in advance.

The name may be, for example, the real name, abbreviation, or pseudonym of the user 4. The name may also be identification information such as a nickname set for the character. Similarly, the place of origin, gender, age, occupation, and friendships may be settings of the user 4, or may be fictitious settings in the game.

Story data DL11 is, for example, text data that shows the story in sentences. However, story data DL11 may also include images, etc.

The first correct answer data DL13 is data that indicates the items to be featured in the game. Specifically, the first correct answer data DL13 is specifications such as the item's appearance, name, effect when used, etc. The first correct answer data DL13, i.e., the specifications for making an item appear, differs depending on the game. Therefore, the first correct answer data DL13 is data that includes items required for an item to appear in the game.

By carrying out the above-mentioned learning, the first learning model A11 learns the correlation of what kind of item to generate from the story, profile, etc. As described above, when generating an item, if a profile, image, etc. are input as auxiliary data in addition to the story, the information in the auxiliary data is taken into consideration when generating the item.

[Examples of second learning data, keywords, log data, and second correct answer data]
8 is a diagram showing an example of a learning process of the second learning model in the RPG. For example, the following data is used for learning the second learning model A12.

The log data DL12 is, for example, similar to the first learning model A11, i.e., the learning process shown in FIG. 7. However, the log data DL12 may have different data content or a different amount of data from the learning process of the first learning model A11.

Keywords DL21 are words or sentences that are used as references when generating story data DL11. For example, keywords DL21 are the meanings of in-game elements, characteristics of in-game elements, events that occur to characters, elements that are desired to appear in the story, or key points of the story.

The story data DL11 is, for example, text data that shows the story in sentences. In other words, the story data DL11 is similar to the learning process shown in FIG. 7.

For example, when keyword DL21 is input, the second learning model A12 learns to generate a story that includes keyword DL21. Note that keyword DL21 does not have to be directly included in story data DL11, and keyword DL21 may be indirectly expressed in the story.

The story is also generated based on the log data DL12. For example, if the parameters in the profile, such as "gender" or "origin" in the profile data DL122, are different, the story data DL11 is generated so that the effect, specifications, type, name, or appearance of the item changes depending on the parameter.

By carrying out the above-mentioned learning, the second learning model A12 learns the correlation of what kind of story to generate from the keywords DL21 and the profile, etc. As described above, when generating a story, if peripheral information, images, etc. are input as auxiliary data in addition to the profile data DL122, the information in the auxiliary data is taken into consideration when generating the story.

It should be noted that story generation may be achieved by a configuration other than AI. For example, when fixing the type of item to be featured based on "gender" or "age" or the like, the fixing conditions (e.g., type of item) may be input from the profile data DL122 to the first trained model A21 (e.g., input as auxiliary data) without going through AI. In this way, when fixed conditions can be input, the items to be featured can be made to appear under specific conditions corresponding to these in a certain profile.

Specifically, if the "occupation" is a specific job type, the item will always be a "weapon" item, and on the other hand, if the job type is different, the item will always be an "armor" item.

[Example of execution process in RPG]
9 is a diagram showing an example of an execution process in an RPG. After a trained model A2 is generated by the learning process shown in FIG. 7 and FIG. 8, the execution process is executed as follows.

The second unknown data DU2, i.e., the unknown log data DU11 and the unknown keyword DU21, are unknown at the time the data is input. In the execution process, the second trained model A22 generates unknown story data DU12 based on the correlation between the learned combination of the keyword DL21 and the log data DL12 and the second correct answer data DL22, and taking into account the combination of the unknown log data DU11 and the unknown keyword DU21. After being generated by the second trained model A22, the unknown story data DU12 is input to the first trained model A21.

In addition, after the unknown story data DU12 is generated, the administrator 5 or the like may edit the story before it is input to the first trained model A21.

Next, unknown log data DU11 and unknown story data DU12 are input as first unknown data DU1 to the first trained model A21. The correct answer for the first unknown data DU1, i.e., the combination of unknown log data DU11 and unknown story data DU12, is unknown at the time the data is input.

Then, in the execution process, the first trained model A21 generates output data D3 based on the trained combination of story data DL11 and log data DL12 and the correlation between the first correct answer data DL13, taking into account the combination of unknown log data DU11 and unknown story data DU12.

When the output data D3 is generated, a game featuring the item generated based on the output data D3 can be played. For example, the game can be played on a user terminal 20, etc.

[Overall processing example]
10 is a diagram showing an example of the overall process. In the following example, the overall process is a sequence of a pre-process and an execution process. Specifically, steps S11 to S14 are the learning process. Steps S21 to S24 are the execution process. However, the overall process may include other processes.

In the following, steps S11 and S12, and steps S13 and S14, i.e., the learning process of the first learning model A11 and the learning process of the second learning model A12, are performed in parallel. However, these learning processes may be performed one after the other, rather than in parallel. For example, the second learning model A12 may be trained first to become the second trained model A22, and then the learning process of the first learning model A11 may be performed using the second trained model A22.

In step S11, the server 11 inputs the first learning data DL1.

In step S12, the server 11 inputs the first learning data DL1 and trains the first learning model A11 to generate the first trained model A21.

In step S13, the server 11 inputs the second learning data DL2.

In step S14, the server 11 inputs the second learning data DL2 and trains the second learning model A12 to generate the second trained model A22.

When steps S11 to S14 are performed, the learning process shown in FIG. 7 and FIG. 8 is performed. As described above, when the learning process is performed to a certain extent, the learning model A1 learns and becomes the trained model A2. Next, the execution process is performed as follows using the trained model A2 generated by the learning process.

In step S21, the server 11 inputs the second unknown data DU2 into the second trained model A22.

In step S22, the server 11 generates unknown story data DU12 using the second trained model A22.

In step S23, the server 11 generates output data D3 using the first trained model A21. Thereafter, the server 11 provides the game to the user terminal 20 based on the output data D3.

In step S24, the server 11 progresses the game so that in-game elements appear based on the output data D3.

[Functional configuration example]
11 is a diagram illustrating an example of a functional configuration. For example, the system 1 is a generative AI system including a learning device 61 and an execution device 62.

The learning device 61 includes a story data generation means 1F11, a first learning data input means 1F12, and a learning means 1F13.

When the keyword DL21 and the log data DL12 are input, the story data generation means 1F11 performs a story data generation procedure to generate the story data DL11. For example, the story data generation means 1F11 is realized by a story generator 12, etc.

The first learning data input means 1F12 performs a first learning data input procedure to input the first learning data DL1 including the story data DL11, the log data DL12, and the first correct answer data DL13. For example, the first learning data input means 1F12 is realized by the communication interface 115, etc.

The learning means 1F13 performs a learning procedure in which the first learning model A11 is trained using the first learning data DL1 to generate the first trained model A21. For example, the learning means 1F13 is realized by the processor 111, etc.

The execution device 62 includes an unknown story data generation means 1F21, a first unknown data input means 1F22, an execution means 1F23, and a progress means 1F24. The unknown story data generation means 1F21 is an example of a first generation means. The execution means 1F23 is an example of a second generation means and an output data generation means.

When unknown keywords DU21 and unknown log data DU11 are input, the unknown story data generating means 1F21 performs an unknown story data generating procedure to generate unknown story data DU12. For example, the unknown story data generating means 1F21 is realized by a story generator 12 or the like.

The first unknown data input means 1F22 performs a first unknown data input procedure for inputting the first unknown data DU1 including the unknown story data DU12 and the unknown log data DU11. For example, the first unknown data input means 1F22 is realized by the communication interface 115 or the like.

The execution means 1F23 performs an execution procedure to generate output data D3 using the first trained model A21. For example, the execution means 1F23 is realized by the processor 111, etc.

The progression means 1F24 performs a progression procedure for progressing the game on the user terminal 20 based on the output data D3. For example, the progression means 1F24 is realized by the processor 111, etc.

The learning device 61 and the execution device 62 are, for example, the server 11. In this way, the learning device 61 and the execution device 62 may be a single information processing device. However, the learning device 61 and the execution device 62 may be different information processing devices.

With the above configuration, when keywords that are the basis of the story and log data indicating the surrounding settings for generating the story are input, a story is generated that matches the characters, etc. Then, in-game elements such as items are generated that correspond to the story and the surrounding settings, etc. Specifically, a story is generated, and items that appear in the story can be made to actually appear in the game. In this way, even if the in-game elements that appear in the game are arranged differently depending on various factors such as the characters, since items are generated to match the story, the load of the setting work related to the in-game elements handled in the game can be reduced.

In addition, items are generated based on the character's profile, making them unique items. For example, if the game is played repeatedly or by different players, different stories and items can be generated for each playthrough.

Furthermore, compared to a method in which items are prepared in advance, a wider variety of items can be provided. If the story or profile, etc., changes, the items also change, so even if user 4 plays the game repeatedly, user 4 will not get bored of the game. In this way, an attractive game can be provided to user 4.

[Modification]
The type of game is not limited to a novel game. For example, the type of game may be a game including RPG, sports, action, puzzle, or fighting. Furthermore, the game may not be one of the above games.

For example, the in-game element may be a character, etc. Therefore, the characters that appear in the game may change depending on the story, profile, etc.

In the above description, the unknown story data DU12 is generated by the second trained model A22, but the embodiment is not limited to this. For example, the unknown story data DU12 may be input to the first trained model A21 by an administrator 5 or the like inputting text to create a sentence indicating a story (including a story that is improved based on a story generated by AI).

Therefore, the story generator 12 is not limited to a configuration using AI, but may be a device that executes a program to assist the administrator 5 or the like in editing text.

The story generator 12 may be the same as the server 11, or may be an information processing device separate from the server 11.

[Other embodiments]
In the above example, the information processing device performs both pre-processing for the learning model and execution processing using the learned model. However, the pre-processing and execution processing do not have to be performed by the same information processing device. In addition, the pre-processing and execution processing do not have to be consistently performed by one information processing device. In other words, each process and data storage, etc. may be performed by an information system, etc., composed of multiple information processing devices.

The learning process may be additionally performed after the execution process or before the execution process.

The above-mentioned processing may be performed auxiliary by an information processing device other than the server 11 and the user terminal 20.

Figure 12 is a diagram showing an example of a configuration using an auxiliary device. Compared to the example shown in Figure 1, the configuration shown in Figure 12 differs in that an auxiliary device 60 is added. Note that the auxiliary device 60 may be a configuration that is used temporarily.

The auxiliary device 60 is an information processing device that is installed near the user terminal 20 (in this example, it is installed near the user terminal 20A, but it may be installed near another device). The auxiliary device 60 then executes a specific process in part or in whole on behalf of the user terminal 20 or the server 11.

For example, the auxiliary device 60 is equipped with a device specialized for graphic processing and performs graphic processing at high speed. In this way, so-called edge computing may be performed by installing the auxiliary device 60 or the like. In this way, the above-mentioned processing may be executed by utilizing the hardware resources of various information processing devices. Therefore, the above-mentioned processing may be executed by an information processing device different from the one described above.

The above-described processes and the data used in the processes executed in this embodiment may be executed and stored by an information processing system. For example, the information processing system may execute or store the processes or storage in a redundant, distributed, parallel, or combination thereof manner in multiple information processing devices. Therefore, the present invention may be realized in devices other than the hardware configurations shown above and in systems other than the devices shown above.

The program according to the present invention is not limited to a single program, but may be a collection of multiple programs. The program according to the present invention is not limited to being executed by a single device, but may be shared and executed by multiple information processing devices. Furthermore, the division of roles among the information processing devices is not limited to the above example. In other words, some or all of the above-mentioned processes may be executed by an information processing device different from the above-mentioned information processing device.

Furthermore, a part or all of the means realized by the program can be realized by hardware such as an integrated circuit. Furthermore, the program may be provided by being recorded on a non-transitory recording medium that can be read by a computer. A recording medium refers to, for example, a hard disk, an SD card (registered trademark), an optical disk such as a DVD, or a server on the Internet. Therefore, the program may be distributed via a telecommunications line such as the Internet.

In addition, the information processing devices that make up the information processing system may be located overseas. In other words, the information processing devices that perform some of the processes performed by the information processing system may be located overseas.

The present invention is not limited to the above-mentioned embodiments. Therefore, the present invention can be modified or have additional components added without departing from the technical gist of the invention. Therefore, all technical matters included in the technical ideas described in the claims are covered by the present invention. The above-mentioned embodiments are preferred examples. Furthermore, a person skilled in the art can realize various modifications from the disclosed contents, and such modifications are included in the technical scope described in the claims.

1: System 1F11: Story data generation means 1F12: First learning data input means 1F13: Learning means 1F21: Unknown story data generation means 1F22: First unknown data input means 1F23: Execution means 1F24: Progression means 4: User 5: Administrator 11: Server 12: Story generator 20: User terminal 60: Auxiliary device 61: Learning device 62: Execution device A1: Learning model A11: First learning model A12: Second learning model A2: Trained model A21: First trained model A22: Second trained model D1: First unknown data D2: Second unknown data D3: Output data D4: Big data DL1: First learning data DL11: Story data DL12: Log data DL121: Image data DL122: Profile data DL13 : First correct answer data DL2 : Second learning data DL21 : Keyword DL22 : Second correct answer data DU1 : First unknown data DU11 : Unknown log data DU12 : Unknown story data DU2 : Second unknown data DU21 : Unknown keyword

Claims (4)

Computer,
a first generating means for generating unknown story data relating to a story of the game including an in-game element when unknown keywords relating to a game and unknown log data relating to a play of the game are input;
an input means for inputting unknown data including unknown story data and unknown log data to a first trained model trained using first training data including story data and log data;
a second generation means for generating output data relating to the in-game element based on the unknown data when the unknown data is input;
A program that functions as a progression means for progressing the game in which the in-game element appears based on the output data. The unknown log data is
2. The program according to claim 1, further comprising image data of an output screen in the game, or profile data relating to a character appearing in the game, or a user's profile. The first generating means is
using a second trained model obtained by training a second learning model using second learning data including keywords related to the game, the log data, and first correct answer data related to a story of the game;
The second trained model is
2. The program according to claim 1, which is a generation AI that is trained to generate the story data when the unknown log data and the unknown keyword are input. a story data generating means for generating story data relating to a story of the game including in-game elements when a keyword relating to the game and log data relating to play of the game are input;
a first learning data input means for inputting first learning data including the story data and the log data;
A learning means for learning a first learning model using the first learning data to generate a first trained model;
an unknown story data generating means for generating unknown story data regarding a story of the game including the in-game elements when unknown keywords related to the game and unknown log data related to play of the game are input;
an unknown data input means for inputting unknown data including the unknown story data and the unknown log data to the first trained model;
an output data generating means for generating output data relating to the in-game element based on the unknown data when the unknown data is input;
and a progression means for progressing the game in which the in-game element appears based on the output data.

Images