Information processing system, information processing method and program

Abstract

To allow an action of a character to be determined.SOLUTION: An information processing system comprises a conversation acquisition unit that acquires a conversation text of a user, and an action estimation unit that provides the conversation text to a machine learning model and estimates an action to be performed by a character that is either a player or an NPC during a game.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an information processing system, an information processing method, and a program.

Background Art

[0002] Patent Document 1 discloses making a player character speak lines according to the game situation.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] If the character's actions do not match the lines, a sense of discomfort will be felt.

[0005] The present invention has been made in view of such a background, and an object thereof is to provide a technology capable of determining the actions of a character.

Means for Solving the Problems

[0006] The main invention of the present invention for solving the above problems is an information processing system, comprising: a conversation acquisition unit that acquires a user's conversation text; and an action estimation unit that gives the conversation text to a machine learning model to estimate actions to be performed by a character who is a player or an NPC during a game.

[0007] Regarding other problems disclosed in the present application and their solutions, they will be clarified by the embodiments of the invention and the drawings.

Effects of the Invention

[0008] According to the present invention, the actions of characters can be determined.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Modes for Carrying Out the Invention

[0010] <First Embodiment> Hereinafter, an information processing system according to the first embodiment of the present invention will be described. The information processing system of this embodiment attempts to automatically determine the actions of characters appearing during a game. Characters can include players (characters operated by users) and NPCs (Non Player Characters; characters not operated by users). In the information processing system of this embodiment, the actions of characters are determined according to the conversations uttered by the characters (those input by the user in voice or text, or those automatically generated by the system for NPCs), and control is performed so that the characters execute the determined actions. The data for operating the characters (action data) can be in any format as long as it can realize, for example, the emotions and motions of the characters.

[0011] FIG. 1 is a diagram showing an example of the overall configuration of an information processing system. The information processing system of the present embodiment includes a management server 2. The management server is communicably connected to the game server 1 via a communication network. The communication network 3 is, for example, the Internet and is constructed by a public telephone line network, a mobile phone line network, a wireless communication path, Ethernet (registered trademark), or the like.

[0012] The game server 1 is a computer that executes games. A user can operate a user terminal 3 such as a smartphone, a game console, or a personal computer to access the game server 1 and play games. In the game server 1, position information, movement history, and conversation data history of characters and NPCs operated by the user are managed.

[0013] The management server 2 may be a general-purpose computer such as a workstation or a personal computer, or may be logically realized by cloud computing.

[0014] <Management Server> FIG. 2 is a diagram showing an example of the hardware configuration of the management server 2. Note that the illustrated configuration is an example, and other configurations may be adopted. The management server 2 includes a CPU 201, a memory 202, a storage device 203, a communication interface 204, an input device 205, and an output device 206. The storage device 203 stores various data and programs, such as a hard disk drive, a solid state drive, or a flash memory. The communication interface 204 is an interface for connecting to the communication network 3, such as an adapter for connecting to Ethernet (registered trademark), a modem for connecting to a public telephone network, a wireless communication device for performing wireless communication, a USB (Universal Serial Bus) connector or an RS232C connector for serial communication. The input device 205 inputs data, such as a keyboard, a mouse, a touch panel, a button, or a microphone. The output device 206 outputs data, such as a display, a printer, or a speaker. Note that each functional unit of the management server 2 described later is realized by the CPU 201 reading a program stored in the storage device 203 into the memory 202 and executing it, and each storage unit of the management server 2 is realized as a part of the storage area provided by the memory 202 and the storage device 203.

[0015] FIG. 3 is a diagram showing an example of the software configuration of the management server 2. The management server 2 includes a machine learning model storage unit 231, an operation data storage unit 232, a conversation acquisition unit 211, an operation estimation unit 212, an operation processing unit 213, a context information acquisition unit 214, and a personality information acquisition unit 215.

[0016] The machine learning model storage unit 231 stores a machine learning model created by machine learning. The machine learning model stored in the machine learning model storage unit 231 can be, for example, a large language model. Also, a pre-trained machine learning model (such as a large language model) may be a model fine-tuned using conversation text and information representing actions as training data. The machine learning model may be constructed by, for example, a neural network or the like. The machine learning model storage unit 231 may not be provided in the management server 2 but may be provided in an external server, and the management server 2 may be configured to use the machine learning model through an API or the like provided by the external server.

[0017] The action data storage unit 232 stores action data. The action data storage unit 232 can store action data for each action generated by the machine learning model. The action data can include, for example, a series of commands for instructing the actions of a character (which may be, for example, the rotation or movement of bones and joints).

[0018] The conversation acquisition unit 211 acquires the conversation text of the character. The conversation text can be, for example, a chat message input by the user operating the player during the game. Also, the conversation text may be lines generated by the game server 1 for the player and / or NPC. Also, the conversation acquisition unit 211 may give a prompt to the machine learning model to generate the lines of the character.

[0019] The motion estimation unit 212 estimates the motions of the character. The motion estimation unit 212 estimates the motions of the character that match the character's lines. The motion estimation unit 212 can estimate the motions of the character by providing the conversation text acquired by the conversation acquisition unit 211 and an instruction to generate the motions that the character should perform during the game to a machine learning model. The motion estimation unit 212 may estimate one motion corresponding to the conversation text, or may estimate a series of motions corresponding to the conversation text. The motion estimation unit 212 may, for example, estimate motions as physical expressions performed without the character moving, or may, for example, estimate the actions of the character in the game space, such as moving to another location within the game space or changing the conversation partner.

[0020] The motion processing unit 213 performs control to cause the character to perform motions. The motion processing unit 213 performs control to cause the character to perform the motions estimated by the motion estimation unit 212. The motion processing unit 213 reads out motion data corresponding to the motions estimated by the motion estimation unit 212 from the motion data storage unit 232 and transmits the read motion data to the game server 1, thereby causing the character to move during the game.

[0021] The context information acquisition unit 214 acquires context information regarding the conversation text. The context information may be information indicating the state of the character during the game. The context information may include, for example, the history of past conversations by the character, the relationship with other characters (which may be only the characters currently in the vicinity), geographical information, events experienced by the character (such as having an argument, being in an argument, failing an exam, winning the lottery, completing a quest, etc.), the situation of the character (such as being with a lover), the current position of the character, the memory of the character, and the like. The context information acquisition unit 214 can, for example, access the game server 1 to acquire the context information. The context information can be, for example, the progress of the game, the time information of the scene, the objects placed in the scene, the equipment and possessions of the character, the equipment and possessions of other characters present in the scene, the lines spoken by other characters, and the like.

[0022] The conversation acquisition unit 211 can include the context information in the prompt and provide it to the machine learning model to generate the lines of the character.

[0023] The action estimation unit 212 can provide the conversation text and the context information to the machine learning model to estimate the action. The action estimation unit 212 can provide a prompt to the machine learning model to estimate the actions of the character according to the context information.

[0024] The personality information acquisition unit 215 acquires the personality information of the user and / or character. The personality information can include, for example, the age, gender, personality, emotions, etc. of the user and / or character. The personality information can include, for example, the demographic attributes, psychographic attributes, geographic attributes, behavioral attributes, etc. of the user and / or character, or can include information indicating other classifications of the user and / or character. The personality information acquisition unit 215 can access the game server 1 and acquire the personality information of the user and / or character grasped in the game server 1. Note that the personality information acquisition unit 215 can, for example, acquire the behavioral attributes based on the browsing history of the user's web browser or the tendency of online shopping from a web server (not shown). Also, the management server 2 may be provided with a personality information management unit that manages the personality information. The personality information acquisition unit 215 can, for example, accept the input of text data explaining the personality of the user and / or character, and give the received text data to a machine learning model to generate personality information. The personality information acquisition unit 215 can also give the conversation text and / or context information and the personality information to a machine learning model and instruct to generate updated personality information. Here, the conversation text to be given may include not only the conversation content uttered by the character but also the conversation content spoken to the character by other characters. For example, according to the context in the game environment, or according to the conversation text uttered by the character and / or received from other characters, it is possible to generate personality information updated so that the emotions of the character change, traumas occur, or the personality changes. The personality information acquisition unit 215 can update the personality information of the character based on the generated personality information. The personality information acquisition unit 215 can, for example, access the game server 1 or the personality information management unit and update the managed personality information.

[0025] The motion estimation unit 212 can estimate motions by providing the conversation text and personality information to a machine learning model. Therefore, the estimated motions can also be changed according to the update of the personality information. Thus, even for the same character and the same context, different motions can be set according to the change in the personality information of the character.

[0026] The motion estimation unit 212 can also estimate motions by providing the conversation text, context information, and personality information to a machine learning model.

[0027] <Motion> FIG. 4 is a diagram for explaining the operation of the management server 2.

[0028] The management server 2 acquires the conversation text, context information, and personality information of the user / character (S301), provides a prompt including the conversation text, context information, and personality information to a machine learning model, and estimates the motion of the character (S302). The management server 2 performs control to cause the character to execute the estimated motion (S303).

[0029] As described above, according to the information processing system of the present embodiment, motions corresponding to the lines of the character can be automatically caused to be performed by the character.

[0030] <Second Embodiment> In the first embodiment, it has been described that motions are generated by providing the conversation text to a machine learning model. In the second embodiment, more specific generation of motions will be defined.

[0031] In the second embodiment, each conversation expression (which can be a word, phrase, sentence, etc.) that can be included in the conversation text is pre-labeled. Also, operation data for causing a character to perform an operation is prepared in advance, and the above labels are also associated with each piece of operation data. Note that different labels may be set for each item included in personality information such as gender and age, or both a label indicating the type of action and a label indicating the item may be generated with the item itself included in the personality information as the label.

[0032] FIG. 5 is a diagram showing an example of the software configuration of the management server 2 according to the second embodiment. As shown in FIG. 5, the management server 2 can further include a learning processing unit 216. In the second embodiment, the machine learning model is assumed to be a large language model. The learning processing unit 216 extracts conversation expressions from the conversation text and selects the label (which may be plural) set for the extracted conversation expression for each of them.

[0033] The action estimation unit 212 can cause the machine learning model to generate a label corresponding to the conversation text by providing the machine learning model with the conversation text and a prompt instructing to generate a label. The action estimation unit 212 can acquire the action data associated with the label from the action data storage unit 232. Also, the action estimation unit 212 may cause the pre-trained learning model (generative AI) that generates action data to generate action data by giving a label to the learning model. The learning model that generates action data can utilize a known generative AI for motion generation.

[0034] <Third Embodiment> In the third embodiment, a label is generated for the machine learning model (large language model). The label represents the action of a character (player or NPC).

[0035] The machine learning model can be made to select from a set of simple and general labels that represent actions, which are preset. In this case, labels such as "feeling happy", "doing push-ups", "riding a bike", etc. can be prepared in advance. The action estimation unit 212 can give a prompt, which is set with the conversation text and an instruction to select a label from the prepared label options, to the machine learning model to select one or a series of labels.

[0036] Action data may be associated with the labels in advance. In this case, the action data storage unit 232 can store the action data in association with the labels. Note that the action data storage unit 232 may associate a plurality of labels with one piece of action data, or may associate a plurality of pieces of action data with one label. The action estimation unit 212 can obtain the action data corresponding to the label generated by the machine learning model from the action data storage unit 232. When a plurality of pieces of action data correspond to one label, the action estimation unit 212 can select one of the corresponding pieces of action data by an arbitrary method (for example, randomly).

[0037] Alternatively, the machine learning model may be made to generate labels. In this case, an instruction may be given to provide information about the personality and context of the character. For example, the machine learning model can generate a label such as "The person is old, he has a pain in his knees, and he is feeling happy". The action estimation unit 212 can give a prompt including the conversation text and an instruction to generate a label to the machine learning model to generate one or a series of labels.

[0038] When generating labels for a machine learning model, it is possible to generate motion data from the labels using a machine learning model (which may be the same as or different from the machine learning model that generates the labels). The learning processing unit 216 can cause the machine learning model to learn, by machine learning, the motion data corresponding to the labels. The motion estimation unit 212 can acquire, from the motion data storage unit 232, the motion data associated with the labels. Further, the motion estimation unit 212 may cause the motion data to be generated by giving labels to a pre-trained learning model (generative AI) that generates the motion data. The learning model that generates the motion data can utilize a known generative AI for motion generation. In this case, by giving labels including personality and / or context to the generative AI, it becomes possible to match the generated motion data to the personality and context of the character.

[0039] <Fourth Embodiment> In the fourth embodiment, conversation text is generated using a machine learning model that is a generator such as a large language model.

[0040] The conversation acquisition unit 211 can cause the machine learning model to generate conversation text by giving the context information acquired by the context information acquisition unit 214 to the machine learning model. Further, the conversation acquisition unit 211 may cause conversation data to be generated by giving context information and personality information to the machine learning model.

[0041] The motion estimation unit 212 can give the generated conversation text to the machine learning model to generate a label representing the motion, and read out, from the motion data storage unit 232, the motion data corresponding to the generated label.

[0042] In addition, the motion estimation unit 212 can acquire motion data associated with the label from the motion data storage unit 232. Further, the motion estimation unit 212 may cause the motion data to be generated by giving a label to a pre-trained learning model (generative AI) that generates motion data. As the learning model for generating motion data, a known generative AI for motion generation can be used. Further, the motion estimation unit 212 may cause the motion data to be generated by giving the conversation text to a machine learning model.

[0043] As described above, the present embodiment has been described. However, the above embodiment is for facilitating the understanding of the present invention and is not for limiting the interpretation of the present invention. The present invention can be changed and improved without departing from its gist, and equivalents thereof are also included in the present invention.

[0044] <Disclosed matters> Note that the present disclosure also includes the following configurations. [Item 1] A conversation acquisition unit that acquires a user's conversation text, An operation estimation unit that gives the conversation text to a machine learning model to estimate an operation to be performed by a character who is a player or an NPC during the game, An information processing system characterized by comprising: [Item 2] The information processing system according to Item 1, Comprising an operation processing unit that performs control for causing the character to perform the operation, An information processing system characterized by comprising: [Item 3] The information processing system according to Item 1, Comprising a context information acquisition unit that acquires context information regarding the conversation text, The operation estimation unit estimates the operation by giving the conversation text and the context information to the machine learning model, An information processing system characterized by comprising: [Item 4] The information processing system according to Item 1, comprising a personality information acquisition unit that acquires the personality information of the character, wherein the action estimation unit gives the conversation text and the personality information to the machine learning model to estimate the action, characterized by the information processing system. [Item 5] The information processing system according to Item 4, wherein the personality information acquisition unit receives an input of text data explaining the personality of the character, gives the received text data to the machine learning model to generate the personality information, and acquires the generated personality information, characterized by the information processing system. [Item 6] The information processing system according to Item 3, comprising a personality information acquisition unit that acquires the personality information of the character, wherein the action estimation unit gives the conversation text, the context information, and the personality information to the machine learning model to estimate the action, characterized by the information processing system. [Item 7] The information processing system according to Item 6, wherein the personality information acquisition unit gives at least any one of the conversation text and the context information and the personality information to the machine learning model to generate the updated personality information, and updates the personality information according to the generation result, characterized by the information processing system. [Item 8] The information processing system according to Item 1, wherein a predetermined label is set for the conversation expression, and at least any one of the labels is set for the action data indicating the action prepared in advance, A learning processing unit that uses, as training data, the conversation text and the labels corresponding to the conversation expressions corresponding to each of the partial texts included in the conversation text, and trains the machine learning model by machine learning. The action estimation unit causes the machine learning model to be given the conversation text to infer the label. An information processing system characterized by the above. [Item 9] The information processing system according to Item 8, The action estimation unit causes the machine learning model to be given the conversation text to select a label representing the action from predetermined labels, and acquires action data for causing the character to perform the action, which is pre-associated with the selected label. An information processing system characterized by the above. [Item 10] The information processing system according to Item 8, The machine learning model is a large language model, The action estimation unit causes the machine learning model to be given the conversation text to select one of predetermined labels pre-associated with the action, representing the action of the character. The action estimation unit causes the label to be given to a trained learning model for generating the action data, thereby generating action data for causing the character to perform the action. An information processing system characterized by the above. [Item 11] The information processing system according to Item 8, The machine learning model is a large language model, The action estimation unit causes the machine learning model to be given the conversation text to generate a label representing the action of the character and the context regarding the character and the conversation text. The action estimation unit causes the label to be given to a trained learning model for generating the action data, thereby generating action data for causing the character to perform the action. An information processing system characterized by [Item 12] The information processing system according to Item 1, comprising a context information acquisition unit that acquires context information regarding the character, wherein the conversation acquisition unit provides the context information to the machine learning model to generate the conversation text, and the action estimation unit provides the conversation text to the machine learning model to cause the machine learning model to select a label representing the action from predetermined labels, and acquires action data for causing the character to perform the action, which is pre-associated with the selected label. An information processing system characterized by [Item 13] The information processing system according to Item 1, comprising a context information acquisition unit that acquires context information regarding the character, wherein the conversation acquisition unit provides the context information to the machine learning model to generate the conversation text, and the action estimation unit provides the conversation text to the machine learning model to generate a label representing the action, and provides the generated label to the machine learning model to generate action data for causing the character to perform the action. An information processing system characterized by [Item 14] The information processing system according to Item 13, comprising a personality information acquisition unit that acquires personality information of the character, wherein the conversation acquisition unit provides the context and the personality information to the machine learning model to generate the conversation data. An information processing system characterized by [Item 15] The step of acquiring the user's conversation text, the step of providing the conversation text to a machine learning model to estimate an action to be performed by a character who is a player or an NPC during the game, An information processing method characterized by being executed by a computer. [Item 16] The step of obtaining the user's conversation text, The step of giving the conversation text to a machine learning model to estimate the actions to be performed by a character who is a player or NPC during the game, A program for causing a computer to execute.

Explanation of symbols

[0045] 1 User terminal 2 Management server

Claims

1. A conversation acquisition unit that acquires a user's conversation text; An action estimation unit that gives the conversation text to a machine learning model and estimates an action to be performed by a character who is a player or NPC during the game; An information processing system, characterized by comprising the above.

2. The information processing system according to Claim 1, further comprising an action processing unit that performs control for causing the character to perform the action; An information processing system, characterized by comprising the above.

3. The information processing system according to Claim 1, further comprising a context information acquisition unit that acquires context information regarding the conversation text, wherein the action estimation unit gives the conversation text and the context information to the machine learning model to estimate the action; An information processing system, characterized by comprising the above.

4. The information processing system according to Claim 1, further comprising a personality information acquisition unit that acquires personality information of the character, wherein the action estimation unit gives the conversation text and the personality information to the machine learning model to estimate the action; An information processing system, characterized by comprising the above.

5. The information processing system according to Claim 4, wherein the personality information acquisition unit receives an input of text data explaining the personality of the character, gives the received text data to the machine learning model to generate the personality information, and acquires the generated personality information; An information processing system, characterized by comprising the above.

6. The information processing system according to Claim 3, It is provided with a personality information acquisition unit that acquires the personality information of the character. The action estimation unit estimates the action by providing the conversation text, the context information, and the personality information to the machine learning model. An information processing system characterized by the above.

7. The information processing system according to claim 6, The personality information acquisition unit provides at least one of the conversation text and the context information and the personality information to the machine learning model to generate the updated personality information, and updates the personality information according to the generation result. An information processing system characterized by the above.

8. The information processing system according to claim 1, A predetermined label is set for the conversation expression. At least one of the labels is set for the action data indicating the action prepared in advance. It is provided with a learning processing unit that uses the conversation text and the label corresponding to the conversation expression corresponding to each of the partial texts included in the conversation text as training data to train the machine learning model by machine learning. The action estimation unit causes the machine learning model to infer the label by providing the conversation text to the machine learning model. An information processing system characterized by the above.

9. The information processing system according to claim 8, The action estimation unit provides the conversation text to the machine learning model to generate a label representing the action, and acquires action data for causing the character to perform the action, which is associated with the generated label in advance. An information processing system characterized by the above.

10. The information processing system according to claim 8, The machine learning model is a large language model, The action estimation unit provides the conversation text to the machine learning model to cause the machine learning model to select one of predetermined labels pre-associated with the action, which represents the action of the character, The action estimation unit generates action data for causing the character to perform the action by providing the label to a trained learning model for generating the action data, An information processing system characterized by the above.

11. An information processing system according to claim 9, The machine learning model is a large language model, The action estimation unit provides the conversation text to the machine learning model to generate a label representing at least one of the action of the character and the context regarding the character and the conversation text, The action estimation unit generates action data for causing the character to perform the action by providing the label to a trained learning model for generating the action data, An information processing system characterized by the above.

12. An information processing system according to claim 1, Comprising a context information acquisition unit that acquires context information regarding the character, The conversation acquisition unit provides the context information to the machine learning model to generate the conversation text, The action estimation unit provides the conversation text to the machine learning model to cause the machine learning model to select a label representing the action from predetermined labels, and acquires action data for causing the character to perform the action, which is pre-associated with the selected label, An information processing system characterized by the above.

13. An information processing system according to claim 1, A context information acquisition unit that acquires context information regarding the character is provided. The conversation acquisition unit gives the context information to the machine learning model to generate the conversation text. The motion estimation unit gives the conversation text to the machine learning model to generate a label representing the motion, and gives the generated label to the machine learning model to generate motion data for causing the character to perform the motion. An information processing system characterized by the above.

14. An information processing system according to claim 13, A personality information acquisition unit that acquires personality information of the character is provided. The conversation acquisition unit gives the context and the personality information to the machine learning model to generate the conversation data. An information processing system characterized by the above.

15. The step of acquiring the user's conversation text, The step of giving the conversation text to a machine learning model to estimate the actions to be performed by a character who is a player or an NPC during the game, An information processing method characterized by the computer executing the above.

16. The step of acquiring the user's conversation text, The step of giving the conversation text to a machine learning model to estimate the actions to be performed by a character who is a player or an NPC during the game, A program for causing a computer to execute the above.

Images