Generation system, generation method, and generation program

Abstract

The purpose of the present invention is to provide a new technique for generating a story in which characters behave in a human-like manner.　This generation system for generating a story includes a plan generation unit, a detailed action generation unit, and a story generation unit. The plan generation unit generates a plan for actions of characters in the story taken after a predetermined period, on the basis of story information including an outline of the story and predetermined period action information related to the actions of the characters taken during the predetermined period. The detailed action generation unit generates detailed actions related to the actions of the characters taken at each point in time, which are more detailed than the actions in the plan, on the basis of the plan and emotion parameters related to emotions of the characters. The story generation unit generates the story on the basis of the detailed actions.

Description

Generation System, Generation Method, and Generation Program

[0001] The present invention relates to a generation system, a generation method, and a generation program.

[0002] Conventionally, there are technologies that assist in creating stories.

[0003] For example, Patent Document 1 discloses a technology for outputting plot data based on information including an operator's input.

[0004] Japanese Patent Application Laid-Open No. 2023-23978

[0005] By the way, recently, there has been an increasing demand to enjoy stories that one can empathize with the characters in the story. However, with the technology of Patent Document 1, while a story can be generated from an operator's input, it is not possible to generate a story in which characters with human-like behaviors that one can empathize with appear.

[0006] The present invention has been made in view of the above circumstances, and an object to be solved is to provide a new technology for generating a story in which characters perform human-like behaviors.

[0007] [1] A generation system for generating a story, the generation system comprising: a plan generation unit, a detailed action generation unit, and a story generation unit, the plan generation unit generating a plan of a character in a predetermined period after the predetermined period based on story information including an outline of the story and predetermined period action information regarding the actions of the characters in the story, the detailed action generation unit generating detailed actions regarding the actions at each time more detailed than the plan of the character based on the plan and an emotion parameter regarding the emotion of the character, and the story generation unit generating the story based on the detailed actions. Generation system.

[0008] [9] A generation method to be performed by a generation system for generating a story, the generation system comprising: a plan generation unit; a detailed action generation unit; and a story generation unit, the method comprising: the plan generation unit generating a plan for a character after a predetermined period based on story information including an outline of the story and predetermined period action information relating to the character's actions during a predetermined period; the detailed action generation unit generating detailed actions relating to the character's actions at each time that are more detailed than the plan, based on the plan and emotional parameters relating to the character's emotions; and the story generation unit generating the story based on the detailed actions.

[0009]

[10] A generation program for generating a story, wherein a computer functions as a plan generation unit, a detailed action generation unit, and a story generation unit, the plan generation unit generates a plan for a character after a predetermined period based on story information including an outline of the story and predetermined period action information relating to the character's actions during a predetermined period, the detailed action generation unit generates detailed actions relating to the character's actions at each time point, which are more detailed than the plan, based on the plan and emotion parameters relating to the character's emotions, and the story generation unit generates the story based on the detailed actions.

[0010] This structure allows the characters in the story to behave in a more human-like manner. Humans sometimes change their behavior based on their emotions (for example, their mood) on any given day. By generating the characters' plans in advance and then generating the characters' actions based on those plans and their emotions, it becomes possible to create human-like behavior that is driven by emotions.

[0011] [2] The generation system according to [1], wherein the story information includes information about locations appearing in the story, the plan generation unit generates a plan including the information about the locations, and the detailed action generation unit generates detailed actions, including interactions between multiple characters, on a time-by-time basis, based on the plans of multiple characters.

[0012] This configuration generates a plan that includes the locations where the characters act, allowing conversations to occur between characters acting in the same location.

[0013] [3] The detailed action generation unit generates detailed actions that are more detailed than the plan at each time based on the plan, the emotional parameters, and the desire parameters relating to the desires of the characters, and further changes the emotional parameters and / or desire parameters based on the detailed actions, as described in [1] or [2].

[0014] This structure allows the characters to behave in a more human way, as they act based on their emotions and desires, often deviating from the plans they initially made. Since people sometimes change plans they made the day before or later, depending on their emotions and desires, this structure allows the characters to exhibit such behavior.

[0015] [4] The detailed action generation unit generates detailed actions that are more detailed than the plan at each time interval, based on the plan, the emotional parameters, and the desire parameters relating to the desires of the characters, and further changes the emotional parameters and / or desire parameters at regular intervals, as described in [1] or [2].

[0016] This structure allows the characters' emotions and / or desires to change over time, resulting in more human-like behavior.

[0017] [5] The generation system according to any one of [1] to [4], wherein the story generation unit generates the detailed actions as a single scene of the story based on the time and relationship of the detailed actions.

[0018] This structure makes it possible to combine the detailed actions of the characters into a single scene, preventing the story from becoming redundant.

[0019] [6] The story generation unit generates the story according to any one of [1] to [4], wherein, based on conflict information relating to the conflict of the character and the plan, it extracts events in the plan in which the conflict of the character is depicted, further extracts events in the plan that occur after those events as the conclusion of the story, further extracts events in the plan that occur before those events as the beginning of the story, and generates the story based on the results of the extraction.

[0020] This structure makes it possible to generate stories that focus on the characters' internal conflicts, allowing users to empathize with the characters who are struggling with these issues.

[0021] [7] The generation system according to any one of [1] to [6], wherein the plan generation unit, the detailed action generation unit, and the story generation unit are all trained learning models.

[0022] This structure allows for the generation of appropriate plans for the characters in the story, appropriate actions based on those plans and the characters' emotions, and an appropriate story based on those actions.

[0023] [8] The generation system according to any one of [1] to [7], comprising a generation unit and a model determination unit, wherein the generation unit receives the story information, and the model determination unit determines the language of the received story information and determines a learning model corresponding to the determined language.

[0024] This configuration makes it possible to generate stories using a learning model that corresponds to the language of the story summary and other information contained in the narrative data, and to generate stories that reflect the characteristics of that learning model. For example, a learning model trained using Japanese data may generate stories that reflect Japanese culture, influenced by the Japanese data used for training, while a learning model trained using English data may generate stories that reflect English-speaking culture, influenced by the English data used for training.

[0025] According to the present invention, it is possible to provide a new technology for generating stories in which characters behave in a human-like manner.

[0026] A block diagram showing the configuration of the generation system in this embodiment. A hardware configuration diagram in this embodiment. An example of the data configuration stored in the memory unit in Embodiment 1. An example of the data configuration stored in the memory unit in Embodiment 1. An example of the data configuration stored in the memory unit in Embodiment 1. A flowchart of the process for generating a story in Embodiment 1. An example of the data configuration stored in the memory unit in Embodiment 2. An example of the data configuration stored in the memory unit in Embodiment 2. An example of the data configuration stored in the memory unit in Embodiment 2. An example of the data configuration stored in the memory unit in Embodiment 2. An example of the data configuration stored in the memory unit in Embodiment 2. An example of the data configuration stored in the memory unit in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation screen in Embodiment 2. An example of the story generation process in Embodiment 2. An example of the story generation process in Embodiment 2. An example of a flowchart for the story generation process in Embodiment 2. An example of a flowchart for the story generation process in Embodiment 2. An example of a flowchart for the story generation process in Embodiment 2. An example of a flowchart for the story generation process in Embodiment 2. An example of a flowchart for the story generation process in Embodiment 2. An example of a flowchart for the story generation process in Embodiment 2. An example of a flowchart for the story generation process in Embodiment 2. An example of a flowchart for the story generation process in Embodiment 2. An example of a flowchart for the story generation process in Embodiment 2.

[0027] The generation system of the present invention will be described below with reference to the drawings. Preferred embodiments are shown in the drawings. However, the present invention can be carried out in many different forms and is not limited to the embodiments described herein.

[0028] For example, while this embodiment describes the configuration and operation of the generation system, similar effects can be achieved by the execution method (steps), apparatus, computer program, etc. The program in this embodiment may be provided as a non-transient recording medium readable by a computer, or it may be provided as a downloadable service from an external server, or the program may be launched on an external computer to perform its functions on a client terminal (so-called cloud computing).

[0029] Furthermore, in this embodiment, "part" may include, for example, hardware resources implemented by circuits in a broad sense, and information processing of software that can be specifically realized by these hardware resources. In this embodiment, "information" can be represented, for example, by the physical values ​​of signal values ​​representing voltage and current, the high or low values ​​of signal values ​​as a set of binary bits composed of 0s or 1s, or by quantum superposition (so-called qubits), and communication and calculations can be performed on circuits in a broad sense.

[0030] In a broad sense, a circuit is a circuit realized by appropriately combining circuits, circuits, processors, and memory. That is, it includes CPUs (Central Processing Units), GPUs (Graphics Processing Units), LSIs (Large Scale Integration), ASICs (Application Specific Integrated Circuits), FPGAs (Field-Programmable Gate Arrays), SoCs (System on a Chip), etc.

[0031] <System Overview> Figure 1 is a block diagram showing the configuration of the generation system in this embodiment. As shown in Figure 1, the generation system 0 comprises a generation device 1, a plan generation device 2, a detailed action generation device 3, a story generation device 4, a setter terminal 5, and a user terminal 6. The generation device 1 is configured to communicate with the plan generation device 2, the detailed action generation device 3, the story generation device 4, the setter terminal 5, and the user terminal 6 via a network NW. The generation device 1 operates as a server, and the setter terminal 5 and the user terminal 6 operate as client terminals.

[0032] The generator 1 receives information via the user terminal 5, including the story title, the story's logline (story outline or summary, etc.), the story's genre, keywords related to the story, locations appearing in the story, events that occur in the story, the theme of each chapter in the story, the number of days elapsed in the story, or the start date and time of the story.

[0033] The generation device 1 receives requests regarding stories via the user terminal 6. The generation device 1 may generate plans, actions, and stories based on some or all of the information received via the setter terminal 5 and / or the user terminal 6. Alternatively, as in this embodiment, some or all of the received information may be transmitted to an external device (for example, a plan generation device 2, a detailed action generation device 3, a story generation device 4, etc.) to generate the plans, actions, and stories.

[0034] The generation device 1 can be a general-purpose server computer or a personal computer, etc. It is also possible to configure the generation device 1 using multiple computers.

[0035] The plan generation device 2 includes a plan generation unit 21 as a component and generates plans for the characters. The detailed action generation device 3 includes a detailed action generation unit 31 as a component and generates detailed actions for the characters. The story generation device 4 includes a story generation unit 41 and generates a story.

[0036] The plan generation device 2, the detailed action generation device 3, and the story generation device 4 are server devices that provide a trained learning model, and the plan generation unit 21, the detailed action generation unit 31, and the story generation unit 41 are thought to be trained learning models. The plan generation device 2, the detailed action generation device 3, and the story generation device 4 are thought to be server devices that provide a natural language processing (NLP) model.

[0037] The plan generation device 2, the detailed action generation device 3, and the story generation device 4 are preferably server devices that provide a large language model (LLM). The natural language processing model enables the processing of data input as natural language by a computer. In this embodiment, the type of natural language processing model employed is not limited, but examples include ChatGPT (Chat Generative Pre-trained Transformer).

[0038] The plan generation unit 21, the detailed action generation unit 31, and the narrative generation unit 41 may be implemented by a large-scale language model, a dedicated model specialized for the processing in question, or a combination thereof. These may be implemented by software or firmware.

[0039] The plan generation device 2, the detailed action generation device 3, and the story generation device 4 may each be multiple devices, for example, they may be server devices that provide learning models learned in different languages. Furthermore, the components of the plan generation device 2, the detailed action generation device 3, and the story generation device 4 may be located in the generation device 1, or in one or more devices configured to communicate with the generation device 1. In addition, a single functional component may perform plan generation, detailed action generation, and story generation.

[0040] The configurator terminal 5 is a terminal used by the configurator, who sets the story settings in order to generate the story, to input information related to the story to be generated. The configurator terminal 5 can be a smartphone, tablet, personal computer, or other terminal device. There may be multiple configurator terminals 5.

[0041] The user terminal 6 is a terminal for a user who enjoys the generated story to input and send a request for the generated story. As the user terminal 6, terminal devices such as smartphones, tablet terminals, and personal computers can be used. There may be a plurality of user terminals 6.

[0042] In the present embodiment, the network NW is an IP (Internet Protocol) network, but there is no limitation on the type of communication protocol, and further, there is no limitation on the type and scale of the network.

[0043] <Hardware Configuration> FIG. 2 is a hardware configuration diagram. As shown in FIG. 2(a), the information processing apparatus 10 (generation apparatus 1) includes a control unit 101, a storage unit 102, and a communication unit 103, and is used to exert the functions of each unit and each process.

[0044] The control unit 101 includes one or more processors such as a CPU (Central Processing Unit), and controls the overall operation processing of the information processing apparatus 10 by executing the generation program according to the present invention, an OS (Operating System), browser software, middleware, and other applications.

[0045] The storage unit 102 is an HDD (Hard Disk Drive), SSD (Solid State Drive), ROM (Read Only Memory), RAM (Random Access Memory), etc., and stores the generation program according to the present invention and data used when the control unit 101 executes processing based on the program. By the control unit 101 executing processing based on the generation program stored in the storage unit 102, the functional configuration described later is realized.

[0046] The communication unit 103 executes communication control with the network NW, and performs inputs necessary for operating the information processing apparatus 10 and outputs related to the operation results.

[0047] As shown in FIG. 2(b), the terminal device 9 (the setter terminal 5 and the user terminal 6) includes a control unit 91, a storage unit 92, a communication unit 93, an input unit 94, and an output unit 95, and is used to exert the functions of each unit and each process.

[0048] The control unit 91 of the terminal device 9 includes one or more processors such as a CPU, and controls the overall operation processing of the terminal device 9. The storage unit 92 of the terminal device 9 is an HDD, SSD, ROM, RAM, etc., and stores the above-mentioned application and data used when the control unit 91 executes processing based on a program.

[0049] The communication unit 93 of the terminal device 9 controls communication with the network NW. The input unit 94 of the terminal device 9 is a mouse, keyboard, etc., and inputs an operation request by the user / provider to the control unit 91. The output unit 95 of the terminal device 9 is a display, etc., and displays the result of the processing of the control unit 91.

[0050] <Functional Components> As shown in FIG. 1, the generation device 1 includes a generation unit 11, a transmission unit 12, a model determination unit 13, and a determination unit 14. The plan generation device 2, the detailed action generation device 3, and the story generation device 4 each include a plan generation unit 21, a detailed action generation unit 31, and a story generation unit 41.

[0051] The arrangement of these functional components is an example, and it is also possible to configure the generation system 0 by implementing these functional components on a plurality of computers. For example, a part of the functional configuration of the generation device 1 may be arranged in one or more devices configured to be communicable with the plan generation device 2, the detailed action generation device 3, the story generation device 4, the setter terminal 5, the user terminal 6, and the generation device 1. Also, the functional components of the plan generation device 2, the detailed action generation device 3, and the story generation device 4 may be arranged in the generation device 1.

[0052] <Data Configuration> FIGS. 3 to 5 are an example of the data configuration stored in the storage unit in Embodiment ١. The generation device 1 receives story information related to the story necessary for generating a story via the setter terminal 5. The generation unit 11 of the generation device 1 may store the received story information in the storage unit.

[0053] Narrative information is information necessary for generating a narrative, and includes some or all of the following: narrative setting information regarding the setting of the narrative; character setting information regarding the settings of the characters appearing in the narrative; character relationship information regarding the relationships between the characters appearing in the narrative; world structure information regarding the settings of the locations appearing in the narrative; chapter information regarding the chapters of the narrative (divides in the narrative (e.g., divisions between scenes, divisions between characters, etc.)); parameter information regarding the parameters of the characters; and so on.

[0054] The arrangement of each data is merely an example, and some or all of the data stored in the memory unit of generation device 1 may also be stored in one or more devices configured to communicate with the plan generation device 2, detailed action generation device 3, story generation device 4, setter terminal 5, user terminal 6, and generation device 1.

[0055] Story setting information is information about the setting of the story to be generated, and includes the story title, story genre, keywords related to the story, story logline, etc., and is managed by a story ID as shown in Figure 3(a).

[0056] Character setting information is information about the settings of characters appearing in the story, and includes the character's name, date of birth, role (role in the story, etc.), motivation (motives for actions in the story, etc.), personality (innate personality, personality towards close friends, personality towards disliked friends, etc.), physical characteristics, speaking style, lifestyle, family structure, etc., and is managed by a character ID as shown in Figure 3(b).

[0057] In addition to the above, character setting information may also include information about the characters' internal conflicts, their past (background, etc.), and their current situation.

[0058] Character relationship information is information about the relationships between characters appearing in the story, and includes the character ID, the relationship with the character, and the level of intimacy indicating the closeness with the character, and is managed by the character ID as shown in Figure 4(c).

[0059] World structure information is information used to set up the world in a story, and includes a hierarchical structure consisting of world information, sector information, arena information, and object information, and is managed by a story ID as shown in Figure 4(d).

[0060] World information refers to information about the setting of the story, within which the story unfolds. Sector information and arena information refer to information about locations that appear in the setting of the world information. Object information refers to things that exist in the real world (for example, tangible objects).

[0061] Parameter information refers to parameters related to the characters, specifically parameters related to the characters' emotions, desires, etc. The parameter information includes emotional parameters (including disgust, fear, sadness, surprise, pleasure, anger, neutrality, etc.) and desire parameters (including hunger (fullness), socialization, enjoyment, health, physical strength, etc.), and is managed by the character ID, as shown in Figure 5(e).

[0062] The emotional parameters may include multiple parameters related to a character's emotions, and the character will perform detailed actions based on these emotional parameters. These emotional parameters may include emotions such as joy and resignation.

[0063] Furthermore, characters may perform detailed actions based on their need parameters. For example, need parameters are parameters other than emotions that determine a character's detailed actions. Need parameters may include multiple parameters related to a character's needs, and may also include parameters such as sleepiness.

[0064] The parameter information stored in the memory unit may be updated periodically. When the plan generation unit 21, detailed action generation unit 31, story generation unit 41, etc. change parameters and the generation unit 11 receives the changed parameters, the generation unit 11 stores the changed parameters in the memory unit.

[0065] Chapter information refers to information about a chapter in a story, including the chapter title, characters, theme (in the chapter), events that occur, story ID, etc., and is managed by the chapter ID, as shown in Figure 5(f). In addition, chapter information may also include a description of the theme, information about locations that appear in the chapter, changes in the characters' emotions (parameters), conflict information regarding the characters' conflicts, keywords in the chapter, etc.

[0066] Conflict information refers to information about the characters' internal struggles, and is highly likely to have a significant impact on their actions within the story. Specifically, conflicts relate to the characters' worries and anxieties within the story, such as gaps between themselves and those around them.

[0067] For example, one might feel a gap between themselves and their peers, such as seeing other students at their school striving towards their goals while they themselves lack any particular goals. Since many people have worries or anxieties, they are more likely to empathize with characters in stories who are also struggling with similar issues.

[0068] <Processing Flowchart> Figure 6 is a flowchart of the process for generating the character plans in Embodiment 1.

[0069] <Receiving and transmitting information for generating a plan> In step S601, the generation unit 11 receives story information necessary for generating a story (for example, some or all of story setting information, character setting information, character relationship information, world structure information, parameter information, chapter information, etc.) via the setter terminal 5.

[0070] In step S602, the transmission unit 12 transmits part or all of the information received by the generation unit 11 to the plan generation device 2. The generation unit 11 may also generate information to be transmitted to the plan generation device 2 based on the story information.

[0071] <Plan Generation> In step S603, the plan generation unit 21 generates a plan for the characters based on the story information. Based on the information received from the transmission unit 12, the plan generation unit 21 generates a plan for a predetermined period (for example, one day (24 hours), several days, one month, etc.) that is set as the period for generating the character's plan. In order to generate a rough plan for the character over the predetermined period, the plan generation unit 21 divides the predetermined period (for example, so that the number of divided periods is three or more).

[0072] Specifically, the plan generation unit 21 divides a predetermined period for generating a plan into shorter periods and generates a plan for each divided period. For example, when the plan generation unit 21 generates a plan for a character's day, it divides that day into shorter periods (e.g., several tens of minutes, one hour, several hours, etc.) and generates plans for each period.

[0073] The generation unit 11 may receive a predetermined period and / or a shorter period for the plan to be generated by the plan generation unit 21 via the user terminal 5, and the transmission unit 12 may transmit that predetermined period and / or a shorter period to the plan generation device 2. The plan generation unit 21 generates the plan based on the period received from the transmission unit 12.

[0074] When the plan generation unit 21 generates a 30-minute schedule for the character ID "CH0001" in Figure 3(b), the plan generation unit 21 generates a schedule that includes events such as "0:00 (until 0:29) sleeping", "0:30 (until 0:59) sleeping", "1:00 (until 1:29) sleeping", ..., "7:00 (until 7:29) waking up", "7:30 (until 7:59) breakfast", "8:00 (until 8:29) going to school", "8:30 (until 8:59) arriving at high school", ..., "23:00 (until 23:29) going to bed", "23:30 (until 23:59) sleeping".

[0075] In this embodiment, "sleeping" is not limited to cases where it is generated as a special event, but also includes treating any continuous period of time without scheduled events in the plan for that date as sleep. Hereafter, the term "sleeping" in this specification shall include the mode in which such unplanned periods of time are considered as sleep.

[0076] The plan generation unit 21 generates plans for each character based on the character setting information received. In this way, plans for all characters appearing in the story are generated, allowing the detailed action generation unit 31 (described later) to simulate the detailed actions of each character.

[0077] The plan generation unit 21 generates a plan for a character after a predetermined period based on the information received from the transmission unit 12 and predetermined period action information (information about the character's actions in the past (before the plan being generated)) regarding the character's actions during that predetermined period. For example, when the plan generation unit 21 generates a plan for a character's day, if there is no information about the plan and / or detailed actions (described later) for days prior to that day, the plan generation unit 21 may generate the same plan as the previous day.

[0078] The plan generation unit 21 can prevent itself from generating the same plan as in the past by using information about actions during a predetermined period to generate a plan for the period after that predetermined period. The information about actions during a predetermined period used by the plan generation unit 21 may be the plan and / or detailed actions themselves for the past predetermined period, or a summary thereof.

[0079] The plan generation unit 21 may further generate a plan based on chapter information. By generating a story composed of multiple chapters, each with its own theme, it is possible to generate a story that readers can enjoy.

[0080] The plan generation unit 21 generates a plan for the characters in the chapter based on information such as the characters appearing in the chapter, themes, and events that occur in that chapter. The plan generation unit 21 may also generate a story for the next chapter based on information from previous chapters (e.g., all past chapters, the chapter immediately preceding, etc.) (e.g., summaries, events that occurred, etc.).

[0081] By utilizing information from previous chapters when generating a plan for a particular chapter of the story, the plan generation unit 21 can generate a consistent story that also reflects the content of previous chapters.

[0082] The plan generation unit 21 may store the generated character's plan as the character's memory. Specifically, the plan generation unit 21 stores the generated plan in the memory unit, linked to the character ID. Furthermore, the plan generation unit 21 may also store the time period in the story that the plan takes place in, linked to the plan (for example, the number of days elapsed in the story, the date in the story, a timestamp, etc.).

[0083] Furthermore, the plan generation unit 21 may store each event included in the generated plan (for example, events at specific times) in the memory unit as a character's memory. When storing an event, the plan generation unit 21 may estimate the importance of that event and store it associated with that importance. In addition, the plan generation unit 21 may store the time period in the story when the event in the plan occurs (for example, elapsed time in the story, date and time in the story, timestamp, etc.) associated with that event.

[0084] Furthermore, the plan generation unit 21 may estimate the importance of events related to a character (for example, events being planned by other characters with whom the character has a high level of closeness, events being planned by other characters involved in the story, etc.) and store the event and its importance as the character's memory. In addition, the plan generation unit 21 may store the time period in the story when the event related to the character takes place (for example, the elapsed time in the story, the date and time in the story, a timestamp, etc.) associated with that event.

[0085] The planning unit 21 estimates the importance of events based on some or all of the information received from the transmission unit 12 (for example, story setting information, character setting information, chapter information, etc.). For example, the planning unit 21 may estimate the importance of events based on the relationship between the characters' conflicts and the events, with higher importance being associated with higher relationships.

[0086] The plan generation unit 21 may use some or all of the plans and / or events stored as the memories of the characters when generating a plan. When the plan generation unit 21 uses some of the events, it generates a plan using a predetermined number of events stored in the memory unit. The predetermined number is a numerical value set in advance by the person who sets the plan, etc.

[0087] Furthermore, the plan generation unit 21 may use some or all of the characters' memories as action information for a predetermined period. When the plan generation unit 21 uses some of the memories, it uses a predetermined number of events stored in the memory unit as action information for a predetermined period.

[0088] The plan generation unit 21 may generate a plan based on events stored as the memories of the characters. For example, the plan generation unit 21 may generate a plan by extracting some or all of the events stored as the memories of the characters based on at least one of the following: relevance to previous content (story, plan, detailed actions), importance of the events, or recency of the events. The plan generation unit 21 may similarly extract some or all of the memories of the characters to be used as action information for a predetermined period.

[0089] For example, if a predetermined number is set in advance, the plan generation unit 21 calculates a score for each event based on at least one of the following: its relevance to previous content (story, plan, detailed actions), the importance of the event, and the recency of the event stored as a character's memory when generating the plan. The predetermined number of events are then used to generate the plan in order of their scores.

[0090] The planning generation unit 21 calculates high scores for events that are highly relevant to previous events, events of high importance, and new events. The planning generation unit 21 may also use the sum of these scores as a single score.

[0091] In addition, the total importance value of events available to the planning generation unit 21 may be set in advance, and the planning generation unit 21 may extract events in order of importance so as not to exceed that total value.

[0092] Furthermore, in order to generate plans and detailed actions, the memory unit may store information about initial values, information about the planning system, information about the action system (completed plans), information about the conversation system, and other information. Information about initial values ​​includes past, present, innate personality (provided_statements), usual lifestyle (lifestyle), and information about any additional information (additional_information).

[0093] Information related to the planning system includes an outline of the entire simulation period used for plan generation (daily_req), the plan itself (day_plan (e.g., daily plan)), and information including the detailed plan and modified detailed plan (minutes_plan (e.g., minute-by-minute plan)).

[0094] Behavioral information includes inferences about the outcome of the behavior (result) and inferences about the specific reasons / causes of the outcome (reason). Conversational information includes a summary of the conversation (dialogue). Other information includes information such as observations of object situations and the behavior of others (event) and the results of introspection (reflect).

[0095] The plan generation unit 21, the detailed action generation unit 31, and the story generation unit 41 may generate plans, detailed plans, etc., based on the information about initial values, planning systems, action systems, conversation systems, and other information stored in the memory unit.

[0096] <Receiving and transmitting plans for generating detailed actions> The plan generation unit 21 transmits the generated plan to the generation device 1. In step S604, the generation unit 11 receives the plans of the characters.

[0097] In step S605, the transmission unit 12 transmits the plan received by the generation unit 11 from the plan generation unit 21 to the detailed action generation device 3. The transmission unit 12 may also transmit some or all of the information received in step S601 to the detailed action generation device 3. The generation unit 11 may generate information to be transmitted to the detailed action generation device 3 based on the character plans and / or story information.

[0098] In step S606, the detailed action generation unit 31 generates detailed actions based on the character's plan. Based on the plan received from the transmission unit 12, the detailed action generation unit 31 generates detailed actions relating to actions that are more detailed than the plan, for the character who is the subject of that plan. The detailed actions may include the time when the character performed the action. The detailed action generation unit 31 may further generate detailed actions based on the story information received from the transmission unit 12.

[0099] Specifically, the detailed action generation unit 31 generates detailed actions for each character at each time interval, based on the plan and the character's parameters, which are more detailed than the plan. By generating detailed actions based on the character's parameters, the characters in the story can be made to behave in a way that is closer to that of a real person, acting according to their mood and other factors of the day.

[0100] The detailed action generation unit 31 simulates the characters' actions every minute (for example, detailed events) based on the received plan. Furthermore, the detailed action generation unit 31 simulates the characters' parameters every minute. For example, the first detailed action simulated (generated) by the detailed action generation unit 31 is set as the number of days that have passed on the first day of the story.

[0101] For example, even if the accepted plan specifies waking up at 7:00, if the sleepiness parameter is high (for example, above a pre-set threshold), the detailed action generation unit 31 generates a detailed action for that character where they are asleep at 7:00. The detailed action generation unit 31 simulates the character's actions minute by minute and generates a detailed action where the character wakes up at a time when the sleepiness parameter has decreased.

[0102] The memory unit may store parameter thresholds for simulating the characters' actions as narrative information. For example, the memory unit may store thresholds for parameters such as the character eating and the character going to sleep, and the transmission unit 12 may transmit these thresholds to the detailed action generation device 3. In this way, the detailed action generation unit 31 generates the characters' detailed actions based on predetermined thresholds.

[0103] The detailed action generation unit 31 simulates (changes) parameters based on the generated detailed actions and / or the passage of time. Since humans get hungry as time passes, for example, the detailed action generation unit 31 may change the parameters at regular intervals regardless of the detailed actions of the characters.

[0104] For humans, sleep reduces drowsiness and restores physical strength, etc., so for example, the detailed behavior generation unit 31 changes emotional parameters and / or desire parameters based on the time of sleep. If a character is sleeping as a detailed behavior, the detailed behavior generation unit 31 changes the character's desire parameters (for example, parameters such as physical strength) (for example, recovery) based on the elapsed time of sleep.

[0105] In addition, since humans get hungry as time passes, for example, the detailed behavior generation unit 31 changes the emotion and / or desire parameters based on the passage of time. The detailed behavior generation unit 31 changes (increases or decreases) the hunger (fullness) parameter based on the passage of time.

[0106] The interval between actions and / or parameters simulated by the detailed action generation unit 31 only needs to be shorter than the interval of the plan generated by the plan generation unit 21, and simulations can be performed every few minutes, for example.

[0107] The memory unit may store the conditions for changing the character's parameters as narrative information. For example, the memory unit may store the time interval at which the character's parameters change, or the character's actions and the values ​​of the changing parameters, and the transmission unit 12 may transmit the threshold values ​​to the detailed action generation device 3. In this way, the detailed action generation unit 31 can change the parameters.

[0108] The detailed action generation unit 31 may further generate detailed actions of the characters in the chapter based on chapter information such as the characters appearing in the chapter, themes, and events that occur in the chapter.

[0109] The detailed action generation unit 31 may generate information for each character, including the detailed actions of the characters in the plan for a predetermined period generated by the plan generation unit 21, and the parameters of the characters at the time of those detailed actions, and may further transmit this information to the generation device 1. The detailed action generation unit 31 may further generate information including the location of the characters at the time of their detailed actions.

[0110] The detailed action generation unit 31 generates detailed actions, including interactions (conversations, etc.) between multiple characters, for each time period, based on the plans or detailed actions of multiple characters. The detailed action generation unit 31 determines the sector information and / or arena information included in the detailed actions for each character and generates detailed actions, including conversations between characters in the same location.

[0111] Even if characters are in the same location, conversations usually do not occur if they are not on good terms. Therefore, the detailed action generation unit 31 may further generate conversations between characters based on character relationship information. For example, the memory unit may store the conditions under which conversations occur between characters (e.g., a threshold for intimacy) as narrative information, and the transmission unit 12 may transmit that threshold to the detailed action generation device 3.

[0112] In this way, the detailed action generation unit 31 can determine whether a conversation will occur based on the location of each character, the conditions under which the conversation will occur, and the character relationship information, and then generate the conversation.

[0113] Furthermore, the detailed action generation unit 31 may generate detailed actions in which multiple characters act together, based on the plans or detailed actions of multiple characters.

[0114] The detailed action generation unit 31 may store the generated detailed actions of the characters as the memories of those characters. Specifically, the detailed action generation unit 31 stores the generated detailed actions in the memory unit, linking them to the character ID. Furthermore, the detailed action generation unit 31 may also store when in the story those detailed actions take place (for example, the number of days elapsed in the story, the date in the story, a timestamp, etc.), linking them to the story.

[0115] Furthermore, the detailed action generation unit 31 may store each event included in the generated detailed action (for example, actions at each time point) in the memory unit as a character's memory. When storing an event, the detailed action generation unit 31 may estimate the importance of that event and store it associated with that importance. In addition, the detailed action generation unit 31 may store the time period in the story when that detailed action event takes place (for example, elapsed time in the story, date and time in the story, timestamp, etc.) associated with that event.

[0116] Furthermore, the detailed action generation unit 31 may estimate the importance of events related to a character (for example, events included in the detailed actions of other characters with whom the character has a high level of intimacy, events included in the detailed actions of other characters involved in the story, etc.) and store those events and their importance as the character's memory. In addition, the detailed action generation unit 31 may store when an event related to a character takes place in the story (for example, the elapsed time in the story, the date and time in the story, a timestamp, etc.) in association with that event.

[0117] The detailed action generation unit 31 estimates the importance of an event based on some or all of the information received from the transmission unit 12 (for example, story setting information, character setting information, chapter information, etc.). For example, the detailed action generation unit 31 may estimate the importance of an event based on the relationship between the character's conflict and the event, with the higher the relationship, the higher the importance.

[0118] The detailed action generation unit 31 may use some or all of the detailed actions and / or events stored as the character's memory when generating detailed actions. When the detailed action generation unit 31 uses some of the events, it generates detailed actions using a predetermined number of events stored in the memory unit. The predetermined number is a numerical value set in advance by the person who created the system, etc.

[0119] The detailed action generation unit 31 may generate detailed actions based on events stored as the characters' memories. For example, the detailed action generation unit 31 may extract some or all of the events stored as the characters' memories and generate detailed actions based on at least one of the following: relevance to previous content (story, plan, detailed actions), importance of the event, or recency of the event.

[0120] For example, if a predetermined number is set in advance, the detailed action generation unit 31 calculates a score for each event based on at least one of the following: its relevance to previous content (story, plan, detailed action), the importance of the event, and the recency of the event stored as a character's memory when generating the detailed action. The predetermined number of events are then used to generate the detailed action in order of their scores.

[0121] The detailed action generation unit 31 calculates high scores for events that are highly relevant to previous events, events of high importance, and new events. The detailed action generation unit 31 may also use the sum of these scores as a single score.

[0122] In addition, the total importance value of events available to the detailed action generation unit 31 may be set in advance, and the detailed action generation unit 31 may extract events in order of importance so as not to exceed that total value.

[0123] In addition, the detailed action generation unit 31 may select information from the generated detailed actions to be added as the character's memory for a predetermined period. The detailed action generation unit 31 may also store the information to be added as memory in the memory unit, linked to the character ID.

[0124] The detailed action generation unit 31 estimates the importance of events related to the characters. The memory unit stores the upper limit of information that each character can add as memory (the upper limit of the total importance), and the detailed action generation unit 31 may sum up the estimated importance and select information to add so that the total importance does not exceed the upper limit.

[0125] If the total importance exceeds the upper limit, the detailed action generation unit 31 deletes items in order of decreasing importance so that the total importance does not exceed the upper limit. In this way, only events with high importance can be added as memories of the characters.

[0126] By using the detailed actions of a character during a predetermined period as the character's memory, the plan generation unit 21 can use the information added as the character's memory as information about actions during that predetermined period when generating a plan for after that period.

[0127] <Receiving and transmitting detailed actions for generating the story> The detailed action generation unit 31 transmits the generated detailed actions to the generation device 1. In step S607, the generation unit 11 receives the detailed actions of the characters.

[0128] In step S608, the transmission unit 12 transmits the detailed actions received by the generation unit 11 from the detailed action generation unit 31 to the story generation device 4. The transmission unit 12 may also transmit some or all of the information received in step S601 to the story generation device 4. The generation unit 11 may generate information to be transmitted to the story generation device 4 based on the detailed actions and / or story information.

[0129] In step S609, the story generation unit 41 generates a story based on the detailed actions received from the transmission unit 12. The story generation unit 41 may further generate a story based on the information received from the transmission unit 12. The story generation unit 41 may also generate a story based on some or all of the detailed actions and story information.

[0130] For example, the narrative generation unit 41 generates a single scene in the story by grouping the detailed actions together based on the timing and relationship between the detailed actions.

[0131] The story generation unit 41 may group actions in the same situation at consecutive times and generate them as a single scene. The story generation unit 41 determines the location of actions at consecutive times and generates actions at the same location as a single scene. For example, if the location of a character from 9 am to 3 pm is a high school, the story generation unit 41 will generate a story as a scene at the high school.

[0132] The narrative generation unit 41 may generate a first-person narrative from the perspective of each character. By having the plan generation unit 21 and the detailed action generation unit 31 generate the plans and detailed actions of each character at the same time, the narrative generation unit 41 can generate a narrative from the perspective of each character at the same time.

[0133] Furthermore, the narrative generation unit 41 may combine the actions of multiple characters into a single scene. The narrative generation unit 41 may determine the location of each character at a given time, and if multiple characters are in the same location, it may combine their actions into a single scene.

[0134] The story generation unit 41 may further generate a story related to a chapter based on its role and purpose (professional writer, commercial-level scenario, etc.), output format and writing rules (style and expression), examples of writing styles, work information, characters appearing in the chapter to be generated, themes, events occurring in that chapter, and other chapter information.

[0135] In addition, the story generation unit 41 generates stories (story scenes) that consist of "introduction," "development," "climax," and "conclusion." The story generation unit 41 generates stories (story scenes) such that the events in the plan generated by the plan generation unit 21 and the detailed actions generated by the detailed action generation unit 31 are arranged in chronological order in the order of "introduction," "development," "climax," and "conclusion."

[0136] "Beginning" refers to an event that marks the start of a story (scene), "development" refers to an event that leads the story (scene) towards its climax, "turning point" refers to an event that is the climax of the story (scene), and "conclusion" refers to an event that marks the end of the story (scene). The story generation unit 41 extracts events that can become the "beginning," "development," "turning point," and "conclusion" of a story in the plan and / or detailed actions.

[0137] Specifically, the story generation unit 41 extracts events in the plan that depict the character's conflict, based on conflict information regarding the character's conflict and the plan generated by the plan generation unit 21. The story generation unit 41 then extracts the events that depict the character's conflict as events that correspond to the "turning point" of the story.

[0138] Furthermore, the story generation unit 41 extracts events that occur after (in chronological order) the event corresponding to the "turning point" in the plan as the conclusion ("end" of the story). The story generation unit 41 may also extract events that are related to the theme of the story as events corresponding to the "end" based on some or all of the information received by the generation unit 11.

[0139] Furthermore, the story generation unit 41 extracts events that occur before (in chronological order) the events corresponding to the "turning point" in the plan as the beginning of the story ("beginning"). The story generation unit 41 may also extract events that are related to the theme of the story as events corresponding to the "beginning" based on some or all of the information received by the generation unit 11.

[0140] The story generation unit 41 then extracts the events that occur between the event corresponding to the "beginning" and the event corresponding to the "turning point" (in chronological order) during the planning process as the "development." The story generation unit 41 then generates a story (story scenes) by arranging the events extracted as "beginning," "development," "turning point," and "conclusion" in the order of "beginning," "development," "turning point," and "conclusion."

[0141] The narrative generation unit 41 generates a narrative by extracting detailed actions corresponding to the events in the extracted plan. The memory unit stores the elapsed time and / or date and time in the narrative in association with the events of the plan and detailed actions, so that the narrative generation unit 41 can extract detailed actions from the events in the extracted plan. For example, a consistent narrative can be generated by using the scene number, the time of the scene, the location, the characters, the surrounding environment / set, the detailed actions, and the content of the next scene.

[0142] The generation unit 11 receives the user's choice of language for the story to be generated via the user terminal 5, and the transmission unit 12 may transmit the story information and the language choice to the plan generation unit 21, the detailed action generation unit 31, and the story generation unit 41. In this way, the plan generation unit 21, the detailed action generation unit 31, and the story generation unit 41 can each generate a plan, detailed action, and story written in any language (for example, Japanese, English, etc.).

[0143] The model determination unit 13 determines the language of the received story information and selects a learning model corresponding to the determined language. The transmission unit 12 then transmits the received story information, plan, detailed actions, etc., to the learning model. A learning model trained in any language is influenced by the culture of that language, and therefore can generate stories that reflect that culture.

[0144] For example, the plan generation device 2 may include multiple plan generation units 21 trained in any language (for example, learning models trained with data in Japanese, English, etc.). Similarly, the detailed action generation device 3 may include multiple detailed action generation units 31 trained in any language (for example, learning models trained with data in Japanese, English, etc.). Furthermore, the story generation device 4 may include multiple story generation units 41 trained in any language (for example, learning models trained with data in Japanese, English, etc.).

[0145] The memory unit stores multiple learning models, associating each learning model with the corresponding language, allowing the transmission unit 12 to transmit information to the corresponding learning model.

[0146] The generation unit 11 receives request information regarding requests for the story (for example, requests related to the development of the story, such as events that will occur) via the user terminal 6. The transmission unit 12 transmits the request information received by the generation unit 11 to either the plan generation unit 21, the detailed action generation unit 31, or the story generation unit 41, thereby allowing the user to influence the generation of the story.

[0147] The plan generation unit 21, the detailed action generation unit 31, and the story generation unit 41 may each generate a plan, detailed actions, and a story based on the received request information.

[0148] The determination unit 14 determines the content of the received request using a trained model, and the transmission unit 12 may transmit the request information to the plan generation unit 21 or the like based on the result of that determination.

[0149] The determination unit 14 may determine whether the content of the request is discriminatory, whether the number of characters in the request is less than or equal to a predetermined number of characters, etc. If the transmission unit 12 determines that the content of the request is discriminatory, it will not transmit the information. Also, if the number of characters in the request is not less than or equal to the predetermined number of characters, the transmission unit 12 may choose not to transmit the information, or it may summarize the content and transmit it.

[0150] In addition, the judgment unit 14 may also determine whether the received content is compatible with the worldview of the story. For example, if the genre of the story is a coming-of-age drama, but the generation unit 11 receives a request that turns the story into a murder mystery, and the plan generation unit 21 etc. reflects that request, the worldview of the story will be ruined and messed up. The judgment unit 14 determines whether the request will not ruin the worldview of the story based on the received request information and the story information.

[0151] <Embodiment 2> In order to improve the accuracy of character simulations and the quality of plans, detailed actions, and the story, the story may be generated using more detailed information for generating the story. In Embodiment 2, the same data as in Embodiment 1 may be used, and the same processing may be performed.

[0152] Figures 7-13 show an example of the data configuration stored in the memory unit in Embodiment 2. The arrangement of each data is just an example, and some or all of the data stored in the memory unit of the generation device 1 may be stored in one or more devices configured to communicate with the plan generation device 2, detailed action generation device 3, story generation device 4, setter terminal 5, user terminal 6, and generation device 1.

[0153] Project information is information used to generate a story, and includes the story's title, genre, logline, status, concept, world setting (time period / background), writing style, style, keywords related to the story, the creation date and update date of the project information, etc., and is managed by a project ID, as shown in Figure 7(a).

[0154] Simulation information is information for running a simulation and includes the ID of the user running the simulation, the project ID, the title of the simulation, a description of the simulation title, the main events that occur in the simulation, the name of the simulation world in which the simulation takes place, the plot structure type (which part of the story the simulation corresponds to), the ID of the character being simulated (the ID of the character appearing), the situation of the character appearing in the simulation (the actions and emotions of the character appearing), the theme information ID related to the theme of the simulation, the roles of the characters (the roles of the characters appearing in the simulation), the change in basic needs (the value of the basic needs of the character in this simulation), a list of the emotions of the characters appearing, the change in intimacy (the value of the intimacy of the character towards other characters in this simulation), the simulation ID of this simulation, the start time, start date, and end date of this simulation in the story, the creation date and time when the simulation information was created, the update date and time, the basic needs setting value ID, etc., and is managed by a document ID as shown in Figure 8(b).

[0155] The basic needs setting information is information about the basic needs of the characters in the simulation, and includes needs, initial values, states when dissatisfied, recommended actions when dissatisfied, probability of decrease at each time step, and explanations about the basic needs, and is managed by a basic needs setting ID as shown in Figure 9(c).

[0156] For example, in the simulation information including the basic needs setting ID in Figure 9(c), the needs values ​​of the characters in the simulation increase or decrease by the values ​​described in the explanation, based on the probability of decrease at each time step in Figure 9(c).

[0157] Character information is information about characters appearing in a story, and includes the project ID of the story in which the character appears, the character's name, age, gender, current situation, habits / quirks, lifestyle (daily routine), appearance / external characteristics, speech patterns, social relationships (closeness to other characters), initial values ​​for introspection triggers (values ​​of information the character can remember during the simulation), the character's motivations (goals / objectives) (in the story), values ​​(preferences / beliefs), innate personality, learned (acquired) personality, personality (negative aspects), a list of memory IDs (past simulation IDs), the creation date and time the character information was created, the update date and time when it was updated, and the last simulation date and time when the simulation including that character was run, etc., and is managed by the character ID as shown in Figure 10(d).

[0158] Theme information is information related to the theme of a simulation and includes the project ID of the story associated with the simulation, the simulation ID, the title and description of the theme, the events that occur, the conflicts of the characters, the opportunities for the characters, the changes in the characters' emotions, the simulation period (days, weeks, months, years, etc.), the simulation units (seconds, minutes (1, 5, 10, 15, 30, 60 minutes, etc.), hours, etc.), and the event ID for specific events, and is managed in the theme information as shown in Figure 11(e).

[0159] Event information includes information about an event, such as the event name, event description, location, participants, and time of the event, and is managed as shown in Figure 12(f). Theme information includes an event ID, which allows for the management of events that occur under that theme.

[0160] Environmental information refers to information about the environment of a story (places and objects that appear in the story), and includes the project ID of the story using that environment, the simulation ID, the creation date and time when the environmental information was created, the update date and time when it was updated, the name of the environment (world, sector, arena, etc.), agent placement (placement of objects, placement of characters, etc.), location information (positional relationship of places that appear in the story), the world (environment) structure, etc., and is managed by the environment ID as shown in Figure 12(g).

[0161] Scenario information is information about the story's scenario, and includes the project ID of the story using that scenario, the simulation ID, the character IDs appearing in that scenario, the episode number in the story of that scenario, the theme number, the daily schedule index number, the language used in the scenario, the title of the scenario, the text of the scenario, the emotional information of the characters, the overall structure, etc., and is managed by the scenario ID as shown in Figure 13(h).

[0162] Request information is information about requests that a user enters to converse with a character in a story, and includes a simulation specification to specify the character the user wants to converse with (even for the same character, their personality etc. may change in each simulation (as the story progresses), so this is information to specify a character in a specific state), input text, conversation history between the user and the character, username, location of the conversation, language of the conversation, target time of the conversation, etc., and is managed by a character ID as shown in Figure 13(i).

[0163] Conversation memory information refers to conversations between the user and characters in the story, and includes information such as the content of the conversation, the type of memory, the importance of the memory, the time the memory was created, a summary of the conversation, the level of intimacy between the user and the character, and the character's feelings towards the user, as shown in Figure 13(j). Conversation memory information may also be managed by conversation IDs.

[0164] Figures 14 to 21 show an example of a story generation screen in Embodiment 2. The generation device 1 may include a display processing unit, and some of the functional configurations of the generation device 1 may be arranged in one or more devices configured to communicate with the plan generation device 2, detailed action generation device 3, story generation device 4, setter terminal 5, user terminal 6, and the generation device 1.

[0165] Figure 14 is a settings screen for receiving project information, and the display processing unit displays a screen like Figure 14 in order to receive project information. The generation unit 11 receives project information via a screen like Figure 14 on the user terminal 5.

[0166] Figures 15-18 show a settings screen for receiving character information about characters appearing in the story. The display processing unit displays screens like those shown in Figures 15-18 to receive the character information. The generation unit 11 receives character information and the like via screens like those shown in Figures 15-18 on the user terminal 5.

[0167] The display processing unit may display a list of characters currently appearing in the registered story, as shown in Figure 15(a), and also display a summary of each character. By performing such display processing, the creator can check the registered characters and consider adding new characters. For example, the creator can add a character by clicking the "New Character" button displayed in Figure 15(a).

[0168] The display processing unit displays a screen for receiving basic information about the characters appearing in the story, as shown in Figure 15(b). The display processing unit displays a screen for receiving the personalities of the characters appearing in the story, as shown in Figure 16(c). The display processing unit displays a screen for receiving the characteristics of the characters appearing in the story, as shown in Figure 16(d).

[0169] The display processing unit displays a screen for receiving the lifestyle of a character appearing in the story, as shown in Figure 17(e). By accepting a start time and end time as part of the lifestyle, it is possible to set the character's lifestyle (daily routine) in increments of a few minutes.

[0170] The display processing unit displays a screen for receiving information about the relationships between characters in the story and other characters (participants), as shown in Figure 18(f). The display processing unit also displays a screen for receiving information that the user has given to the character's initial memory, as shown in Figure 18(g).

[0171] Figure 19 is a settings screen for receiving simulation information to run the simulation, and the display processing unit displays a screen like Figure 19 in order to receive the simulation information. The generation unit 11 receives the simulation information via a screen like Figure 19 on the user terminal 5.

[0172] Figure 20 is a screen used to check the status of characters appearing in the story during the simulation. In Figure 20, the display processing unit displays the status of the characters participating in the simulation at the time "23:11" during the simulation.

[0173] The display processing unit displays the character's emotions and location at the selected time. Furthermore, by scrolling the bar W20 in Figure 20, which is processed by the display processing unit, left or right, the display processing unit may also display the character's status at a time that the setter and / or the user wish to check.

[0174] Figure 21 is an example of a screen displaying the story's environment (setting). When the simulation starts, the display processing unit displays the story's setting as shown in Figure 21. The story's setting has a hierarchical structure as shown in Figure 21.

[0175] For example, as shown in Figure 21, there is a world (Hana's hometown), and within the world there are places (Hana's house, high school, local music event venue, etc.), and within the places within the world there are further places (Hana's bedroom, living room, etc.), and within those places there are objects or further places (audience seats, aisles, standing areas, etc.).

[0176] Figures 22 to 30 are examples of flowcharts for the story generation process in Embodiment 2. Figure 22 is an overview of the flowchart for the story generation process. In step S2201, the generation unit 11 receives data input via the user terminal 5 or the like.

[0177] In step S2202, the detailed action generation unit 31 executes a simulation. In step S2203, the story generation unit 41 generates a story (scenario). In step S2204, the person who set the story (scenario) reviews it, and the generation unit 11 accepts any modifications to the scenario.

[0178] Figure 23 is an example of a flowchart of the simulation process in Embodiment 2. In step S1, the detailed action generation unit 31 reads project information (Figure 7(a)), simulation information (Figure 8(b)), and character information (Figure 10(d)) from the storage unit.

[0179] In step S2, the detailed action generation unit 31 determines whether environmental information (Figure 12(g)) exists (is stored in the memory unit). If environmental information exists (step S2: Y (Yes)), in step S3, the detailed action generation unit 31 reads the environmental information.

[0180] If no environmental information exists (step S2: N (No)), in step S4, the detailed action generation unit 31 generates environmental information. In step S5, the detailed action generation unit 31 stores the generated environmental information in the storage unit.

[0181] In step S6, the detailed action generation unit 31 initializes the agent. Specifically, the detailed action generation unit 31 initializes the positions of objects and characters to be placed in locations within the world.

[0182] In step S7, the detailed action generation unit 31 determines whether theme information (Figure 11(e)) exists (is stored in the memory unit). If theme information exists (step S7: Y), in step S8, the detailed action generation unit 31 reads the theme information. If theme information does not exist (step S8: N), in step S9, the detailed action generation unit 31 generates the theme information.

[0183] In steps S10 to S28, the detailed action generation unit 31 executes loop A. In loop A, the detailed action generation unit 31 performs a simulation (loop) for a specified number of days. The detailed action generation unit 31 performs a loop for a minimum of 1 day and a maximum of 90 days.

[0184] In step S11, the detailed action generation unit 31 sets the start and end dates and times of the simulation. The detailed action generation unit 31 sets the dates and times based on the start and end dates and times received from the setter, etc. In step S12, the detailed action generation unit 31 generates events.

[0185] In step S13, the detailed action generation unit 31 sets the simulation unit (1, 5, 10, 15, 30, 60 minutes, etc.). The detailed action generation unit 31 performs simulations at time intervals based on the simulation unit. The detailed action generation unit 31 performs simulations based on the simulation unit received from the setter, etc.

[0186] The simulation unit may be automatically set to a unit corresponding to the specified number of days. For example, the detailed action generation unit 31 performs a simulation in time units of the same numerical value as the specified number of days (1 minute if the specified number of days is 1 day, 30 minutes if it is 30 days, 60 minutes if it is 60 days).

[0187] In steps S14 to S27, the detailed action generation unit 31 executes loop B. In loop B, the detailed action generation unit 31 performs a simulation for one day. The detailed action generation unit 31 performs simulations on a daily basis.

[0188] In step S15, the detailed action generation unit 31 sets the current (start) time for the day's simulation.

[0189] In step S16, the detailed action generation unit 31 initializes the agent's state. Specifically, the detailed action generation unit 31 initializes the positions of objects and characters placed in locations within the world.

[0190] In step S17, the detailed action generation unit 31 updates the character's personality. By performing multiple simulations, the character's personality can be updated by learning from past simulations. In step S18, the detailed action generation unit 31 initializes the agent's state.

[0191] In step S19, the detailed action generation unit 31 sets the end time for the day's simulation. The end time in step S19 is the start time in step S15 plus the simulation unit.

[0192] In steps S20 to S26, the detailed action generation unit 31 executes loop C. In loop C, the detailed action generation unit 31 simulates the time between the current time set in step S15 and the end time set in step S19.

[0193] In steps S211 to S22N, the detailed action generation unit 31 updates agents 1 to N, respectively. In step S22, the detailed action generation unit 31 collects logs of the updated agents. By running simulations in parallel for each character, the simulation time can be shortened.

[0194] In step S23, the detailed action generation unit 31 processes the conversation between the characters. In step S24, the detailed action generation unit 31 saves the log.

[0195] In step S25, the simulation unit is added to the current time. In this way, the current (start) time in step S15 becomes the new time. In S29, the detailed action generation unit 31 stores the simulation results in the storage unit.

[0196] Figure 24 is an example of a flowchart of scenario (story) processing in Embodiment 2. In step S31, the story generation unit 41 reads project information (Figure 7(a)), simulation information (Figure 8(b)), character information (Figure 10(d)), and theme information (Figure 11(e)) from the storage unit.

[0197] In step S32, the story generation unit 41 initializes the agent. In step S33, it is determined whether there has been a change in the agent's initial state.

[0198] If there is a change in the initial state (step S33: Y), in step S34, the story generation unit 41 overwrites the initial state of the agent. If there is no change in the initial state (step S33: N), the process proceeds to step S35.

[0199] In step S35, the story generation unit 41 determines whether the simulation is the second or later. If it is the second or later simulation (step S35: Y), in step S36, since there is scenario information generated based on past simulations, the story generation unit 41 reads the previous scenario information (Figure 13(h)) from the memory unit. If it is not the second or later simulation (step S35: N), the process proceeds to step S37.

[0200] In step S37, the story generation unit 41 determines whether information for the next simulation exists. If information for the next simulation exists (step S37: Y), in step S38, the story generation unit 41 reads the information for the next simulation from the memory unit. If information for the next simulation does not exist (step S37: Y), the process proceeds to step S39.

[0201] In steps S39 to S67, the story generation unit 41 executes loop A. In loop A, the story generation unit 41 performs a simulation (loop) for a specified number of days. The story generation unit 41 performs a loop for a minimum of 1 day and a maximum of 90 days. In step S40, the story generation unit 41 sets the start and end dates and times of the simulation.

[0202] In steps S41 to S66, the story generation unit 41 executes loop B. In loop B, the story generation unit 41 performs a simulation of one day. The story generation unit 41 performs a simulation on a daily basis. In step S42, the end time of the daily simulation is set.

[0203] In steps S43 to S45, the story generation unit 41 executes the C loop. In the C loop, the story generation unit 41 performs a loop for each character (on a character-by-character basis). In step S44, the story generation unit 41 obtains the simulation results for each character.

[0204] In step S46, the story generation unit 41 generates the overall structure of the story (introduction, development, climax, and conclusion). In step S47, the story generation unit 41 sets the simulation unit (1, 5, 10, 15, 30, 60 minutes, etc.).

[0205] In step S48, the story generation unit 41 determines whether a prologue exists in the overall structure. If a prologue exists in the overall structure (step S48: Y), in step S49, the story generation unit 41 determines whether it is the first simulation. If a prologue does not exist in the overall structure (step S48: N), the process proceeds to step S52.

[0206] If it is the first simulation (step S49: Y), the story generation unit 41 generates a prologue. If it is not the first simulation (step S49: N), in step S51, the story generation unit 41 adds the prologue to the beginning of the story's structure.

[0207] In step S52, the story generation unit 41 determines whether an epilogue exists in the overall structure. If an epilogue exists in the overall structure (step S52: Y), in step S53, the story generation unit 41 adds an epilogue to the conclusion part of the story's introduction, development, turn, and conclusion. If an epilogue does not exist in the overall structure (step S52: N), the process proceeds to step S54.

[0208] In step S54, the narrative generation unit 41 determines whether the simulation unit is less than 60 minutes. If the simulation unit is less than 60 minutes (step S54: Y), the process proceeds to step S55. If the simulation unit is not less than 60 minutes (step S54: N), the process proceeds to step S63.

[0209] In steps S55 to S57, the story generation unit 41 executes the D loop. In the D loop, the story generation unit 41 performs a loop through the overall structure (introduction, development, climax, and conclusion). In step S56, the story generation unit 41 acquires scene information.

[0210] In steps S58 to S62, the story generation unit 41 executes the E: loop. In the E: loop, the story generation unit 41 performs a loop for each scene. In step S59, the story generation unit 41 determines whether the scene information is in the format of introduction, development, climax, and conclusion.

[0211] If the scene information is either introduction, development, climax, or conclusion (step S59: Y), in step S60, the story generation unit 41 generates a scenario from the scene information. In step S61, the story generation unit 41 adds the current episode number.

[0212] In step S63, the story generation unit 41 generates a scenario (story) from the overall structure (introduction, development, turn, and conclusion). In step S64, the story generation unit 41 adds the current episode number.

[0213] In step S65, the story generation unit 41 stores the scenario information in the memory unit. In step S68, the story generation unit 41 writes the scenario to external storage.

[0214] Figure 25 is an example of a flowchart for user-character conversation processing in Embodiment 2. Figure 25(a) is a flowchart for the process of generating the character's response, and Figure 25(b) is a flowchart for the process when a request (conversation) from the user is received.

[0215] The generation system may include a conversation generation device that generates conversations between the user and the character. In addition, conversations between the user and the character may be generated by any of the following: the plan generation device 2, the detailed action generation device 3, or the story generation device 4.

[0216] A conversation generation device is a server device that provides a pre-trained model, and the conversation generation unit of the conversation generation device is thought to be a pre-trained model. A conversation generation device is thought to be a server device that provides a natural language processing (NLP) model.

[0217] In step S71, the conversation generation unit reads the request information from the memory unit. In step S72, the conversation generation unit reads the simulation information and character information. In step S73, the conversation generation unit initializes the agent.

[0218] In step S74, the conversation generation unit determines whether memory information for the conversation exists. If memory information for the conversation exists (step S74: Y), in step 75, the conversation generation unit reads the memory information for the conversation from the memory unit.

[0219] In step S76, the conversation generation unit overrides the agent's (character's) state (emotional intimacy) based on the stored conversation information. In step S77, the conversation generation unit calculates the total length of past conversations between the user and a specific character based on the stored information.

[0220] If no conversation memory exists (step S74: N), the process proceeds to step 78. In step S78, the conversation generation unit refers to the conversation history between the user and a specific character. In step S79, the conversation generation unit generates a response from the specific character.

[0221] In step S80, the conversation generation unit adds the generated response to the conversation history (stores it in the memory unit as conversation history information). In step S81, the conversation generation unit calculates the time taken for the conversation (calculating based on 240 characters taking 1 minute).

[0222] In step S82, the conversation generation unit determines whether the conversation took longer than 15 minutes. If the conversation took longer than 15 minutes (step S82: Y), in step S83, the conversation generation unit generates a summary of the conversation. If the conversation took longer than 15 minutes (step S82: N), the process proceeds to step S86.

[0223] In step S84, the conversation generation unit increases or decreases the intimacy level between the user and a specific character based on the conversation content. In step S85, the conversation generation unit adds a summary of the conversation as the character's memory information (stores it in the memory unit). In step S86, the conversation generation unit provides the generated response to the user as a response.

[0224] In step S91, if a user requests a conversation with a specific character, a Push notification is sent. In step S92, the conversation generation unit aggregates the requests. In step S93, a lottery process is performed to add to the aggregated requests (e.g., randomly selecting requests, summarizing requests, etc.).

[0225] In step S94, the conversation generation unit determines whether the content of the request is harmful. If the request is not harmful (step S94: Y), in step S95, the conversation generation unit writes the request information to the simulation information.

[0226] Figure 26 is a flowchart of the process in step S9, "Generating Theme Information," in Figure 23. In step S101, the detailed action generation unit 31 decomposes events in chronological order. In step S102, the detailed action generation unit 31 estimates the starting point of the change in events.

[0227] In steps S103 to S110, the detailed action generation unit 31 performs loop processing for each event. In step S104, the detailed action generation unit 31 determines whether the event occurred in the past.

[0228] If the event is not in the past (step S104: Y), in step S105, the detailed action generation unit 31 generates a summary of the event. In step S106, the detailed action generation unit 31 estimates the reason / cause of the event. In step S107, the detailed action generation unit 31 estimates the duration of the event.

[0229] In step S108, the detailed action generation unit 31 determines whether the estimated duration of the event is greater than 0 days. If the estimate is greater than 0 days (step S108: Y), in step S109, the detailed action generation unit 31 saves the theme information.

[0230] In step S111, the detailed action generation unit 31 determines whether there is one or more theme information. If there is no one or more theme information (step S111: Y), in step S112, the detailed action generation unit 31 saves the theme information with an estimate of 1 day.

[0231] Figure 27(a) is a flowchart of the process in step S12, "Generating an event," in Figure 23. Figure 27(b) is a flowchart of the process in step S128, "Generating a special event," in Figure 27(a).

[0232] In step S121, the detailed action generation unit 31 determines whether an event (synopsis) exists. If no event exists (step S121:N), in step S122, the detailed action generation unit 31 generates an event (synopsis).

[0233] In step S123, the detailed action generation unit 31 determines whether a conflict exists. If no conflict exists (step S123:N), in step S124, the detailed action generation unit 31 generates a conflict.

[0234] In step S125, the detailed behavior generation unit 31 determines whether an emotional change exists. If no emotional change exists (step S125:N), in step S126, the detailed behavior generation unit 31 generates an emotional change.

[0235] In step S127, the detailed action generation unit 31 determines whether a special event exists. If no special event exists (step S127:N), in step S128, the detailed action generation unit 31 generates a special event.

[0236] In step S131, the detailed action generation unit 31 generates an event that constitutes a conflict (external factor). In step S132, the detailed action generation unit 31 generates a suggestion (foreshadowing). In step S133, the detailed action generation unit 31 randomly generates an event as a suggestion (foreshadowing) for the next chapter.

[0237] In step S134, the detailed action generation unit 31 randomly generates a plan as an indication (foreshadowing) for the next chapter. In step S135, the detailed action generation unit 31 generates events related to the past as an indication (foreshadowing) for the future.

[0238] In step S136, the detailed action generation unit 31 generates the events described in the request. In step S137, the detailed action generation unit 31 sorts the events in chronological order and deletes duplicate events.

[0239] Figure 28 is a flowchart of the "personality update" process in step S17 of Figure 23. In step S141, the detailed behavior generation unit 31 generates three questions. In step S142, the detailed behavior generation unit 31 adds new questions about knowledge, skills, and experience.

[0240] In steps S143 to S146, the detailed action generation unit 31 performs loop processing for each question. In step S144, the detailed action generation unit 31 reads the memory information from the previous simulation. In step S145, the detailed action generation unit 31 generates insights.

[0241] In step S147, the detailed behavior generation unit 31 summarizes the four insights. In step S148, the detailed behavior generation unit 31 adds the summary of insights to the personality (present).

[0242] Figure 29 is a flowchart of the "agent update" process in steps S211 to S21N in Figure 23. In step S151, the detailed action generation unit 31 updates the current open date and time.

[0243] In step S152, the detailed action generation unit 31 determines whether the current date and time is the simulation start date and time. If the current date and time is the simulation start date and time (step S152: Y), in step S153, the detailed action generation unit 31 generates a plot for each character.

[0244] If the current date and time is not the simulation start date and time (step S152: N), proceed to step S154. In step S154, the detailed action generation unit 31 determines whether the current time is the simulation start date and time.

[0245] If the current time is the simulation start date and time (step S154: Y), in step S155, the detailed action generation unit 31 generates the entire day's schedule. If the current time is not the simulation start date and time (step S154: N), the process proceeds to step S156. In step S156, the detailed action generation unit 31 determines whether it is the time when the schedule begins.

[0246] If it is the time when the scheduled event should begin (step S156: Y), in step S157, the detailed action generation unit 31 generates a detailed schedule. If it is not the time when the scheduled event should begin (step S156: N), the process proceeds to step S175. In step S158, the detailed action generation unit 31 determines whether there is a previous scheduled event that has been completed.

[0247] If there is a previously completed appointment (step S158: Y), in step S159, the detailed action generation unit 31 calculates the elapsed time based on the current time and the start time of the previous appointment. If there is no previously completed appointment (step S158: N), the process proceeds to step S163.

[0248] In step S160, the detailed action generation unit 31 determines whether the previously completed scheduled activity was something other than "conversation". If the previously completed scheduled activity was something other than "conversation" (step S160: Y), in step S161, the detailed action generation unit 31 analyzes the result. If the previously completed scheduled activity was not something other than "conversation" (step S160: N), the process proceeds to step S162.

[0249] In step S162, the detailed behavior generation unit 31 updates the inferred values ​​of emotions and desires. In step S163, the detailed behavior generation unit 31 determines whether the emotion and desire values ​​are below a threshold. If the emotion and desire values ​​are below a threshold (step S163: Y), in step S164, the detailed behavior generation unit 31 updates the status information.

[0250] If the emotion / desire value is not below the threshold (step S163: N), proceed to step S165. In step S165, the detailed action generation unit 31 infers the reason for the planned (action). In step S166, the detailed action generation unit 31 determines whether the desire value is below the threshold.

[0251] If the desire value is below the threshold (step S166: Y), in step S167, the detailed action generation unit 31 infers a change in the plan. In step S168, the detailed action generation unit 31 adds the changed plan to the agent's memory. If the desire value is not below the threshold (step S166: N), the process proceeds to step S169.

[0252] In step S169, the detailed action generation unit 31 performs world position inference. In step S170, the detailed action generation unit 31 performs sector position inference. In step S171, arena position inference. In step S172, affected objects inference.

[0253] In step S173, the detailed action generation unit 31 generates a detailed description of the landscape. In step S174, the detailed action generation unit 31 performs a location update process (update of environmental information). In step S175, the detailed action generation unit 31 adds surrounding events (from the environmental information) to the recognition.

[0254] In step S176, the detailed action generation unit 31 determines whether a new recognition exists. If a new recognition exists (step S175: Y), in step S177, the detailed action generation unit 31 sorts the recognitions by importance. In step S178, the detailed action generation unit 31 adds the top eight recognitions to memory.

[0255] In step S179, the detailed action generation unit 31 subtracts the sum of the importance of recognition from the introspection threshold (the value of information that the character can remember during the simulation). In step S180, the detailed action generation unit 31 determines whether the introspection threshold is 0 or less.

[0256] If the threshold for introspection is 0 or less (step S180: Y), in step S181, the detailed action generation unit 31 generates three questions for introspection. If the threshold for introspection is not 0 or less (step S180: N), the detailed action generation unit 31 terminates processing.

[0257] In step S182, the detailed action generation unit 31 executes a loop for each question. In step S183, the detailed action generation unit 31 reads memories related to the question (vector search). In step S184, the detailed action generation unit 31 generates insights from memories related to the question.

[0258] In step S185, the detailed action generation unit 31 adds the insight to memory (the memory unit as the character's memory). In step S187, the detailed action generation unit 31 sets the introspection threshold to MAX (predefined).

[0259] Figure 30 is a flowchart of the process related to step S23 "Conversation Processing" in Figure 23. In step S191, the detailed action generation unit 31 executes loop A. In loop A, the detailed action generation unit 31 performs a loop for each arena of environmental information.

[0260] In step S192, the detailed action generation unit 31 determines whether the interaction is digital (telephone call, message). If it is digital (step S192: Y), in step S193, the detailed action generation unit 31 adds everyone to the conversation participants. If it is not digital (step S192: N), the process proceeds to step S194.

[0261] In step S194, the detailed action generation unit 31 executes loop B. In loop B, the detailed action generation unit 31 performs a loop for each conversation participant. In step S195, the detailed action generation unit 31 determines whether the participant is sleeping or has already started talking.

[0262] If the person is asleep or has already started talking (step S195: Y), in step S196, the detailed action generation unit 31 excludes them from the conversation participants. If the person is asleep or has already stopped talking (step S195: N), the process proceeds to step S197.

[0263] In step S198, the detailed action generation unit 31 determines whether there is one or more conversation participants (agents). If there is one or more conversation participants (step S198: Y), in step S199, the detailed action generation unit 31 executes loop C. In loop C, the detailed action generation unit 31 performs a loop for each conversation participant. If there are no conversation participants (step S198: N), the process proceeds to step S239.

[0264] In step S200, the detailed action generation unit 31 cancels the current planned action and infers the need for the conversation participant to speak. In step S201, the detailed action generation unit 31 determines whether the conversation participant needs to speak.

[0265] If a conversation participant needs to speak (step S201:Y), in step S202, the detailed action generation unit 31 adds the conversation participant to the list of speakers. If a conversation participant does not need to speak (step S201:N), the process proceeds to step S203.

[0266] In step S204, the detailed action generation unit 31 determines whether there are two or more speakers. If there are two or more speakers (step S204: Y), in step S205, the detailed action generation unit 31 performs inference to determine which speaker will speak first. If there are no two or more speakers (step S204: N), the process proceeds to step S206.

[0267] In step S206, the detailed action generation unit 31 executes the D: loop. In the D: loop, the detailed action generation unit 31 performs a loop for each speaker. In step S207, the detailed action generation unit 31 removes the speaker from the conversation participants.

[0268] In step S208, the detailed action generation unit 31 executes loop E. In loop E, the detailed action generation unit 31 performs 15 loops of conversation turns. In step S209, the detailed action generation unit 31 determines whether there is a conversation target or a main point. If there is neither a conversation target nor a main point (step S209: Y), in step S210, the detailed action generation unit 31 determines the conversation target and the main point.

[0269] If both a conversation subject and a subject exist (step S209: N), proceed to step S211. In step S211, the detailed action generation unit 31 determines whether a conversation subject exists. If a conversation subject exists (step S211: Y), in step S212, the detailed action generation unit 31 skips the turn with a 10% probability. If no conversation subject exists (step S211: N), proceed to step S232.

[0270] In step S213, the detailed action generation unit 31 infers the content of the statement. In step S214, the detailed action generation unit 31 determines whether a statement exists. If a statement exists (step S214: Y), in step S215, the detailed action generation unit 31 reverses the content of the statement. If no statement exists (step S214: N), the process proceeds to step S232.

[0271] In step S216, the detailed action generation unit 31 generates a statement from the speaker. In step S217, the detailed action generation unit 31 adds the generated statement to the conversation history. In step S218, the detailed action generation unit 31 executes loop F. In loop F, the detailed action generation unit 31 executes a loop for each conversation participant.

[0272] In step S219, the detailed action generation unit 31 determines whether there is neither a conversation target nor a main point. If there is neither a conversation target nor a main point (step S219: Y), in step S220, the detailed action generation unit 31 determines the conversation target and the main point. If there is both a conversation target and a main point (step S219: N), the process proceeds to step S221.

[0273] In step S221, the detailed action generation unit 31 determines whether a conversation target exists and whether the speaker is the conversation target. If a conversation target exists and the speaker is the conversation target (step S221: Y), in step S222, the detailed action generation unit 31 skips the turn with a 10% probability. If step S221: N, the process proceeds to step S230.

[0274] In step S223, the detailed action generation unit 31 infers the content of the speaker's statement. In step S224, the detailed action generation unit 31 determines whether a statement exists. If a statement exists (step S224: Y), in step S225, the detailed action generation unit 31 reverses the statement. In step S226, the detailed action generation unit 31 generates a statement from the speaker.

[0275] In step S227, the detailed action generation unit 31 adds the generated statement to the conversation history. In step S224: If N, in step S229, the detailed action generation unit 31 removes the speaker from the conversation participants.

[0276] In step S230, the detailed action generation unit 31 determines whether there are conversation participants. If there are conversation participants (step S230: Y), the process proceeds to step S231; if there are no conversation participants (step S230: N), the process proceeds to step S232. In step S232, the detailed action generation unit 31 calculates the time taken for the conversation.

[0277] In step S233, the detailed action generation unit 31 generates a summary of the conversation. In step S234, the detailed action generation unit 31 executes a loop for the speaker and all conversation participants. In step S235, the detailed action generation unit 31 adds the conversation content to the memory of the speaker and all conversation participants.

[0278] In step S236, the detailed behavior generation unit 31 infers emotions. In step S237, the detailed behavior generation unit 31 updates the conversation participants' schedules according to the time taken for the conversation. In step S238, the detailed behavior generation unit 31 adds the updated schedules to the conversation participants' memories.

[0279] Figure 31 is a flowchart of the process in step S46, "Generating the overall structure (introduction, development, climax, and conclusion)," in Figure 24. In step S241, the story generation unit 41 breaks down the simulation results into scenes and summarizes them.

[0280] In step 242, the story generation unit 41 determines whether there is one or more scenes. If there is one or more scenes (step S242: Y), in step S243, the story generation unit 41 generates an embedded synopsis (of the target characters).

[0281] In step S244, the story generation unit 41 executes loop A. In loop A, the story generation unit 41 loops through each scene. In step S245, the story generation unit 41 generates embedded summaries of the scenes.

[0282] In step S247, the story generation unit 41 calculates the cosine similarity between the embedding of the synopsis and the scene summary. In step S248, the story generation unit 41 sorts the scenes (scenes) by similarity. In step S249, the story generation unit 41 selects only one scene that is most relevant to the event described in the plot.

[0283] In step S250, the story generation unit 41 designates the first scene as the "conclusion". In step S251, the story generation unit 41 determines whether there are scenes prior to the "conclusion". If there are scenes prior to the "conclusion" (step S251: Y), in step S252, the story generation unit 41 selects one or more scenes in which the events described in the plot are depicted first.

[0284] In step S253, the story generation unit 41 sorts the selected scenes. If there are no scenes prior to "Conclusion" (step S251: N), the process proceeds to step S254. In step S254, the story generation unit 41 designates the first scene as "Beginning". In step S255, the story generation unit 41 includes the scenes between "Beginning" and "Turning Point" in the overall structure.

[0285] In step S256, the narrative generation unit 41 determines whether an event for suggestion has been generated. If an event for suggestion has been generated (step S256: Y), in step S257, the narrative generation unit 41 searches the plans for a plan that matches the event for suggestion. If an event for suggestion has not been generated (step S256: N), the process proceeds to step S258.

[0286] In step S258, the narrative generation unit 41 determines whether an initial simulation and a suggestive event exist. If an initial simulation and a suggestive event exist (step S258: Y), in step S259, the narrative generation unit 41 selects only one scene related to the suggestive event.

[0287] If there is no initial simulation and no suggestive events (step S258: N), the process proceeds to step S260. In step S260, the narrative generation unit 41 determines whether there is the end of the theme information and whether there is a next simulation.

[0288] If there is an end to the theme information and a next simulation (step S260: Y), in step S261, the story generation unit 41 selects only one scene related to an event in the next simulation. If there is no end to the theme information and no next simulation (step S260: N), the process ends.

[0289] Figure 32 is a flowchart of the process in step S60, "Generating a scenario from scene information," in Figure 24. In step S271, the story generation unit 41 acquires the necessary scene information.

[0290] In step S272, the story generation unit 41 executes loop A. In loop A, the story generation unit 41 performs the loop up to three times. In step S273, the story generation unit 41 determines whether a review exists. If a review exists (step S273: Y), in step S274, the story generation unit 41 generates a scenario from the review and scene information.

[0291] If no criticism exists (step S273:N), in step S276, the story generation unit 41 generates a scenario from the scene information. In step S275, the story generation unit 41 executes loop B. In loop B, the story generation unit 41 performs the loop three times. In each of steps S277 to S279, the story generation unit 41 compares the existing scenario (A) with the generated scenario (B). The existing scenario is thought to be a scenario written by a human writer as training data.

[0292] In step S281, the story generation unit 41 determines whether the generated scenario (B) has one vote or less. If the generated scenario (B) has one vote or less (step S281: Y), in step S282, the story generation unit 41 generates a summary of the evaluation. If the generated scenario (B) has more than one vote or less (step S281: N), the process ends.

[0293] In step S283, the narrative generation unit 41 generates the initial critique. In step S284, the narrative generation unit 41 executes loop C. In loop C, the narrative generation unit 41 performs the loop three times. In step S285, the narrative generation unit 41 shuffles the order of statements.

[0294] In step S286, the story generation unit 41 executes loop D. In loop D, the story generation unit 41 loops for each speaker. In step S287, the story generation unit 41 generates criticism. In step S290, the E loop is executed. In loop E, the story generation unit 41 loops for the number of times a speaker has made a statement.

[0295] In step S291, the narrative generation unit 41 generates a critique that adds the content of the statements to the initial critique. In step S292, the narrative generation unit 41 determines whether there is a contradiction with the previous critique. If there is a contradiction (step S292: Y), in step S293, the narrative generation unit 41 executes loop F. In loop F, the narrative generation unit 41 performs the loop up to 5 times.

[0296] If there is no contradiction (step S292: N), proceed to step S297. In step S294, the story generation unit 41 corrects the contradiction. In step S295, the story generation unit 41 determines whether the contradiction has been resolved. If the contradiction has been resolved (step S295: Y), proceed to step S297.

[0297] In step S298, the narrative generation unit 41 generates the final critique. In step S299, the narrative generation unit 41 determines whether a critique from the previous loop exists. If a critique from the previous loop exists (step S299: Y), in step S300, the narrative generation unit 41 integrates it with the previous critique.

[0298] If there is no review of the previous loop (step S299: N), proceed to step S301. In step S302, the story generation unit 41 summarizes the entire text. In step S303, the story generation unit 41 generates a title. In step S304, the story generation unit 41 saves the scenario information (stores it in the memory unit).

[0299] As described above, the configuration of the present invention provides a new technology for generating stories in which characters behave in a human-like manner.

[0300] 0 Generation System 1 Generation Device 11 Generation Unit 12 Transmission Unit 13 Model Determination Unit 14 Judgment Unit 2 Plan Generation Device 21 Plan Generation Unit 3 Detailed Action Generation Device 31 Detailed Action Generation Unit 4 Story Generation Device 41 Story Generation Unit 5 Setter Terminal 6 User Terminal NW Network

Claims

1. A generation system for generating a story, comprising: a plan generation unit; a detailed action generation unit; and a story generation unit, wherein the plan generation unit generates a plan for a character after a predetermined period based on story information including an outline of the story and predetermined period action information relating to the character's actions during a predetermined period; the detailed action generation unit generates detailed actions relating to the character's actions at each time point, which are more detailed than the plan, based on the plan and emotion parameters relating to the character's emotions; and the story generation unit generates the story based on the detailed actions.

2. The generation system according to claim 1, wherein the story information includes information about locations appearing in the story, the plan generation unit generates a plan including information about the locations, and the detailed action generation unit generates the detailed actions, including interactions between the multiple characters, on a time-by-time basis, based on the plans of the multiple characters.

3. The detailed action generation unit generates detailed actions that are more detailed than the plan at each time interval, based on the plan, the emotional parameters, and the desire parameters relating to the desires of the characters, and further changes the emotional parameters and / or desire parameters based on the detailed actions, as described in claim 2.

4. The detailed action generation unit generates detailed actions that are more detailed than the plan at each time interval, based on the plan, the emotional parameters, and the desire parameters relating to the desires of the characters, and further changes the emotional parameters and / or desire parameters at regular intervals, as described in claim 2.

5. The generation system according to claim 1, wherein the story generation unit generates the detailed actions as a single scene of the story based on the time and relationship of the detailed actions.

6. The story generation system according to claim 1, wherein the story generation unit extracts events in the plan in which the conflict of the character is depicted, based on conflict information relating to the character's conflict and the plan, further extracts events in the plan that occur after those events as the conclusion of the story, further extracts events in the plan that occur before those events as the beginning of the story, and generates the story based on the results of the extraction.

7. The generation system according to any one of claims 1 to 6, wherein the plan generation unit, the detailed action generation unit, and the story generation unit are all trained learning models.

8. The generation system according to claim 7, comprising a generation unit and a model determination unit, wherein the generation unit receives the story information, and the model determination unit determines the language of the received story information and determines a learning model corresponding to the determined language.

9. A generation method to be performed by a generation system for generating a story, the generation system comprising: a plan generation unit; a detailed action generation unit; and a story generation unit, the method comprising: the plan generation unit generating a plan for a character after a predetermined period based on story information including an outline of the story and predetermined period action information relating to the character's actions during a predetermined period; the detailed action generation unit generating detailed actions relating to the character's actions at each time point, which are more detailed than the plan, based on the plan and emotional parameters relating to the character's emotions; and the story generation unit generating the story based on the detailed actions.

10. A generation program for generating a story, wherein a computer functions as a plan generation unit, a detailed action generation unit, and a story generation unit, the plan generation unit generates a plan for a character after a predetermined period based on story information including an outline of the story and predetermined period action information relating to the character's actions during a predetermined period, the detailed action generation unit generates detailed actions relating to the character's actions at each time point in time that are more detailed than the plan based on the plan and emotion parameters relating to the character's emotions, and the story generation unit generates the story based on the detailed actions.

Images