Computer system and scenario information provision method

Abstract

A server system (1100) receives given generation instruction information for generating scenario information and gives the generation instruction information to a prescribed generation unit that can generate scenario information to acquire candidate scenario information that is a candidate for scenario information. The server system (1100) generates an evaluation criterion for the candidate scenario information, evaluates the candidate scenario information on the basis of the evaluation criterion, and outputs the candidate scenario information as generated scenario information when the evaluation results satisfy a given acceptability condition.

Description

Computer System and Scenario Information Providing Method

[0001] The present invention relates to a computer system and the like.

[0002] Patent Document 1 describes training an AI (Artificial Intelligence) model from multiple game plays of a scenario using training state data collected from multiple game plays of the scenario and success criteria associated with each of the multiple game plays.

[0003] Japanese Patent Application Laid-Open No. 2022-33164

[0004] As one of the generation units represented by generative AI, a generative AI (generation unit) capable of generating scenario data is known. In the field of game production, methods for using the scenarios generated by the generation unit in game production are being explored.

[0005] In order to cause the generation unit to generate data, it is necessary to give the generation unit information (generation instruction information) for a generation instruction called a prompt. The generation instruction information includes multiple types of contexts that describe what to generate, under what preconditions to generate, and what data to refer to for generation.

[0006] The user creates generation instruction information so that a scenario (product) with the desired specifications and content can be obtained, but the scenario generated is not necessarily as desired by the user. If the generated scenario does not have the specifications and content desired by the user, the user modifies the generation instruction information and gives it to the generation unit again to re-execute the generation. The user has to repeat this, which is one of the reasons why the production efficiency does not increase as expected despite using the generation unit.

[0007] Such problems are not limited to the use of the generation unit in the video game production site. For example, the same applies when using scenarios generated by the generation unit in movie production, drama production, novel writing, etc.

[0008] The problem that this invention aims to solve is to provide a new technology that enables improved efficiency when generating and acquiring scenarios using a generation unit.

[0009] The first disclosure is a computer system comprising: a generation instruction information receiving unit that receives given generation instruction information for generating scenario information; a candidate scenario information acquisition control unit that performs control to acquire candidate scenario information, which is a candidate for scenario information, by providing instruction information based on the generation instruction information to a predetermined generation unit capable of generating scenario information; and an evaluation unit that generates evaluation criteria for the candidate scenario information relating to scenario theme information, which is scenario premise information and / or scenario character information based on the generation instruction information, and evaluates the candidate scenario information based on the evaluation criteria.

[0010] According to the first disclosure, the computer system will be able to evaluate the scenarios generated by the generation AI. Previously, users had to read and evaluate the content of each scenario they generated themselves, but this will be eliminated. Users will be able to acquire scenarios more efficiently than before.

[0011] The second disclosure is a computer system further comprising an output control unit that controls the output of candidate scenario information as generated scenario information, wherein the evaluation result of the evaluation unit satisfies a given pass condition.

[0012] According to the second disclosure, the computer system will be able to output a successful scenario. Users will be able to obtain the scenario more efficiently.

[0013] The third disclosure is a computer system in which, in the above-mentioned computer system, the candidate scenario information acquisition control unit includes feedback information based on the evaluation result in the instruction information and provides it to the generation unit as new instruction information, and performs reacquisition control to acquire new candidate scenario information.

[0014] Furthermore, the fourth disclosure is that in the above-mentioned computer system, the candidate scenario information acquisition control unit is a computer system that performs the reacquisition control when the evaluation result does not meet the passing conditions.

[0015] According to the third or fourth disclosure, the computer system will be able to receive feedback based on the evaluation results and re-acquire (re-generate) new scenarios. For the user, this will allow them to acquire scenarios more efficiently.

[0016] The fifth disclosure is a computer system in which, in the above-described computer system, the candidate scenario information acquisition control unit sets a randomness control parameter value for controlling the randomness of the generated information by the generation unit, and performs control to acquire the candidate scenario information by including the randomness control parameter value in the instruction information and providing it to the generation unit.

[0017] Furthermore, the sixth disclosure is a computer system in which, in the above-mentioned computer system, the candidate scenario information acquisition control unit sets a randomness control parameter value for controlling the randomness of the generated information by the generation unit in a variable manner according to the number of times the reacquisition control is executed, and performs control to acquire the candidate scenario information by including the randomness control parameter value in the instruction information and providing it to the generation unit.

[0018] A scenario generation unit can be implemented as a pre-trained AI model (e.g., LLM; Large Language Model) using a vast number of scenarios. During the learning process, the generation unit learns and understands the structure, context, and word order trends (probability of occurrence) of scenarios.

[0019] The generation unit then determines, in a sense, the scenario to be generated probabilistically based on the randomness control parameter value, under the generation conditions (prerequisites) specified in the generation instruction information. The randomness control parameter value can be said to be one of the parameters that determine the diversity and randomness of the generated scenarios.

[0020] According to the fifth or sixth disclosure, the computer system will be able to perform reacquisition while controlling the degree to which the content differs from the previously generated scenario.

[0021] The seventh disclosure is a computer system in which, in the above-mentioned computer system, the scenario information has story information consisting of multiple stages, and the candidate scenario information acquisition control unit changes the order of the stages to be generated among the multiple stages.

[0022] "Multiple stages" refers to, for example, the four divisions of introduction, development, turn, and conclusion, or the acts in the three-act structure of filmmaking. Furthermore, when generating a scenario in the generation unit, even if the generation instruction information has similar content, the output may differ depending on how the generation instruction information is written.

[0023] According to the seventh disclosure, the computer system can change the order of the generation steps when generating a single scenario.

[0024] The eighth disclosure is a computer system in which, in the above-mentioned computer system, the generation unit has element-specific generation units for each element that constitutes the scenario information, and the candidate scenario information acquisition control unit generates the instruction information to be given to the element-specific generation unit corresponding to each element that constitutes the candidate scenario information, based on the generation instruction information.

[0025] According to the eighth disclosure, the computer system will be able to generate a single scenario by generating the elements that make up the scenario information separately.

[0026] The ninth disclosure is a computer system in which, in the above-mentioned computer system, the candidate scenario information acquisition control unit performs control to variably determine the order in which the element-specific generation units that provide the instruction information are given.

[0027] According to the ninth disclosure, the computer system will be able to variably determine the order in which element-specific generating AIs provide instruction information.

[0028] The tenth disclosure is a computer system in which, in the above-mentioned computer system, the generation instruction information includes user information, and the candidate scenario information acquisition control unit generates the instruction information based on the user information.

[0029] According to the tenth disclosure, the computer system can include user information (for example, information linked to a user ID) in the generation instruction information and issue generation instructions to the generation unit based on this user information.

[0030] The eleventh disclosure is a computer system in which, in the above-mentioned computer system, the user information includes user attribute information, and the candidate scenario information acquisition control unit generates the instruction information based on the attribute information.

[0031] According to the eleventh disclosure, the computer system can include user attribute information in user information and issue generation instructions to the generation unit based on this attribute information.

[0032] The twelfth disclosure is a computer system in which, in the above-mentioned computer system, the user information includes user preference information, and the candidate scenario information acquisition control unit generates the instruction information based on the preference information.

[0033] According to the 12th disclosure, the computer system can include user preference information in user information and issue generation instructions to the generation unit based on this preference information.

[0034] Furthermore, the 13th disclosure is a computer system in which the candidate scenario information acquisition control unit has an external information retrieval unit that searches for related information related to the generation instruction information from an external information data source, and generates the instruction information based on the generation instruction information and the related information.

[0035] According to the 13th disclosure, the computer system will be able to generate instruction information by searching for and obtaining relevant information related to the generation instruction information from an external information data source.

[0036] The fourteenth disclosure is a computer system further comprising a keynote information generation unit that generates the scenario keynote information based on the generation instruction information in the computer system described above.

[0037] According to the 14th disclosure, the computer system will be able to generate scenario background information based on generation instruction information.

[0038] The 15th disclosure is a computer system in which the evaluation criteria are criteria for determining the extent to which the content of the scenario background information is satisfied.

[0039] According to the 15th disclosure, the computer system will be able to output candidate scenario information that satisfies the passing criteria as generated scenario information, based on the extent to which it satisfies the content of the scenario background information.

[0040] Furthermore, the sixteenth disclosure is a computer system in which, in the above-mentioned computer system, the evaluation unit analyzes the content of the scenario basic information for each given analysis item, generates criteria for each analysis item to determine the degree to which the content of the scenario basic information is satisfied as the evaluation criteria, the pass conditions include requirements based on the criteria for each analysis item, and the feedback information is information relating to the requirements included in the pass conditions that the evaluation result has not met.

[0041] According to the sixteenth disclosure, the computer system can analyze the content of the scenario background information for each given analysis item and determine whether or not it meets the passing criteria. It can then provide feedback to improve the analysis items that fail to pass. Therefore, it becomes possible to generate scenarios more efficiently.

[0042] The seventeenth disclosure is a computer system in which the scenario premise information includes at least one of the following: formal conditional information, information indicating the foundation of the scenario story, and information indicating the degree of realism of the scenario.

[0043] According to the 17th disclosure, the computer system can generate and output scenarios that meet at least any one of formal condition information, information indicating the basis of the scenario story, and information indicating the degree of realism of the scenario.

[0044] The 18th disclosure includes receiving given generation instruction information for generating scenario information, providing instruction information based on the generation instruction information to a predetermined generation unit capable of generating scenario information to perform control for obtaining candidate scenario information that is candidate scenario information, generating evaluation criteria for the candidate scenario information related to scenario premise information and / or scenario character information based on the generation instruction information, and evaluating the candidate scenario information based on the evaluation criteria. It is a scenario information providing method.

[0045] According to the 18th disclosure, a scenario information providing method having the same operational effects as the above-described computer system can be realized.

[0046] System configuration diagram showing a configuration example of a scenario providing system. Diagram for explaining the process of generating and providing a new scenario. Diagram for explaining the process of generating and providing a new scenario. Diagram showing an example of a user interface. Diagram for explaining the machine learning of plot generation AI. Diagram for explaining the machine learning of scene generation AI. Diagram for explaining the machine learning of dialogue generation AI. Diagram for explaining the machine learning of defect detection AI. Diagram for explaining the machine learning of scenario classification AI. Diagram showing an example of programs and data stored in a server system. Diagram showing a functional configuration example of a server processing unit. Flowchart for explaining the flow of scenario providing processing. Flowchart continuing from Figure 12. System configuration diagram of a modification example. Flowchart for explaining the flow of scenario providing processing of a modification example.

[0047] Hereinafter, examples of embodiments of the present invention will be described, but it is needless to say that the forms to which the present invention can be applied are not limited to the following embodiments.

[0048] FIG. 1 is a system configuration diagram showing a configuration example of a scenario providing system according to the present embodiment. The scenario providing system 1000 is a computer system that provides scenario data to a user 2.

[0049] The scenario providing system 1000 is a computer system including a server system 1100, an external information data source 1200, and user terminals 1500 for each user, which are communicably connected via a network 9. [[ID=...]]

[0050] The network 9 means a communication path capable of data communication. That is, the network 9 includes, in addition to a dedicated line (dedicated cable) for direct connection and a LAN (Local Area Network) such as Ethernet (registered trademark), a telephone communication network, a cable network, the Internet, and the like.

[0051] The server system 1100 is a computer system that performs various processes such as management control of registered user information and control related to scenario provision.

[0052] The server system 1100 mounts a control board ********** onto the main body device 110********** The control board 1150 is equipped with various microprocessors such as a CPU (Central Processing Unit) 1151, a GPU (Graphics Processing Unit), and a DSP (Digital Signal Processor), various IC memories 1152 such as VRAM, RAM, and ROM, and a communication device 1153. Note that some or all of the functions mounted on the control board 1150 may be realized by an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a SoC (System on a Chip). [[ID=..]]

[0053] In Figure 1, the server system 1100 is depicted as a single server device, but it may also be implemented using multiple devices. For example, the server system 1100 may be configured with multiple server devices, each responsible for a specific function, connected to each other via an internal bus or network 9 for data communication. Furthermore, the server system 1100 may include a database and online storage.

[0054] The external information data source 1200 is a database, data management server computer, or online storage to which the server system 1100 connects via the network 9. Various information (user information) linked to the user count assigned upon user registration is stored in the external information data source 1200. Of course, the user information may also be stored in the server system 1100.

[0055] The server system 1100 includes a scenario generation AI 10, a defect detection AI 14, and a scenario classification AI 16. These AI models are Large Language Models (LLMs) implemented through deep learning based on hardware employing a multi-core architecture (e.g., a group of GPUs and memory, a group of AI chips, etc.).

[0056] The scenario generation AI 10 generates text data for a scenario based on the given generation instruction information (prompt). Specifically, the scenario generation AI 10 has element-specific generation AIs 12 (12a, 12b, ...) that generate elements that constitute the scenario.

[0057] In this specification, three elements are adopted as constituting a scenario: plot, scene, and dialogue. Therefore, the scenario generation AI 10 has a plot generation AI 12a, a scene generation AI 12b, and a dialogue generation AI 12c. The construction of the AI ​​models for these generation AIs will be described later. Note that elements other than these may be set as constituting the scenario.

[0058] The defect detection AI14 detects plot holes, deus ex machina, dream sequences, and other flaws in the scenario.

[0059] The scenario classification AI 16 classifies the scenarios generated by the scenario generation AI 10. The classification items are set in correspondence with the prompt items (items that can be specified by prompt) of the generation instruction information given to the scenario generation AI 10.

[0060] Furthermore, the scenario generation AI 10, defect detection AI 14, and scenario classification AI 16 are not exclusive to the server system 1100. For example, these generation AIs may be implemented in external devices that the server system 1100 can connect to via the network 9, and may be external systems that can be used by customers other than the server system 1100.

[0061] The user terminal 1500 serves as the Man-Machine Interface (MMIF) for the scenario provision system 1000. Although only one terminal is depicted in Figure 1, in actual operation, it is typical for multiple user terminals 1500 to communicate with the server system 1100 simultaneously.

[0062] The user terminal 1500 is a computer system that can connect to the network 9, such as a personal computer, smartphone, wearable computer, portable game console, home game console, or tablet computer.

[0063] The user terminal 1500 is a computer comprising an operation input device, an image display device, a communication device, and a control board 1550 for performing calculations. Examples of the operation input device include a touch panel 1506, a keyboard, a game controller, and a mouse. Examples of the image display device include a touch panel 1506, a head-mounted display, and a glasses-type display.

[0064] The control board 1550 is equipped with a CPU 1551, various microprocessors such as a GPU and DSP, various IC memories 1552 such as VRAM, RAM, and ROM, and a communication module 1553 connected to the network 9. These elements mounted on the control board 1550 are electrically connected via bus circuits and the like, enabling data reading and writing, and signal transmission and reception. Part or all of the control board 1550 may be an ASIC, FPGA, or SoC.

[0065] The control board 1550 stores programs and various data necessary to realize the functions of the user terminal 1500 in the IC memory 1552. The user terminal 1500 realizes its functions as a terminal by executing a predetermined application program (for example, a client program) on the CPU 1551.

[0066] Figures 2 and 3 are diagrams illustrating the process of generating and providing a new scenario in the scenario provision system 1000. As shown in Figure 2, user 2 can request the scenario provision system 1000 to generate and provide a new scenario by performing a predetermined request operation on the user terminal 1500.

[0067] Figure 4 shows an example of a user interface (UI) related to request operations. When the server system 1100 detects that a predetermined request function has been invoked on the user terminal 1500, it displays the request setting UI 5 (5a, 5b, ...) for each setting item on the user terminal 1500.

[0068] The settings items mentioned here are not limited to the example shown in Figure 4. Also, while the request settings UI5 is illustrated as a pull-down menu, the UI design is not limited to this. For example, request settings UI5 could be a text input field that allows natural language input for each request item.

[0069] Request setting UI5 is provided for each prompt item that describes the generation conditions (prerequisites), quality, supplementary information, etc., in the generation instruction information given to the scenario generation AI 10. For example, Request setting UI5 (5a, 5b, ...) is provided separately for "genre," "plot keywords," "character settings," "dialogue," "plot points," "story length," "user-specified plot," etc. Multiple options are presented for selection in a single Request Setting UI5. Some items include options that the user can input in their own words (natural language).

[0070] The user terminal 1500 transmits the input result in the request setting UI 5 to the server system 1100 along with a predetermined request signal.

[0071] Returning to Figure 2, the server system 1100 creates origin generation instruction information 710, which is the origin of scenario generation, based on the input results in the request setting UI 5 received from the user terminal 1500. In other words, it creates a prompt to be given to the scenario generation AI 10.

[0072] The origin generation instruction information 710 includes scenario premise information 712, scenario character information 720, randomness control parameter values ​​722, user-specified plot information 724, and user information 726.

[0073] The scenario prerequisite information 712 includes formal conditional information 714, information 716 indicating the foundation of the scenario story, and information 718 indicating the degree of realism of the scenario. The formal conditional information 714 indicates, for example, the number of characters and lines in the scenario. For example, it may be described as a predetermined number of pages of 400-character manuscript paper.

[0074] Information 716, which forms the foundation of the scenario story, includes, for example, the genre and keywords in the synopsis.

[0075] Information 718 indicating the degree of realism in the scenario is, for example, the result of setting the genre in the request setting UI 5a (see Figure 4) to "Realistic," "High Fantasy," "Low Fantasy," etc.

[0076] Scenario character information 720 corresponds to the result of the setting operation on the request setting UI 5c regarding the settings of the characters appearing in the scenario.

[0077] The randomness control parameter value 722 is a parameter value that defines the probability distribution of the range in which the generating AI will select the next word, and it affects the creativity and randomness of the generated content. The parameter value known as "Temperature" also falls under this category.

[0078] The generation AI constructs generated data by sequentially selecting words with a high probability of appearing based on the type and context of the surrounding words. The randomness control parameter value 722 is initially set to a predetermined low value, which sets the range of "probability" to the narrowest possible state.

[0079] The randomness control parameter value 722 is changed so that its range gradually increases depending on the number of times the scenario is regenerated by the AI. As the number of regenerations increases, the range of the "probability" of selecting the next word widens, and the regenerated scenario becomes more creative, diverse, and random than the previous scenario. In other words, the content is more likely to differ from the previous one.

[0080] User-specified plot information 724 is a plot devised by user 2. User-specified plot information 724 is, for example, text data in the form of bullet points, categorized by the stages of story development (introduction, development, climax, and conclusion; acts in a three-act structure such as a movie).

[0081] User information 726 consists of attribute information (e.g., gender, age, place of residence, etc.) and preference information linked to user 2's personal identification information (e.g., user count). Preference information includes statistical data categorized by genre and synopsis keywords (see Figure 4) of novels and comics purchased by the user, and statistical data categorized by genre and synopsis keywords of works purchased or watched on video sites. The underlying data is stored in the external information data source 1200, and the information of the user who made the request is searched and set as appropriate.

[0082] Moving to Figure 3, the server system 1100 creates the origin generation instruction information 710, and then creates element-specific generation instruction information 750 from the origin generation instruction information 710, which is given to each of the element-specific generation AIs 12. Specifically, the content of the information included in the element-specific generation instruction information 750 differs depending on how the element-specific generation AI 12 was constructed through machine learning.

[0083] The plot generation AI 12a is trained to take the origin generation instruction information 710 and annotations as input, as shown in Figure 5, for example, and output plot information (text data of the plot). Plot information is information that shows the chain of causal relationships between events described in units of important events in the story, and can also be called a storyline. Each event unit in the plot information is associated with information indicating which stage of the story development (e.g., introduction, development, turn, conclusion) it corresponds to. For example, appropriate divisions may be made in the description of the plot information, and the divisions may indicate the stages.

[0084] Therefore, the element-specific generation instruction information 750 for the plot generation AI 12a is created by copying the origin generation instruction information 710. If the origin generation instruction information 710 contains user-specified plot data for each story development stage, this will also be included in the element-specific generation instruction information 750 for the plot generation AI 12a.

[0085] The scene generation AI 12b is trained to take origin generation instruction information 710, plot information, and annotations as input, as shown in Figure 6, for example, and output scene information. The scene information is text data of the scene, and is generated as text data that fills in the gaps between event units indicated by the plot information and does not change the causal relationships of the events, and includes one or more scene units such as stage directions and dialogue.

[0086] Therefore, the element-specific generation instruction information 750 for the scene generation AI 12b is created to include the origin generation instruction information 710 and the plot information generated by the plot generation AI 12a.

[0087] The dialogue generation AI 12c is trained to take as input, for example as shown in Figure 7, origin generation instruction information 710 as character setting information, scene information, and annotations, and to output scene information with adjusted dialogue.

[0088] Character setting information corresponds to the settings in the Request Setting UI5c for the characters appearing. Dialogue-adjusted scene information is scene information where the dialogue has been changed to a style more appropriate for the character indicated in the input character setting information.

[0089] Therefore, the element-specific generation instruction information 750 for the dialogue generation AI 12c is created to include the origin generation instruction information 710 (or just the character setting information from the origin generation instruction information 710), and scene information.

[0090] The server system 1100 creates and provides element-specific generation instruction information 750 for the plot generation AI 12a from the origin generation instruction information 710, and obtains plot information. Next, the server system 1100 takes the obtained plot information, includes it in element-specific generation instruction information 750 for the scene generation AI 12b, creates and provides it, and obtains scene information. Then, the server system 1100 includes the obtained scene information in element-specific generation instruction information 750 for the dialogue generation AI 12c, creates it, and provides it to the dialogue generation AI 12c, and obtains dialogue-adjusted scene information.

[0091] When the scene information with adjusted dialogue is integrated into the scene order, it becomes candidate scenario information that contains story information indicating the development of the story.

[0092] Returning to Figure 3, the server system 1100 creates and provides generation instruction information for the defect detection AI 14, which includes candidate scenario information.

[0093] The defect detection AI 14 is trained to take scenario information (plot information + dialogue-adjusted scene information; corresponding to candidate scenario information) and annotations as input, as shown in Figure 8, for example, and to output whether or not there are defects. The scenario information used as the training dataset contains a mix of scenarios that contain defects and those that do not. Therefore, the generation instruction information for the defect detection AI 14 includes candidate scenario information.

[0094] Examples of "defects" as used here include, for example, the following (1) to (4): (1) Inconsistencies in time and space. For example, descriptions that cannot be made unless characters teleport, descriptions that make past events seem like they will happen in the future, descriptions of conversations between characters or cause and effect that cannot be made without magic or super technology, etc. (2) Logical inconsistencies. For example, simple calculation errors, errors in deductive / inductive reasoning, etc. (3) Moderation violations. For example, descriptions of characters' actions or words that deviate from ethical and social good sense and common sense, etc. (4) Inconsistency in characters' language use outside of dramatic purposes. For example, a character who is supposed to be refined mixes vulgar actions and remarks, etc. (5) Character language use that is inappropriate for the time, place, and occasion. For example, a serious character using "ra-nuki" (omitting the "ra" sound in verbs), etc.

[0095] Furthermore, if the origin generation instruction information 710 includes a specification of the target audience (for example, the age range of the readers, the language used, etc.), the use of language or difficult words unsuitable for the target audience may also be included as one of the defects. Additionally, readability issues such as sentence length and formal style, including the use of punctuation, may also be included as one of the defects.

[0096] Alternatively, the defect detection AI 14 may output an evaluation score regarding the logic and consistency of a series of texts called a scenario, rather than outputting whether or not there is a defect. In that case, if the evaluation score reaches a predetermined passing score (for example, 80 points), it may be considered "no defect".

[0097] If there are defects in the candidate scenario information, the server system 1100 will perform a re-acquisition control of the scenario information. Specifically, it will restart the process from providing the plot generation AI 12a with element-specific generation instruction information 750 for the plot generation AI 12a.

[0098] If the candidate scenario information is free of defects, the server system 1100 creates generation instruction information for the scenario classification AI 16, including the candidate scenario information, provides it, and obtains the classification result of the candidate scenario information.

[0099] The scenario classification AI 16 is trained to take scenario information and annotations as input and output scenario classification information, as shown in Figure 9, for example. The classification items in the scenario classification information are set to correspond to the prompt items of the origin generation instruction information 710 (see Figure 4). Therefore, the generation instruction information for the scenario classification AI 16 includes candidate scenario information.

[0100] Returning to Figure 3, once the server system 1100 obtains the classification results of the candidate scenario information, it then generates evaluation criteria. The "evaluation criteria" are standards for determining the degree to which the content of the scenario background information (scenario prerequisite information 712 and scenario character information 720 in Figure 2) is satisfied. Specifically, the evaluation criteria are set as the match rate / similarity rate between the content of the prompt items in the origin generation instruction information 710 and the content of the classification results by the scenario classification AI 16. For example, it may be set as "match rate = 80% or higher".

[0101] When setting the evaluation criteria by similarity rate, a similarity recognition rule between the content of the prompt item and the content of the classification result must be defined separately. When "Automatic" (for example, Request Setting UI5b in Figure 4) is set in Request Setting UI5, the evaluation criteria may be set to none or essentially unconditional.

[0102] Furthermore, the evaluation criteria are not limited to a single standard; for example, different criteria may be set depending on the genre of the scenario. For instance, in the "realistic" genre, the evaluation criteria may be set more strictly to emphasize realism, while in the "high fantasy" genre, the premise of fantasy inherently allows for a more lenient approach, where readers are more likely to accept whatever happens. Therefore, in the "high fantasy" genre, the evaluation criteria may be set more leniently.

[0103] Next, the server system 1100 evaluates the candidate scenario information based on evaluation criteria and determines whether it passes or fails. Specifically, the server system 1100 extracts items whose content does not match between the prompt item and the classification item, and calculates the percentage of all items whose content matches (match rate). If the match rate meets the evaluation criteria, the candidate scenario information is judged to be successful. In other words, the passing condition includes multiple requirements for each prompt item, namely "match between the content of the prompt item and the content of the classification result corresponding to that item," and the evaluation criteria indicate how many of these multiple requirements are met.

[0104] If the candidate scenario information is unsatisfactory, the server system 1100 generates feedback information based on the evaluation result, adds this to the element-specific generation instruction information 750 used in the previous generation, and executes control to reacquire the candidate scenario information.

[0105] "Feedback information" refers to information used to improve analysis items related to requirements that were not met in the evaluation results, among the requirements included in the passing criteria. For example, it may include additional prompts or extended prompts aimed at bringing prompt items that did not match the classification results in a matching manner.

[0106] For example, suppose that in the origin generation instruction information 710, the first character (see request setting UI 5c in Figure 4) is set to "female" or "boy / girl," but the classification result of the candidate scenario information indicates that the first character is "female" or "adult." In this case, the feedback information would be such that the first character is classified as "boy / girl." Specifically, in a configuration where the origin generation instruction information 710 allows natural language input, additional prompts or extended prompts such as "Please have the first character engage in conversation befitting a boy / girl" may be generated as feedback information. Adding feedback information makes it easier to generate candidate scenario information that will pass the re-acquisition control.

[0107] For example, suppose the classification result of the candidate scenario information includes characters that are not specified in the origin generation instruction information 710 (this includes an omnipotent god and the reader). In this case, the feedback information would add a condition that characters not specified in the origin generation instruction information 710 should not appear.

[0108] For example, in the classification results of candidate scenario information, if the genre, theme, or situation (including so-called dream endings) of the candidate scenario information differs from the settings of the origin generation instruction information 710, the feedback information will contain content that reinforces the conditions related to the genre and situation of the origin generation instruction information 710.

[0109] If the candidate scenario information is satisfactory, the server system 1100 outputs the approved candidate scenario information as generated scenario information.

[0110] Internally, the server system 1100 may repeat the re-acquisition control several times. However, from the user 2's perspective, when a request operation is performed, it appears as if generated scenario information that satisfies the request content is output.

[0111] In light of the structure of this specification, conventionally, candidate scenario information was output directly from the generating AI as generated scenario information. Therefore, conventionally, user 2 had to read and evaluate the content of the candidate scenario information (= generated scenario information) each time it was generated, and if it was insufficient, repeat the request. However, according to this embodiment, user 2 is freed from this repetitive work.

[0112] Figure 10 shows an example of programs and data stored by the server system 1100. The server system 1100 stores the server program 501 and the distribution client program 503, which is the original client program provided to the user terminal 1500, in the IC memory 1152. The server system 1100 also stores the UI definition data 510 of the request setting UI 5, the trained generation AI model 512, the feedback definition data 514, and the origin generation instruction information 710. The server system 1100 also stores the number of request outputs 740 related to one request, element-specific generation instruction information 750, candidate scenario information 752, evaluation criterion data 754, and evaluation result data 756. The server system 1100 also stores the number of generation executions 758 related to one request, feedback information 760, and generation scenario information 762. Of course, other data may be stored as appropriate.

[0113] The server system 1100 performs the server program 501 and calculates the results on the CPU 1151, thereby realizing the function of a server processing unit 200s, as shown in Figure 11. The server program 501 may include one or more generation AI programs 502 for realizing the functions of a scenario generation AI 10, a defect detection AI 14, and a scenario classification AI 16. Alternatively, the generation AI programs 502 may be stored separately from the server program 501.

[0114] The server processing unit 200s performs various controls from generating a scenario using AI in response to a request operation from user 2 until it is output. Specifically, as shown in Figure 11, the server processing unit 200s includes a generation instruction information receiving unit 204, a candidate scenario information acquisition control unit 210, a generation unit 218, an evaluation unit 220, and an output control unit 230.

[0115] The generation instruction information receiving unit 204 receives given generation instruction information for generating scenario information. This includes control related to the acceptance of request operations.

[0116] The candidate scenario information acquisition control unit 210 performs control to acquire candidate scenario information, which is a candidate for scenario information, by providing instruction information based on the generation instruction information to a predetermined generation AI capable of generating scenario information. The candidate scenario information acquisition control unit 210 also performs re-acquisition control to acquire new supplementary scenario information by including feedback information based on the evaluation result in the instruction information and providing it to the generation AI as new instruction information.

[0117] The candidate scenario information acquisition control unit 210 includes an external information retrieval unit 212 and a key information generation unit 214.

[0118] The external information retrieval unit 212 retrieves user 2's attribute information and preference information from the user registration information 600 (see Figure 10) stored in the external information data source 1200 as related information related to the generation instruction information.

[0119] The key information generation unit 214 generates scenario key information based on the generation instruction information. Specifically, it automatically generates prompt items that are set to "automatic" in the origin generation instruction information 710. For example, it automatically generates scenario prerequisite information 712 and scenario character information 720 (see Figure 2).

[0120] The generation unit 218 generates a scenario in response to a request operation. The generation unit 218 is implemented by the scenario generation AI 10. Therefore, in a configuration where the scenario generation AI 10 is implemented outside the server system 1100, the generation unit 218 is not included in the server processing unit 200s.

[0121] The evaluation unit 220 generates evaluation criteria for candidate scenario information relating to scenario prerequisite information and / or character information based on the generation instruction information, and evaluates the candidate scenario information and determines whether it passes or fails based on these evaluation criteria (see Figure 3).

[0122] In this embodiment, the evaluation unit 220 is implemented by a defect detection AI 14 and a scenario classification AI 16. In a configuration where the defect detection AI 14 and the scenario classification AI 16 are implemented outside the server system 1100, the evaluation unit 220 is not included in the server processing unit 200s.

[0123] The output control unit 230 controls the output of candidate scenario information as generated scenario information, where the evaluation result of the evaluation unit 220 satisfies the given passing conditions.

[0124] Returning to Figure 10, the feedback definition data 514 is the data that forms the basis for generating the feedback information 760, and various patterns are provided for different situations where the content of the prompt item and the classification item of the candidate scenario do not match.

[0125] A single feedback definition data 514 stores the application requirements and the base data that forms the basis of the feedback information 760. The application requirements are described by the type of non-conforming prompt item, which is the condition that must be met for that type of non-conforming prompt item, and the item content, which is the condition that must be met for the content of that item. The base data, for example, defines the content of an additional prompt.

[0126] The feedback information 760 is generated based on the base data of the feedback definition data 514 that satisfies the application requirements, in order to conform to the format of the generation instruction for the element-specific generation AI 12 that generates scenario components related to non-conforming items.

[0127] Figures 12 and 13 are flowcharts illustrating the flow of scenario provision processing related to the generation and output of scenarios executed by the server system 1100. The server system 1100 receives a request operation from user 2 (step S10) and receives the number of requested outputs, which is the number of scenarios to be output in relation to the request (step S12).

[0128] Next, the server system 1100 creates origin generation instruction information 710 in response to the request operation (step S14), initializes the execution count 758 to "1" (step S16), and performs the initial acquisition control of candidate scenario candidates (step S18).

[0129] In step S14, the server system 1100, for prompt items set to "automatic," refers to user 2's attribute information and preference information from the external information data source 1200 and automatically sets content that matches them. For example, suppose the setting for the "genre" item is "automatic." In this case, if it is determined from the genre-specific statistical data of movies watched, which is preference information, that user 2 tends to prefer watching romance movies, the setting for the "genre" item may be set to "romance."

[0130] In the initial acquisition control of step S18, element-specific generation instruction information 750 is created and provided to the element-specific generation AI 12 in a predetermined order (plot generation AI 12a → scene generation AI 12b → dialogue generation AI 12c) to acquire the scenario components. Then, the acquired scenario components are integrated into a single candidate scenario information 752.

[0131] Next, the server system 1100 performs defect evaluation of the candidate scenarios (step S40). That is, it creates and provides generation instruction information for the defect detection AI 14, which includes candidate scenario information 752.

[0132] If there is a defect in the candidate scenario (YES in step S42), the server system 1100 provides the scenario generation AI 10 with the same generation instruction information (element-specific generation instruction information 750) used in the initial generation to control the reacquisition of new candidate scenario information 752 (step S44). Note that the generation instruction information for reacquisition may include feedback information (for example, a description of additional conditions) to correct defects detected in the defect evaluation, in addition to the generation instruction information used in the initial generation.

[0133] If there are no defects in the candidate scenario (NO in step S42), the server system 1100 creates instruction information including candidate scenario information 752 and has the scenario classification AI 16 classify the candidate scenario (step S50). Then, the server system 1100 creates evaluation criteria (step S52) and determines whether the candidate scenario is pass or fail (step S54).

[0134] Moving to Figure 13, if the candidate scenario is unsuccessful (NO in step S60), the server system 1100 generates feedback information 760 for the prompt items whose content does not match the classification result (step S62). Then, it adds and updates the feedback information 760 to the element-specific generation instruction information 750 (step S64).

[0135] Next, the server system 1100 determines whether the origin generation instruction information 710 contains user-specified plot information. If the determination is positive (YES in step S66), the server system 1100 changes the order in which the user-specified plot information is described in the element-specific generation instruction information 750 for the plot generation AI 12a (step S68).

[0136] Next, the server system 1100 changes the randomness control parameter value of the element-specific generation instruction information 750 according to the execution count 758 (step S70), performs reacquisition control (step S72), and increases the execution count 758 by "1" (step S74).

[0137] The candidate scenarios regenerated by the reacquisition control are again classified by the scenario classification AI 16 and judged as pass or fail (steps S50 to S54 in Figure 12).

[0138] If a candidate scenario is successful (YES in step S60), the server system 1100 saves it and counts the number of successful candidate scenario information 752. If the count does not reach the requested output number 740 (NO in step S62), the server system 1100 controls the reacquisition of the instruction information (element-specific generation instruction information 750) to be given to the scenario classification AI 16 (step S82).

[0139] If the number of approved candidate scenario information 752 reaches the number of requested outputs 740 (YES in step S80), the server system 1100 outputs the approved candidate scenario information 752 as generated scenario information (step S84) and terminates the series of processes.

[0140] As described above, this embodiment makes it possible to improve the efficiency of generating and acquiring scenarios using the generation unit. The generated scenario information 762 is free from defects such as plot holes and conforms to the request. Unlike in the past, users do not need to read the content themselves and determine whether it is acceptable or not as soon as the scenario is generated, allowing users to efficiently obtain scenarios.

[0141] [Modifications] Although examples of embodiments to which the present invention is applied have been described above, the forms to which the present invention can be applied are not limited to the above forms, and components can be added, omitted, or modified as appropriate.

[0142] (Modification 1) For example, although the scenario provision system 1000 was exemplified as a client-server type, multiple user terminals 1500 may be implemented in a P2P (Peer to Peer) architecture. In this case, programs and data corresponding to the division of functions are stored in the user terminals 1500, and functions corresponding to the server processing unit 200s in the above embodiment are implemented in a distributed manner by the user terminals 1500 which become P2P nodes. The same effects as in the above embodiment can be obtained in this configuration as well.

[0143] (Modification 2) In the above embodiment, the server system 1100 was exemplified as having a scenario generation AI 10, a defect detection AI 14, and a scenario classification AI 16, but it is not limited to this. As shown in the scenario provision system 1000B in Figure 14, the user terminal 1500B may be configured to have these AIs. In this configuration, the server system 1100B transmits instruction information (element-specific generation instruction information 750) to the AI ​​of the user terminal 1500B via the network 9 and acquires the generated data.

[0144] (Modification 3) In the above embodiment, the scenario provisioning system 1000 may be implemented not as a client-server type, but as a single computer system which was the user terminal 1500 in the above embodiment.

[0145] Specifically, the user terminal 1500B in Figure 14 has a scenario generation AI 10, a defect detection AI 14, and a scenario classification AI 16, and stores all the data that the server system 1100 in the above embodiment is supposed to store (see Figure 10). However, instead of the server program 501 and the distribution client program 503, a scenario provision program is provided as an application program for the user terminal 1500.

[0146] In the above embodiment, the scenario provision program enables the user terminal 1500B to implement all of the functional units of the server system 1100 (see Figure 11). In this modified example, the scenario provision program is executed by the user terminal 1500B. The processing flow in the above embodiment (see Figures 12 and 13) can be interpreted by simply replacing the execution entity from the server system 1100 to the user terminal 1500B.

[0147] (Modification 4) In the above embodiment, the scenario components constituting the candidate scenario were plot, story, and dialogue. However, it is also possible to divide the plot into story development stages (for example, introduction, development, climax, and conclusion), and make each of these into a single component.

[0148] In this configuration, the plot generation AI 12a is constructed as an AI model capable of generating plots for a specified stage in the story development process. The element-specific generation instruction information for the plot generation AI 12a in this configuration is created to include origin generation instruction information 710, generation stage information indicating which stage of the story development process the plot for will be generated, and plot information for other stages that have already been generated. The plot generation AI 12a in this configuration generates plots for the stage specified in the generation stage so as to connect to plot information for other stages.

[0149] Figure 15 is a flowchart illustrating the flow of the scenario provision process (scenario provision process B) in the modified example, and corresponds to the flowchart in Figure 12. Figure 15 leads to Figure 13.

[0150] In scenario provisioning process B, steps S22 to S30 are executed instead of step S18 in the above embodiment. Also, step S46 is executed instead of step S44.

[0151] Specifically, the server system 1100 initially sets the order in which to instruct the generation of each element (generation instruction order) to be in the order of "introduction plot → development plot → turning point plot → conclusion plot → scene → dialogue" (step S22). Then, it executes loop A for each scenario component according to this correct generation instruction order (steps S24 to S30).

[0152] In loop A, the server system 1100 creates element-specific generation AI for the element-specific generation AI 12 corresponding to the scenario component to be processed (step S26). Then, it acquires the scenario component to be processed (step S28) and ends loop A for the processing target (step S30). Then, as in the above embodiment, once all the scenario components have been acquired, the candidate scenario information 752 is established.

[0153] In scenario provision process B, if a candidate scenario has a defect (NO in step S42), the generation instruction order for the plot's scenario components is changed to "Conclusion → Turning Point → Development → Beginning" (step S46), and the process returns to loop A. In other words, the generation instruction order is reversed, and candidate scenario information 752 is reacquired. By reversing the order, the final "Conclusion" stage is determined first, the "Turning Point" stage that leads to this "Conclusion" is generated, followed by the "Development" and "Beginning" stages. Therefore, by generating candidate scenarios that follow the chronological story development in reverse order, it is possible to generate candidate scenarios that are less likely to have defects.

[0154] (Modification 5) In the above embodiment, the element-specific generation AI 12 used for the initial generation and the element-specific generation AI 12 used for reacquisition control were described as the same AI. However, the element-specific generation AI 12 used for reacquisition control may be a different AI from the one used previously.

[0155] For example, the server system 1100 may have multiple AI models for plot generation AI 12a, scene generation AI 12b, and dialogue generation AI 12c, each with different machine learning approaches.

[0156] In step S18 (see Figure 12), which is the initial generation step, the server system 1100 selects one of the AI ​​models for plot generation AI 12a, scene generation AI 12b, and dialogue generation AI 12c, and provides it with generation instruction information. The server system then remembers which AI model was used, and in step S44 and / or step S72 (see Figure 13), it may select a different AI model than the one used previously and provide it with generation instruction information.

[0157] The use of a different AI model than the one used previously may be applied when the number of executions reaches a predetermined number.

[0158] (Modification 6) The scenario generation AI 10 may also have a function for evaluating defects in candidate scenarios. In that case, the defect detection AI 14 and the process for when defects are found from the defect evaluation (steps S40 to S44 in Figure 12) may be omitted.

[0159] Furthermore, if an AI is available that integrates the defect detection AI 14 and the scenario classification AI 16, the process from defect evaluation to the case where a defect is found may be integrated with step S54.

[0160] (Modification 7) In the above embodiment, if there are multiple requested outputs, step S84 provides multiple similar scenarios that satisfy the acceptance criteria originating from one origin generation instruction information 710, and presents all of them to the user 2, but is not limited to this.

[0161] For example, in step S84, the generation scenario information prepared may be configured to output the one with the highest evaluation in step S54. Alternatively, for example, the scenario generation AI 10 may be trained to output the evaluation result of a document based on the evaluation criteria described in the generation instruction information, which is written in natural language and requests an evaluation of a document. The generation instruction information may be prepared in advance, and in step S84, the scenario generation AI 10 may be given the generation instruction information along with multiple generation scenario information, and only the one with the highest evaluation among the evaluations it receives may be output.

[0162] 2...User 10...Scenario generation AI 12...Element-specific generation AI 12a...Plot generation AI 12b...Scene generation AI 12c...Dialogue generation AI 14...Defect detection AI 16...Scenario classification AI 200s...Server processing unit 204...Generation instruction information receiving unit 210...Candidate scenario information acquisition control unit 212...External information retrieval unit 214...Key information generation unit 218...Generation unit 220...Evaluation unit 230...Output control unit 501...Server program 512...Trained generation AI model 514...Feedback definition data 710...Origin generation instruction information 712...Scenario prerequisite information 714...Formal condition information 716...Information indicating the foundation of the scenario story 718...Information indicating the degree of realism of the scenario 720...Scenario character information 722...Randomness control parameter value 724...User-specified plot information 726...User information 750... Element-specific generation instruction information 752... Candidate scenario information 754... Evaluation criteria data 760... Feedback information 762... Generated scenario information 1000... Scenario provision system 1100... Server system 1200... External information data source 1500... User terminal

Claims

1. A computer system comprising: a generation instruction information receiving unit that receives given generation instruction information for generating scenario information; a candidate scenario information acquisition control unit that performs control to acquire candidate scenario information, which is a candidate for scenario information, by providing instruction information based on the generation instruction information to a predetermined generation unit capable of generating scenario information; and an evaluation unit that generates evaluation criteria for the candidate scenario information relating to scenario theme information, which is scenario premise information and / or scenario character information based on the generation instruction information, and evaluates the candidate scenario information based on the evaluation criteria.

2. The computer system according to claim 1, further comprising: an output control unit that performs control to output candidate scenario information as generated scenario information if the evaluation result of the evaluation unit satisfies a given pass condition.

3. The computer system according to claim 2, wherein the candidate scenario information acquisition control unit includes feedback information based on the evaluation result in the instruction information and provides it to the generation unit as new instruction information, and performs reacquisition control to acquire new candidate scenario information.

4. The computer system according to claim 3, wherein the candidate scenario information acquisition control unit performs the reacquisition control when the evaluation result does not satisfy the passing conditions.

5. The computer system according to any one of claims 1 to 4, wherein the candidate scenario information acquisition control unit sets a randomness control parameter value for controlling the randomness of the information generated by the generation unit, and performs control to acquire the candidate scenario information by including the randomness control parameter value in the instruction information and providing it to the generation unit.

6. The computer system according to claim 3 or 4, wherein the candidate scenario information acquisition control unit sets a randomness control parameter value for controlling the randomness of the generated information by the generation unit in a variable manner according to the number of times the reacquisition control is executed, and performs control to acquire the candidate scenario information by including the randomness control parameter value in the instruction information and providing it to the generation unit.

7. The computer system according to any one of claims 1 to 6, wherein the scenario information comprises story information composed of multiple stages, and the candidate scenario information acquisition control unit changes the order of the stages to be generated from among the multiple stages.

8. The computer system according to any one of claims 1 to 7, wherein the generation unit has element-specific generation units for generating elements constituting the scenario information, and the candidate scenario information acquisition control unit generates instruction information to be given to the element-specific generation units corresponding to each element constituting the candidate scenario information based on the generation instruction information.

9. The computer system according to claim 8, wherein the candidate scenario information acquisition control unit performs control to variably determine the order in which the element-specific generation units that provide the instruction information.

10. The computer system according to any one of claims 1 to 9, wherein the generation instruction information includes user information, and the candidate scenario information acquisition control unit generates the instruction information based on the user information.

11. The computer system according to claim 10, wherein the user information includes user attribute information, and the candidate scenario information acquisition control unit generates the instruction information based on the attribute information.

12. The computer system according to claim 10 or 11, wherein the user information includes user preference information, and the candidate scenario information acquisition control unit generates the instruction information based on the preference information.

13. The computer system according to any one of claims 1 to 12, wherein the candidate scenario information acquisition control unit includes an external information retrieval unit that retrieves related information related to the generation instruction information from an external information data source, and generates the instruction information based on the generation instruction information and the related information.

14. The computer system according to any one of claims 1 to 13, further comprising: a keynote information generation unit that generates the scenario keynote information based on the generation instruction information.

15. The computer system according to any one of claims 1 to 14, wherein the evaluation criteria are criteria for determining the extent to which the content of the scenario background information is satisfied.

16. The computer system according to claim 3 or 4, wherein the evaluation unit analyzes the content of the scenario basic information for each given analysis item, generates criteria for each analysis item to determine the degree to which the content of the scenario basic information is satisfied as the evaluation criteria, the pass condition includes requirements based on the criteria for each analysis item, and the feedback information is information relating to the requirements included in the pass condition that the evaluation result has not met.

17. The computer system according to any one of claims 1 to 16, wherein the scenario prerequisite information includes at least one of formal conditional information, information indicating the foundation of the scenario story, and information indicating the degree of realism of the scenario.

18. A method for providing scenario information, comprising: receiving a given generation instruction information for generating scenario information; providing instruction information based on the generation instruction information to a predetermined generation unit capable of generating scenario information, thereby controlling the acquisition of candidate scenario information which is a candidate for scenario information; and generating evaluation criteria for the candidate scenario information relating to scenario prerequisite information and / or character information appearing in the scenario, based on the generation instruction information, and evaluating the candidate scenario information based on the evaluation criteria.

Images