Proxy existence model management device, proxy existence model management method, proxy existence model management program, and recording medium

Abstract

Provided is a proxy existence model management device that makes it possible to easily operate a plurality of proxy existence models.　A proxy existence model management device according to the present disclosure includes: a model pool management unit that associates at least one proxy existence model and model feature information of the proxy existence model with each other and manages the same as a proxy existence model pool; an input information acquisition unit that acquires input information of a user; an input analysis unit that analyzes the input information; a model selection unit that selects at least one proxy existence model simulating a target person from the proxy existence model pool, on the basis of the analysis result of the input information and the model feature information; an information generation unit that gives the input information to the selected proxy existence model to generate information; and an output unit that outputs the information generated by the proxy existence model. Here, the proxy existence model is a model configured to simulate the behavior of a model target person with respect to a large-scale learning model.

Description

Agent Presence Model Management Device, Agent Presence Model Management Method, Agent Presence Model Management Program, and Recording Medium

[0001] The present disclosure relates to an agent presence model management device, an agent presence model management method, an agent presence model management program, and a recording medium.

[0002] Recently, a technique for executing a predetermined task by a machine learning model called generative artificial intelligence has been known. For example, in Patent Document 1, a reception unit that receives questions about products, and an AI agent that generates and outputs information according to the input information, inputs input information including the information of the question and information for outputting an answer to the question, and causes the AI agent to generate answer information that is information indicating an answer to the question. And a providing unit that provides the answer information generated by the generating unit. In Patent Document 1, when a target is given to the AI agent, the AI agent generates a task for achieving the target for a large language model, collects information for causing the generated task to be executed by the large language model, and repeats the process of causing the large language model to execute the task. It is an AI agent that outputs information for achieving the target.

[0003] Japanese Patent No. 7509972

[0004] By the way, the inventors of the present disclosure have constructed an agent presence model as an AI agent that can learn the personality information of the model target person and behave as the model target person. Here, as the number of types of the constructed agent presence models increases, there is a problem that it becomes difficult to select an appropriate model for the purpose.

[0005] Therefore, an object of the present disclosure is to provide an agent presence model management device, an agent presence model management method, an agent presence model management program, and a recording medium that enable easy operation of a plurality of agent presence models.

[0006] To achieve the above objective, the proxy existence model management device of this disclosure includes a model pool management unit, an input information acquisition unit, an input analysis unit, a model selection unit, an information generation unit, and an output unit, wherein the model pool management unit manages a proxy existence model pool by linking at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition unit acquires user input information, the input analysis unit analyzes the input information, the model selection unit selects at least one proxy existence model that imitates a target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation unit provides the selected proxy existence model with the input information to generate information, and the output unit outputs the information generated by the proxy existence model.

[0007] The proxy existence model management method of this disclosure includes a model pool management step, an input information acquisition step, an input analysis step, a model selection step, an information generation step, and an output step, wherein the model pool management step manages a proxy existence model pool by associating at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition step acquires user input information, the input analysis step analyzes the input information, the model selection step selects at least one proxy existence model that imitates a target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation step provides the selected proxy existence model with the input information to generate information, and the output step outputs the information generated by the proxy existence model, each step being performed by a computer.

[0008] The proxy existence model management program of this disclosure includes a model pool management procedure, an input information acquisition procedure, an input analysis procedure, a model selection procedure, an information generation procedure, and an output procedure, wherein the model pool management procedure manages a proxy existence model pool by associating at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition procedure acquires user input information, the input analysis procedure analyzes the input information, the model selection procedure selects at least one proxy existence model that imitates a target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation procedure provides the selected proxy existence model with the input information to generate information, and the output procedure outputs the information generated by the proxy existence model, thereby causing the computer to execute each of these procedures.

[0009] The recording medium of this disclosure includes a model pool management procedure, an input information acquisition procedure, an input analysis procedure, a model selection procedure, an information generation procedure, and an output procedure, wherein the model pool management procedure manages a proxy existence model pool by associating at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition procedure acquires user input information, the input analysis procedure analyzes the input information, the model selection procedure selects at least one proxy existence model that imitates a target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation procedure provides the selected proxy existence model with the input information to generate information, and the output procedure outputs the information generated by the proxy existence model. The recording medium is a computer-readable recording medium that records a proxy existence model management program that causes a computer to execute each of the above procedures.

[0010] According to this disclosure, multiple surrogate existence models can be easily operated.

[0011] Figure 1 is a block diagram showing the configuration of an example of the proxy existence model management device of this disclosure. Figure 2 is a block diagram showing an example of the hardware configuration of the proxy existence model management device of this disclosure. Figure 3 is a flowchart showing an example of processing in the proxy existence model management device of this disclosure. Figure 4 is a conceptual diagram showing an example of the construction of a proxy existence model pool. Figure 5 is a block diagram showing the configuration of an example of the proxy existence model manufacturing device of this disclosure. Figure 6 is a block diagram showing an example of the hardware configuration of the proxy existence model manufacturing device of this disclosure. Figure 7 is a flowchart showing an example of processing in the proxy existence model manufacturing device of this disclosure.

[0012] In this disclosure, the “surrogate existence model” is a machine learning model that has learned the personality information of a model subject so that it can behave like the model subject. The method for manufacturing the surrogate existence model is not particularly limited and can be manufactured by any method, but it can be manufactured, for example, by the method for manufacturing the surrogate existence model described in this disclosure. The surrogate existence model in this disclosure may be a machine learning model that has learned the personality information of a predetermined mentor, counselor, researcher, coach, teacher, consultant, etc., as the model subject.

[0013] Next, embodiments of the present disclosure will be described with reference to the drawings. The present disclosure is not limited to the following embodiments. In the following drawings, the same parts are denoted by the same reference numerals. Furthermore, unless otherwise specified, the descriptions of each embodiment can be used interchangeably with those of the others, and unless otherwise specified, the configurations of each embodiment can be combined.

[0014] [Embodiment 1] The proxy existence model management device of this embodiment will be described with reference to Figure 1. Figure 1 is a block diagram showing the configuration of an example of the proxy existence model management device 10 of this embodiment. As shown in Figure 1, the proxy existence model management device 10 (hereinafter also referred to as "this device 10") includes a model pool management unit 11, an input information acquisition unit 12, an input analysis unit 13, a model selection unit 14, an information generation unit 15, and an output unit 16. In addition, although not shown, this device 10 may also include, for example, an input unit, an output unit, a display unit and / or a storage unit.

[0015] The device 10 may be, for example, a single device including the aforementioned parts, or it may be a device in which the aforementioned parts can be connected via a communication network. Furthermore, the device 10 can be connected to an external device described later via a communication network. The communication network is not particularly limited and a known network can be used, for example, it may be wired or wireless. Examples of communication networks include the Internet, WWW (World Wide Web), telephone lines, LAN (Local Area Network), SAN (Storage Area Network), DTN (Delay Tolerant Networking), LPWA (Low Power Wide Area), L5G (Local 5G), etc. Examples of wireless communication include Wi-Fi®, Bluetooth®, Local 5G, LPWA, etc. The wireless communication may be in the form of direct communication between devices (Ad Hoc communication), infrastructure communication, or indirect communication via an access point. The device 10 may, for example, be incorporated into a server as a system. The device 10 may also be, for example, a personal computer (PC, e.g., desktop or notebook), smartphone, or tablet terminal on which the program disclosed herein is installed. Furthermore, the device 10 may be in the form of cloud computing or edge computing, for example, in which at least one of the aforementioned parts is on a server and the other parts are on a terminal. As a specific example, the device 10 may be in the form of a model pool management device including a model pool management unit 11, and an interactive execution device with a surrogate existence model, which includes at least one configuration selected from the group consisting of an input information acquisition unit 12, an input analysis unit 13, a model selection unit 14, an information generation unit 15, and an output unit 16.

[0016] Figure 2 illustrates a block diagram of the hardware configuration of the device 10. The device 10 includes, for example, a central processing unit 101, memory 102, bus 103, storage device 104, input device 105, output device 106, communication device (communication unit) 107, etc. Each part of the device 10 is interconnected via the bus 103 through its respective interface (I / F).

[0017] The central processing unit 101 operates in coordination with other components via controllers (system controller, I / O controller, etc.) and is responsible for the overall control of the device 10. In the device 10, the central processing unit 101 executes, for example, the program disclosed herein and other programs, and also reads and writes various types of information. Specifically, for example, the central processing unit 101 functions as a model pool management unit 11, an input information acquisition unit 12, an input analysis unit 13, a model selection unit 14, an information generation unit 15, and an output unit 16. The device 10 may also include other computing devices such as a CPU, a GPU (Graphics Processing Unit), or an APU (Accelerated Processing Unit), or a combination thereof, as computing devices.

[0018] Bus 103 can also be connected to external devices, for example. Examples of such external devices include external storage devices (external databases, etc.), electrocardiographs, printers, external input devices, external display devices, audio output devices such as speakers, external imaging devices such as cameras, and various sensors such as acceleration sensors, geomagnetic sensors, and direction sensors. The device 10 can be connected to an external network (the aforementioned communication network) by a communication device 107 connected to bus 103, for example, and can also be connected to other devices via the external network.

[0019] Memory 102 may be, for example, main memory. When the central processing unit 101 performs processing, memory 102 reads various operational programs, such as the program of this disclosure, stored in the storage device 104 (described later), and the central processing unit 101 receives data from memory 102 and executes the program. The main memory may be, for example, RAM (random access memory). Alternatively, memory 102 may be, for example, ROM (read-only memory).

[0020] The storage device 104 is also called an auxiliary storage device, for example, in relation to the main memory (primary memory). As described above, the storage device 104 stores an operating program including the program of this disclosure. The storage device 104 may be, for example, a combination of a recording medium and a drive for reading and writing to the recording medium. The recording medium is not particularly limited and may be internal or external, for example, an HD (hard disk), CD-ROM, CD-R, CD-RW, MO, DVD, flash memory, memory card, etc. The storage device 104 may be, for example, a hard disk drive (HDD) in which the recording medium and the drive are integrated, or a solid state drive (SSD). If the device 10 includes, for example, the storage device 104 functions as the storage unit. The storage device 104 may store, for example, a proxy existence model pool, which will be described later.

[0021] In this device 10, the memory 102 and storage device 104 can also store various types of information, such as log information, information obtained from an external database (not shown) or external devices, information generated by this device 10, and information used by this device 10 when executing processing. At least some of the information may be stored on an external server other than the memory 102 and storage device 104, or it may be stored in a distributed manner across multiple terminals using blockchain technology or the like.

[0022] The device 10 further includes, for example, an input device 105 and an output device 106. The input device 105 may include, for example, a pointing device such as a touch panel, trackpad, or mouse; a keyboard; imaging means such as a camera or scanner; a card reader such as an IC card reader or magnetic card reader; an audio input means such as a microphone; and so on. The output device 106 may include, for example, a display device such as an LED display or liquid crystal display; an audio output device such as a speaker; a printer; and so on. In this embodiment 1, the input device 105 and the output device 106 are configured separately, but the input device 105 and the output device 106 may be configured as an integrated unit, such as a touch panel display.

[0023] Next, an example of the proxy existence model management method of this embodiment will be described based on the flowchart in Figure 3. The proxy existence model management method of this embodiment can be implemented as follows, for example, using the proxy existence model management device 10 shown in Figures 1 and 2. Note that the proxy existence model management method of this embodiment is not limited to the use of the proxy existence model management device 10 shown in Figures 1 and 2.

[0024] The model pool management unit 11 manages a proxy existence model pool by linking at least one proxy existence model with the model feature information of the proxy existence model (S1, model pool management process). The proxy existence model is a model constructed to behave in imitation of a model subject for a large-scale learning model. For an explanation of the proxy existence model, refer to the description in Embodiment 3, which will be described later. The model feature information is information that indicates the characteristics of the proxy existence model. Examples of the model feature information include information about the model subject of the proxy existence model and information about the knowledge system handled by the proxy existence model. The model pool management unit 11 may, for example, manage the model feature information by linking the occupation of the person (subject) that each proxy existence model is modeled on and the tasks corresponding to the occupation with the proxy existence model. Alternatively, the model pool management unit 11 may, for example, manage the model feature information by linking the knowledge system handled by each proxy existence model. "Knowledge system" refers to information about the knowledge that the proxy existence model can use. The aforementioned knowledge system may be, for example, a database containing predetermined knowledge, or a trained model that has learned the predetermined knowledge. The trained model that has learned the predetermined knowledge may be, for example, the explicit knowledge model (large-scale language model with domain knowledge) of Embodiment 3 described later. The model feature information can be set appropriately based, for example, on information of the person (subject) that served as the model for the surrogate existence model. The model pool management unit 11 may manage the surrogate existence model pool by, for example, recording the surrogate existence model and the model feature information in association with each other in the storage unit of the device 10, or it may manage the surrogate existence model pool recorded on a recording medium outside the device 10.

[0025] The model pool management unit 11 may, for example, generate an embedding vector representing the knowledge system handled by each surrogate model as model feature information, and register the embedding vector as an index in the surrogate model pool. Specifically, the model pool management unit 11 may, for example, vectorize metadata or representative text describing the knowledge system for each unit text, and record the average vector or representative vector of these vectors as model feature information. This allows for high-speed retrieval of surrogate models based on similarity to vectors representing the required knowledge system.

[0026] The model selection unit 14 may, for example, further generate embedding vectors for the necessary knowledge systems for each task or divided task identified by the input analysis unit 13, and calculate the similarity (e.g., cosine similarity) between these vectors and the embedding vectors of the model feature information of each proxy existence model managed by the model pool management unit 11. The model selection unit 14 can then, for example, select a predetermined number of proxy existence models in descending order of similarity from among the proxy existence models whose similarity is above a predetermined threshold, thereby narrowing down the number of proxy existence models necessary to satisfy the necessary knowledge systems to the minimum number. In this way, by dynamically selecting the proxy existence models to be executed according to the content of the input information, it is possible to reduce the startup of unnecessary models and token processing compared to the case where only general-purpose large-scale language models are always used, thereby reducing both computational resources and response time.

[0027] The input information acquisition unit 12 acquires user input information (S2, input information acquisition step). The input information is not particularly limited as long as it is information that allows the proxy existence model to generate information via the device 10. The input information may be, for example, information that reveals the content of the user's consultation or the content of the mission that the user wants to solve. The input information may be, for example, text data, image data, audio data, or a combination of these. The input information acquisition unit 12 may, for example, record the acquired input information in the storage unit of the device 10.

[0028] The input analysis unit 13 analyzes the input information (S3, input analysis step). The input analysis unit 13 may analyze the input information and extract at least one task contained in the input information. For example, the input analysis unit 13 may analyze the input information, extract at least one task contained in the input information, and analyze the necessary knowledge system for each extracted task. Alternatively, the input analysis unit 13 may, for example, further divide the task into stepwise subdivided tasks and analyze the necessary knowledge system for each subdivided task. The necessary knowledge system is information that identifies the combination of knowledge that needs to be referenced in order to solve the task. The necessary knowledge system may, for example, be information that identifies a database from which predetermined knowledge has been collected, or information that identifies a trained model that has learned the predetermined knowledge. The trained model that has learned the predetermined knowledge may, for example, be the explicit knowledge model (large-scale language model with domain knowledge) of Embodiment 3 described later. The input analysis unit 13 may, for example, pass the input information to the LLM to extract the task or subdivided task contained in the input information, and for each extracted task or subdivided task, analyze the knowledge system to be used in information processing. Furthermore, the input analysis unit 13 may, for example, analyze the tasks included in the input information and the necessary knowledge system for each task based on the task identification information. The task identification information is, for example, information in which keywords related to the task and the necessary knowledge system for each task are linked and organized.

[0029] The input analysis unit 13 may, for example, preprocess the input information and convert each unit of text into an embedding vector. Specifically, the input analysis unit 13 may, for example, generate an embedding vector consisting of hundreds to thousands of dimensions of real-valued elements for each sentence or paragraph contained in the input information, and then analyze the generated sequence of embedding vectors using clustering or a classifier to identify the tasks contained in the input information and the necessary knowledge systems for each task. It should be noted that such vectorization and clustering / classification processes consist of numerous matrix operations, making it practically impossible for a human to perform them mentally or with paper and pencil, and therefore rely on automated processing by electronic computers such as processors and GPUs.

[0030] The model selection unit 14 selects at least one surrogate existence model that mimics the target person from the surrogate existence model pool based on the analysis results of the input information and the model feature information (S4, model selection step).

[0031] The model selection unit 14 can, for example, select at least one surrogate existence model that mimics a person belonging to the occupation corresponding to the task. If multiple tasks are extracted in step S3, the model selection unit 14 may select a surrogate existence model that mimics a person belonging to the occupation corresponding to each of the multiple tasks.

[0032] The model selection unit 14 may select a surrogate existence model for each task based on the necessary knowledge system. Alternatively, the model selection unit 14 may select a surrogate existence model for each partitioned task based on the necessary knowledge system.

[0033] The information generation unit 15 provides the selected proxy existence model with the input information to generate information (S5, information generation step). The information generation unit 15 can, for example, input the input information to each selected proxy existence model to cause each proxy existence model to generate information. The information output unit 15 may, for example, cause the proxy existence model to generate a proxy statement text of the target person based on the text information as the information, and output the proxy statement text. If the input information is a question, the information generation unit 15 may, for example, cause the proxy existence model to generate an answer to the question as the information, and output the answer. The information generation unit 15 may, for example, cause the proxy existence model to generate an answer to the question from another large-scale language model as the information, input the instruction information to the large-scale language model to generate an answer from the large-scale language model, and output the answer. In this case, the information generation unit 15 may, for example, input the response generated by the large-scale language model as further input information to the surrogate existence model, correct the response generated by the large-scale language model to an output that is more in line with the target person, and output the corrected response as information.

[0034] The information generation unit 15 may, for example, manage the order in which the selected multiple proxy existence models generate information based on the input information. The information generation unit 15 may, for example, aggregate the information generated by the selected multiple proxy existence models and correct the generated information so that the information output with respect to the input information is consistent. Specifically, when the information generation unit 15 has multiple proxy existence models generate information, it may generate a directed graph representing the execution order based on the model feature information and task dependencies of each proxy existence model. For example, the dependencies between divided tasks can be represented as nodes and edges, such as using the output of task "requirements definition" as the input of task "architecture design," and the result of that as the input of task "risk assessment," and the execution order of the proxy existence models can be determined based on a topological sort of the graph. This allows each proxy existence model to be executed with the necessary prerequisite information in place, compared to the case where each model is executed in parallel while ignoring dependencies, and the occurrence of unnecessary re-executions can be suppressed.

[0035] Furthermore, when the information generation unit 15 aggregates the information generated by the selected multiple proxy existence models, it may perform consistency evaluation processing to detect logical inconsistencies and duplications in the content of each generated piece of information. Specifically, it may convert each generated piece of information into a feature vector, calculate the similarity between them, cluster semantically similar information, and integrate it so that only representative information remains for information belonging to the same class. Also, if information with contradictory preconditions is detected between different classes, it is possible to generate output information that maintains consistency with the input information by excluding information with low confidence based on predetermined priorities and fitness scores.

[0036] The output unit 16 outputs the information generated by the proxy existence model (S6, output process). The output unit 16 may, for example, calculate the degree of suitability of the information generated by the proxy existence model to the input information and output the degree of suitability linked to the information. For example, when using the device 10 for system development, the degree of suitability to customer requirements included in the input information (e.g., cost-conscious, processing speed-conscious, design-conscious, etc.) may be calculated, and the output may be corrected if the degree of suitability is too low. In addition, the customer requirements may not be single but may include multiple conditions or priorities. In this case, the output unit 16 may, for example, calculate and judge the degree of suitability for each customer requirement and output the result, presenting it as a candidate to multiple users for selection.

[0037] The output unit 16 may, for example, calculate evaluation values ​​for each of the multiple requirement items included in the input information for the information generated by the proxy existence model, and calculate a numerical fitness by integrating them. Specifically, an evaluation vector can be defined with each requirement item extracted from the input information as a dimension, and the satisfaction level for each requirement can be determined as a numerical value in the range of 0 to 1 based on the dot product or distance with the evaluation item vector extracted from the generated information. The output unit 16 may then calculate the overall fitness level by a linear combination of the weight vector for each requirement set by the user and the satisfaction level, and if the overall fitness level falls below a predetermined threshold, it may output a control signal to the model selection unit 14 instructing it to select or re-select another proxy existence model. In this way, by not only displaying the evaluation results but also dynamically controlling the model configuration and generation process based on the numerical fitness level, it is possible to obtain a response that is more closely suited to the user requirements compared to relying on the one-time inference result of a general-purpose large-scale language model.

[0038] (Example of Use 1) The following describes examples of how the device 10 can be used. The following examples describe how the device 10 can be used in the IT field, but this disclosure is not limited to the following examples.

[0039] First, prior to using the device 10, a proxy existence model that mimics an IT expert is generated, the occupations of the individuals on whom each proxy existence model is based (e.g., project manager (PM), architect, business expert, legal checker, etc.) are organized, and tasks that can be handled by each occupation are linked to construct a proxy existence model pool.

[0040] Suppose a user of this device 10 (for example, a project leader) is launching a new development project and wants to proceed with the architecture based on the project's objectives and requirements gathered from the customer. In this case, the project leader first inputs documents containing the project's objectives and requirements (for example, a request for proposal, a requirements list, etc.) into this device 10 as input information. This device 10 analyzes the input information and extracts "Support for launching projects in the IT field" as a task. Then, based on the extracted task and the proxy existence model pool, this device 10 selects a proxy existence model (IT) that mimics an IT expert and, based on the input information, executes a dialogue between the proxy existence model (IT) and the user. This device 10 analyzes the user's input to the proxy existence model (IT) and extracts the necessary subdivided tasks (create a requirements list based on the input information, create an architecture from the created requirements list). Based on the extracted subdivided tasks and the proxy existence model pool, this device 10 selects a proxy existence model for each subdivided task. In this example, for the divided task "Create a requirements list based on input information," a proxy existence model (PM) that mimics a project manager is selected, and for the divided task "Create an architect from the created requirements list," a proxy existence model (architecture) that mimics an architecture is selected, and the information related to each task is generated by the respective proxy existence model. The device 10 then outputs the results in stages as they are executed (1. Requirements list, 2. Architecture diagram). The device 10 may also acquire user feedback on the output results (e.g., "This requirement is missing," "I want to change the layout of the diagram") to confirm whether the output (response) is in line with the user's input information and make corrections.

[0041] (Example of Use 2) The following describes an example of using the device 10 with another specific example. In the following example, an example of using the device 10 in the agricultural field will be described, but this disclosure is not limited to the following example. First, using Figure 4, an example of creating a proxy existence model pool for the knowledge system necessary for autumn-sown wheat will be described, but this disclosure is not limited to the following description. First, the tasks (divided tasks) necessary for autumn-sown wheat (task) are listed for each work period. Specifically, the divided tasks necessary to carry out the autumn-sown wheat task are, as shown in Figure 4, "compost spreading," "plowing," "soil crushing and leveling," "fertilization and sowing," "herbicide application," "pest and disease control," "snow melting promotion," "top dressing," "pest and disease control," "harvesting," "transportation," "drying and processing," and "residue disposal." Then, a surrogate existence model that has learned the necessary knowledge system for each segmented task (for example, for "fertilization and sowing," the knowledge systems would be "genetics and breeding," "crop cultivation," and "plant nutrition" for crops; "crop production science," "agricultural machinery," "soil science," "agricultural environmental engineering," "agricultural information engineering," and "plant pathology" for technology; and "genetics and breeding," "pesticides," and "soil and fertilizer science" for materials) is recorded as a surrogate existence model corresponding to each segmented task, and a surrogate existence model pool is constructed.

[0042] A user of this device 10 (for example, an agricultural worker) inputs a question ("I want to increase my wheat yield...") to the device 10 for purposes such as early acquisition of agricultural knowledge, improvement of productivity and profitability, and mastery of the latest technologies related to smart agriculture and reduction of environmental impact. The device 10 analyzes the input question and extracts "answers to questions in the agricultural field" as a task. Then, based on the extracted task and the surrogate existence model pool, the device 10 selects a surrogate existence model (agriculture) that mimics an expert in the agricultural field and generates and outputs an answer to the question ("To increase wheat yield, extending the ripening period can be expected to increase yield by 2-3% per day. Therefore, it is necessary to grasp the growth status of wheat in each plot as accurately as possible and determine the optimal harvest time for each plot."). If the user asks a further question ("How can I understand the growth status of wheat in each plot?") in response to the generated answer, the device 10 causes the surrogate existence model (agriculture) to generate and output a further answer ("We use high-throughput phenotyping technology using drones."). Then, if the user inputs a further question ("Please tell me how to perform high-throughput phenotyping."), the device 10 extracts a more detailed subtask, "Answers to questions about high-throughput phenotyping technology," from the "Answers to questions in the agricultural field" task. Based on the extracted subtask and the surrogate existence model pool, the device selects a surrogate existence model (high-throughput phenotyping) that mimics an expert in high-throughput phenotyping technology, and generates and outputs an answer to the question ("We will explain how to perform high-throughput phenotyping for wheat. We can customize the analysis pipeline and assist with the analysis as needed.").

[0043] Thus, according to the present apparatus 10, it is possible to select an appropriate agent presence model for generating an answer to the question input by the user and output a highly accurate answer. At least a part of the agent presence models managed by the model pool management unit 11 may include the personality model and the formal knowledge model described later in Embodiment 3. In this case, the model selection unit 14 selects one or both of the personality model and the formal knowledge model according to the task or the necessary knowledge system specified by the input analysis unit 13, and the information generation unit 15 cooperatively operates the selected personality model and the formal knowledge model, so that a response based on the necessary knowledge system can be generated while matching the personality of the target person. As a result, compared with a configuration in which only the prompt is changed for a single general-purpose large-scale language model to switch tasks, both personality hallucination and knowledge hallucination can be reduced, and the consistency and reliability of the generated information can be further improved.

[0044] [Embodiment 2] The agent presence model management program of the present embodiment is a program for causing a computer to execute each step of the above-described agent presence model management method. Specifically, the agent presence model management program of the present embodiment is a program for causing a computer to execute a model pool management procedure, an input information acquisition procedure, an input analysis procedure, a model selection procedure, an information generation procedure, and an output procedure.

[0045] The model pool management procedure manages at least one agent presence model and the model feature information of the agent presence model as an agent presence model pool. The agent presence model is a model constructed to behave like a model target person with respect to a large-scale learning model. The input information acquisition procedure acquires the input information of the user. The input analysis procedure analyzes the input information. The model selection procedure selects at least one agent presence model that imitates the target person from the agent presence model pool based on the analysis result of the input information and the model feature information. The information generation procedure gives the input information to the selected agent presence model to generate information. The output procedure outputs the information generated by the agent presence model.

[0046] Also, the proxy presence model management program of this embodiment can also be said to be a program that causes a computer to function as a model pool management procedure, an input information acquisition procedure, an input analysis procedure, a model selection procedure, an information generation procedure, and an output procedure.

[0047] The proxy presence model management program of this embodiment can incorporate the descriptions in the proxy presence model management device and the proxy presence model management method of the present disclosure. Each of the above procedures can be read as "processing" instead of "procedure", for example. Also, the program of this embodiment may be recorded on a computer-readable recording medium, for example. The recording medium is not particularly limited, and examples include random access memory (RAM), read-only memory (ROM), hard disk (HD), flash memory (e.g., SSD (Solid State Drive), USB flash memory, SD / SDHC card, etc.), optical disk (e.g., CD-R / CD-RW, DVD-R / DVD-RW, BD-R / BD-RE, etc.), magneto-optical disk (MO), floppy (registered trademark) disk (FD), etc. Further, the proxy presence model management program of this embodiment (also referred to as a programming product or a proxy presence model management program product, for example) may be in a form distributed from an external computer, for example. The "distribution" may be, for example, distribution via a communication network or distribution via a device connected by wire. The proxy presence model management program of this embodiment may be installed and executed on the distributed device, or may be executed without being installed.

[0048] [Embodiment 3] The proxy presence model manufacturing apparatus of this embodiment will be described with reference to FIG. 8. FIG. 5 is a block diagram showing an example configuration of the proxy presence model manufacturing apparatus 20 of this embodiment. As shown in FIG. 5, the proxy presence model manufacturing apparatus 20 (hereinafter also referred to as "this apparatus 20") includes a knowledge information acquisition unit 21, a construction information extraction unit 22, and a model construction unit 23. Also, although not shown, this apparatus 20 may include, for example, an input unit, an output unit, a display unit, and / or a storage unit.

[0049] The device 20 may be, for example, a single device including the aforementioned parts, or it may be a device in which the aforementioned parts can be connected via a communication network. Furthermore, the device 20 can be connected to an external device described later via a communication network. The communication network is not particularly limited and a known network can be used, for example, it may be wired or wireless. Examples of communication networks include the Internet, WWW (World Wide Web), telephone lines, LAN (Local Area Network), SAN (Storage Area Network), DTN (Delay Tolerant Networking), LPWA (Low Power Wide Area), L5G (Local 5G), etc. Examples of wireless communication include Wi-Fi®, Bluetooth®, Local 5G, LPWA, etc. The aforementioned wireless communication may be in the form of direct communication between devices (Ad Hoc communication), infrastructure communication, or indirect communication via an access point. The device 20 may, for example, be incorporated into a server as a system. Alternatively, the device 20 may be, for example, a personal computer (PC, e.g., desktop or notebook), smartphone, or tablet terminal on which the program disclosed herein is installed. Furthermore, the device 20 may be in the form of cloud computing or edge computing, for example, in which at least one of the aforementioned parts is on a server and the other parts are on a terminal.

[0050] Figure 6 illustrates a block diagram of the hardware configuration of the device 20. The device 20 includes, for example, a central processing unit 201, memory 202, bus 203, storage device 204, input device 205, output device 206, communication device (communication unit) 207, etc. Each part of the device 20 is interconnected via the bus 203 through its respective interface (I / F).

[0051] The central processing unit 201 operates in coordination with other components via controllers (system controller, I / O controller, etc.) and is responsible for the overall control of the device 10. In the device 20, the central processing unit 201 executes, for example, the program of this disclosure and other programs, and also reads and writes various types of information. Specifically, for example, the central processing unit 201 functions as a knowledge information acquisition unit 21, a construction information extraction unit 22, and a model construction unit 23. The device 20 may also include other computing devices such as a CPU, GPU (Graphics Processing Unit), APU (Accelerated Processing Unit), or a combination thereof as computing devices.

[0052] Bus 203 can also be connected to external devices, for example. Examples of such external devices include external storage devices (external databases, etc.), electrocardiographs, printers, external input devices, external display devices, audio output devices such as speakers, external imaging devices such as cameras, and various sensors such as acceleration sensors, geomagnetic sensors, and direction sensors. The device 20 can be connected to an external network (the aforementioned communication network) by a communication device 207 connected to bus 203, for example, and can also be connected to other devices via the external network.

[0053] Memory 202 may be, for example, main memory. When the central processing unit 201 performs processing, memory 202 reads various operational programs, such as the program of this disclosure, stored in the storage device 204 (described later), and the central processing unit 201 receives data from memory 202 and executes the program. The main memory may be, for example, RAM (random access memory). Alternatively, memory 202 may be, for example, ROM (read-only memory).

[0054] The storage device 204 is also called an auxiliary storage device, for example, in relation to the main memory (primary memory). As described above, the storage device 204 stores an operating program including the program of this disclosure. The storage device 204 may be, for example, a combination of a recording medium and a drive for reading and writing to the recording medium. The recording medium is not particularly limited and may be internal or external, for example, an HD (hard disk), CD-ROM, CD-R, CD-RW, MO, DVD, flash memory, memory card, etc. The storage device 104 may be, for example, a hard disk drive (HDD) in which the recording medium and the drive are integrated, or a solid state drive (SSD). If the device 20 includes, for example, the storage device 204 functions as the storage unit.

[0055] In this device 20, the memory 202 and storage device 204 can also store various types of information, such as log information, information obtained from an external database (not shown) or external devices, information generated by this device 20, and information used by this device 20 when executing processing. At least some of the information may be stored, for example, on an external server other than the memory 202 and storage device 204, or distributed and stored across multiple terminals using blockchain technology or the like.

[0056] The device 20 further includes, for example, an input device 205 and an output device 206. The input device 205 may include, for example, a pointing device such as a touch panel, trackpad, or mouse; a keyboard; imaging means such as a camera or scanner; a card reader such as an IC card reader or magnetic card reader; an audio input means such as a microphone; and so on. The output device 206 may include, for example, a display device such as an LED display or liquid crystal display; an audio output device such as a speaker; a printer; and so on. In this embodiment 3, the input device 205 and the output device 206 are configured separately, but the input device 205 and the output device 206 may be configured as an integrated unit, such as a touch panel display.

[0057] Next, an example of the proxy existence model manufacturing method of this embodiment will be described based on the flowchart in Figure 7. The proxy existence model manufacturing method of this embodiment can be carried out as follows, for example, using the proxy existence model manufacturing apparatus 20 shown in Figures 5 and 6. Note that the proxy existence model manufacturing method of this embodiment is not limited to the use of the proxy existence model manufacturing apparatus 20 shown in Figures 5 and 6.

[0058] First, the knowledge information acquisition unit 21 acquires the subject's knowledge information (S21, knowledge information acquisition step). The format of the knowledge information is not particularly limited; for example, it may be text information, image information, audio information, or a combination thereof. The knowledge information is, for example, information linked to predetermined information and subject identification information that identifies the creator of the information (the subject). The predetermined information is, for example, information that includes the subject's personality information and explicit knowledge information. The personality information is, for example, information that represents the subject's thoughts from the knowledge information. The personality information is also called, for example, a partial stance. The explicit knowledge information is, for example, the part of the knowledge information excluding the personality information, and includes objective knowledge. If the knowledge information is a book, paper, etc., the explicit knowledge information may, for example, be information such as technical terms and experimental results, but is not limited to these. The subject identification information may, for example, be a name, address, telephone number, email address, identification number (for example, My Number (individual number), etc.). Specific examples of the aforementioned knowledge information include, but are not limited to, books and papers written by the subject, video data of lectures given by the subject, audio data of lectures given by the subject, and image data created by the subject. The knowledge information acquisition unit 21 may, for example, acquire knowledge information recorded in the storage unit of the device 20, or it may acquire the aforementioned knowledge information from outside the device 20 via the input device 205. The knowledge information acquisition unit 21 may, for example, acquire one type of knowledge information of the subject, or it may acquire two or more types. The knowledge information acquisition unit 21 may, for example, record the acquired knowledge information in the storage unit of the device 20.

[0059] The information extraction unit 22 for construction extracts personality information of the subject from the knowledge information (S22, information extraction step for construction). The information extraction unit 22 for construction may also, for example, further extract explicit knowledge information from the knowledge information of the subject. The information extraction unit 22 for construction may, for example, use known natural language processing techniques to extract at least one of the personality information and explicit knowledge information of the subject, or use a large-scale language model to extract at least one of the personality information and explicit knowledge information of the subject. If the knowledge information is textual information such as a book or a paper, the information extraction unit 22 for construction can, for example, extract the personality information or explicit knowledge information based on the end of a sentence in a document, the chapter structure of a book, etc. The information extraction unit 22 for construction may, for example, record at least one of the extracted personality information and explicit knowledge information in the storage unit of the device 20. In this case, the information extraction unit 22 for construction can, for example, record personality information and explicit knowledge information, the knowledge information from which they were extracted, and the creator identification information of the knowledge information in a linked manner.

[0060] The information extraction unit 22 for construction can, for example, analyze the knowledge information and extract sentences whose sentence ends with a word that expresses the author's thoughts as the personality information. Examples of words that express the author's thoughts include, but are not limited to, words such as "I want to," "I think," "I believe," and "I want." The information extraction unit 22 for construction can also analyze the knowledge information and extract sentences whose sentence ends with a word that indicates explicit knowledge as the explicit knowledge information. Examples of words that indicate explicit knowledge include, but are not limited to, words such as "It is," "It was," and "As a result."

[0061] The information extraction unit 22 for construction may, for example, analyze the knowledge information and extract sentences contained in chapters that describe the author's ideas as personality information. Examples of chapters that describe the author's ideas include the "preface," "introduction," and "afterword."

[0062] The information acquisition unit 22 for construction may, for example, use AI to extract personality information and explicit knowledge information from knowledge information. In this disclosure, AI may refer to, for example, a large-scale learning model called a "foundation model." The foundation model is a machine learning model pre-trained on predetermined big data and is not limited to a large-scale language model (LLM) that has learned natural language, but may also include a large-scale model for speech, a large-scale model for images, and a multimodal model (such as a visual language model) that handles language, images, speech, and video across the board. Furthermore, a configuration may be adopted in which a small-scale language model (SLM) is placed on the terminal side and cooperates with the large-scale model on the cloud side, a configuration that includes search extension generation (RAG) using an external knowledge source, tool execution / function call, agent-oriented control logic, etc. The providers of large-scale language models are not particularly limited. Examples include, but are not limited to, various LLM / multimodal models provided by companies such as OpenAI, Anthropique, Alphabet (Google), META, Microsoft, Cohere, Mistral, xAI, NEC Corporation, and NTT.

[0063] When the information acquisition unit 22 for construction uses AI to extract personality information and explicit knowledge information from knowledge information, the information acquisition unit 22 can cause the AI ​​to extract the user's personality information and explicit knowledge information by inputting extraction instruction information, which instructs the AI ​​to extract personality information and explicit knowledge information, along with the knowledge information. The extraction instruction information may be recorded in the storage unit of the device 10, stored externally, or input by the user each time. Specific examples of the extraction instruction information include, for example, "classifying information into the following two categories: sentences and paragraphs that show the author's thoughts and opinions, and sentences and paragraphs that show knowledge, such as scientific verification results and historical facts." This disclosure is not limited to the above examples.

[0064] Furthermore, the AI ​​may be, for example, a model finely tuned to extract personality information and explicit knowledge information from knowledge information. In this case, for example, fine tuning for extracting personality information and explicit knowledge information from knowledge information can be performed by training the AI ​​with knowledge information and a set of personality information and explicit knowledge information previously extracted from the knowledge information. In this case, the AI ​​may be, for example, a multilayer neural network model having a large number of parameters (e.g., 100,000 or more, and in some cases, several million or more). Such fine tuning processing requires the rapid and repeated execution of a huge amount of numerical calculations, making it practically impossible for a human to perform it in their head or with paper and pencil, and thus relies on automated processing by electronic computers such as processors and GPUs. During the fine tuning, in order to suppress overfitting of the training data, known regularization methods such as L2 regularization, dropout, and early stopping may be combined and applied. This makes it possible to obtain a model that can extract personality information and explicit knowledge information with high generalization performance even from unknown knowledge information, and improves the accuracy and robustness of information extraction compared to simple threshold judgment or rule-based processing.

[0065] Furthermore, the information extraction unit 22 for construction may extract the personality information of the subject by, for example, providing a large-scale language model with the subject's knowledge information and instruction information that instructs the model to extract the subject's personality information from the knowledge information based on the subject's knowledge information, thereby causing the model to extract the personality information from the knowledge information. The large-scale language model is not particularly limited and includes, but is not limited to, OpenAI®'s GPT-3®, GPT-4®, Alphabet Inc. (Google®)'s BERT, LaMDA, PaLM2, META®'s LlaMA, NEC Corporation's LLM, NTT®'s LLM, etc. The instruction information that instructs the model to extract the subject's personality information from the knowledge information based on the subject's knowledge information is not particularly limited as long as it is a document that instructs the model to divide the knowledge information into parts that contain thoughts and parts that contain knowledge. Specific examples of instruction information that instructs the extraction of personality information of the subject from the knowledge information of the subject include, but are not limited to, documents such as, "Classify the information into the following two categories: - Sentences and paragraphs that show the author's thoughts and opinions, such as the author's ideas; - Sentences and paragraphs that show knowledge, such as scientific verification results and historical facts."

[0066] The information extraction unit 22 for construction may, for example, preprocess the knowledge information and extract personality information and explicit knowledge information. Specifically, for example, the information extraction unit 22 for construction may convert the knowledge information into a vectorized embedding vector sequence for each unit text, and then classify or group the knowledge information into personality information and explicit knowledge information based on the embedding vector sequence, thereby extracting personality information and explicit knowledge information based on the knowledge information. The embedding vectors may, for example, contain hundreds to thousands of real-valued elements for each unit text. For this reason, for example, a large number of embedding vectors are generated for the entire knowledge information, and the classification or grouping process consists of a large number of numerical operations, mainly matrix operations, which is a process that is practically impossible for a human to perform in their head or with paper and pencil.

[0067] The model building unit 23 constructs a surrogate existence model that mimics the target person based on the personality information (S23, model building step). The model building unit 23 can construct the surrogate existence model by, for example, providing a large-scale learning model with the personality information and instruction information that instructs the large-scale learning model to construct a surrogate existence model that mimics the target person based on the personality information. The large-scale learning model is, for example, a machine learning model that has been trained using predetermined big data. The large-scale learning model may be, for example, a model that has been trained on big data of natural language (large-scale language model), a model that has been trained on big data of speech (large-scale speech model), or a model that has been trained on big data of images (large-scale image model). The model building unit 23 can construct the surrogate existence model by, for example, providing a large-scale language model, as the large-scale learning model, with the personality information and instruction information that instructs the large-scale learning model to construct a surrogate existence model that mimics the target person based on the personality information. The instruction information is, for example, instruction information (prompt) for generating the behavior of a person handling knowledge. Examples of the aforementioned instructional information include, but are not limited to, documents such as, "When generating text, follow the rules below: - Describe examples that reflect personal information. - Do not reflect personal information in terms of knowledge. - Structure the text such that, for example, you state a fact that is correct as knowledge, and then express an opinion on that fact as personal information."

[0068] Furthermore, the model building unit 23 may generate multiple sets of training data showing the correspondence between the subject's input information and responses based on the extracted personality information. The model building unit 23 may input the training data in mini-batch units into a large-scale learning model, calculate a loss function based on the error between the model's output for the training data and the response, and train a personality model capable of outputting behavior that mimics the subject by iteratively updating a large number of model parameters using gradient descent or a modified algorithm thereof.

[0069] Furthermore, the model building unit 23 may, for example, construct an explicit knowledge model by further training a large-scale language model with the explicit knowledge information. The model building unit 23 can construct the explicit knowledge model by, for example, providing the explicit knowledge information to the large-scale language model and fine-tuning it. The explicit knowledge model is also called, for example, a large-scale language model with domain knowledge.

[0070] The aforementioned surrogate existence model is, for example, a model that has learned the personality information of the subject from the information that constitutes the knowledge information. Therefore, the surrogate existence model has learned, for example, the subject's way of thinking and understanding when handling knowledge, values ​​such as thoughts and beliefs, and how they interact with others (personality). For this reason, the surrogate existence model manufacturing method of this disclosure makes it possible to easily manufacture a model that reflects the thoughts of the information creator. Furthermore, the surrogate existence model manufactured by the surrogate existence model manufacturing method of this disclosure is capable of outputting products that are characteristic of the subject. In other words, according to this disclosure, for example, it becomes possible to construct a model that suppresses personal hallucination. Personal hallucination refers to, for example, a hallucination (illusion) of how to handle knowledge and output policies that the creator of the knowledge information would not say. Therefore, according to the surrogate existence model of this disclosure, for example, it becomes possible to more accurately extract and output the knowledge that knowledge information (e.g., a book) explicitly or implicitly contains. The output that is characteristic of the target person is not particularly limited and may be, for example, text output, voice output, or instructions to a designated machine.

[0071] The model building unit 23 may, for example, use extracted personality information to train a personality model capable of outputting behavior that mimics the subject by iteratively updating the parameters of a large-scale learning model having many parameters based on gradient descent to reproduce the correspondence between the subject's past input information and responses. Similarly, the model building unit 23 may, for example, use extracted explicit knowledge information to train an explicit knowledge model that outputs explicit knowledge information contained in the knowledge information. In this case, the model building unit 23 can, for example, generate proxy behavior information consistent with the subject's personality and output the proxy behavior information by coordinating the trained personality model and the explicit knowledge model. By configuring the personality model and the explicit knowledge model separately in this way, and coordinating the output of explicit knowledge based on the knowledge information with the output based on the personality information, it is possible to reduce the inclusion of personal expressions not included in the knowledge information and improve the consistency and reliability of the generated responses, compared to, for example, simply providing prompts to a general-purpose large-scale language model.

[0072] According to this disclosure, it is possible to create a model that reflects the thoughts of the information creator. Therefore, according to this disclosure, for example, it is possible to construct a surrogate model for individuals with limited human resources (e.g., busy researchers, supervisors, managers, teachers, etc.), thereby reducing the burden on those individuals.

[0073] Although the present disclosure has been described above with reference to embodiments, the present disclosure is not limited to the embodiments described above. Various modifications to the structure and details of the present disclosure are possible, which can be understood by those skilled in the art within the scope of the present disclosure.

[0074] This application claims priority based on Japanese Patent Application No. 2024-229567, filed on 26 December 2024, and incorporates all of its disclosures herein.

[0075] <Note> Some or all of the above embodiments may be described as follows, but are not limited to the following. (Note 1) A proxy existence model management device comprising a model pool management unit, an input information acquisition unit, an input analysis unit, a model selection unit, an information generation unit, and an output unit, wherein the model pool management unit manages a proxy existence model pool by linking at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition unit acquires user input information, the input analysis unit analyzes the input information, the model selection unit selects at least one proxy existence model that imitates a target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation unit provides the selected proxy existence model with the input information to generate information, and the output unit outputs the information generated by the proxy existence model. (Note 2) The proxy existence model management device according to Note 1, wherein the model pool management unit manages the occupation of the model subject in each proxy existence model and the tasks corresponding to the occupation as model feature information, linking them with the proxy existence model; the input analysis unit analyzes the input information and extracts at least one task included in the input information; and the model selection unit selects at least one proxy existence model that mimics a person belonging to the corresponding occupation based on the task. (Note 3) The proxy existence model management device according to Note 2, wherein the input analysis unit extracts a plurality of tasks included in the input information; and the model selection unit selects a proxy existence model that mimics a person belonging to the occupation corresponding to each of the plurality of tasks. (Note 4) The proxy existence model management device as described in Note 1, wherein the model pool management unit manages the knowledge systems handled by each proxy existence model as model feature information, the input analysis unit analyzes the input information, extracts at least one task contained in the input information, analyzes the necessary knowledge system for each extracted task, and the model selection unit selects a proxy existence model for each task based on the necessary knowledge system.(Note 5) The proxy existence model management device according to Note 4, wherein the input analysis unit further divides the task into stepwise divided tasks, analyzes the necessary knowledge system for each divided task, and the model selection unit selects a proxy existence model for each divided task based on the necessary knowledge system. (Note 6) The proxy existence model management device according to any one of Notes 1 to 5, wherein the information generation unit manages the order in which the selected proxy existence models generate information based on the input information. (Note 7) The proxy existence model management device according to any one of Notes 1 to 6, wherein the information generation unit collects the information generated by the selected proxy existence models and corrects the generated information so that the information output with respect to the input information is consistent. (Note 8) The proxy existence model management device according to any one of Notes 1 to 7, wherein the output unit calculates the degree of fitness for the information generated by the proxy existence models with respect to the input information and outputs the degree of fitness linked to the information. (Note 9) A method for managing a proxy model, comprising a model pool management step, an input information acquisition step, an input analysis step, a model selection step, an information generation step, and an output step, wherein the model pool management step manages a proxy model pool by linking at least one proxy model with the model feature information of the proxy model, the proxy model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition step acquires user input information, the input analysis step analyzes the input information, the model selection step selects at least one proxy model that imitates a target from the proxy model pool based on the analysis results of the input information and the model feature information, the information generation step provides the selected proxy model with the input information to generate information, and the output step outputs the information generated by the proxy model, each step being performed by a computer.(Note 10) The method for managing surrogate existence models according to Note 9, wherein the model pool management step manages the occupation of the model subject in each surrogate existence model and the tasks corresponding to the occupation as model feature information, linked to the surrogate existence model; the input analysis step analyzes the input information and extracts at least one task included in the input information; and the model selection step selects at least one surrogate existence model that imitates a person belonging to the corresponding occupation based on the task. (Note 11) The method for managing surrogate existence models according to Note 10, wherein the input analysis step extracts a plurality of tasks included in the input information; and the model selection step selects a surrogate existence model that imitates a person belonging to the occupation corresponding to each of the plurality of tasks. (Note 12) The proxy existence model management method according to Note 9, wherein the model pool management step manages the knowledge system handled by each proxy existence model as model feature information, the input analysis step analyzes the input information, extracts at least one task included in the input information, analyzes the necessary knowledge system for each extracted task, and the model selection step selects a proxy existence model for each task based on the necessary knowledge system. (Note 13) The proxy existence model management method according to Note 12, wherein the input analysis step further divides the task into stepwise divided tasks, analyzes the necessary knowledge system for each divided task, and the model selection step selects a proxy existence model for each divided task based on the necessary knowledge system. (Note 14) The proxy existence model management method according to any one of Notes 9 to 13, wherein the information generation step manages the order in which the selected proxy existence models generate information based on the input information. (Note 15) The information generation step is the proxy existence model management method according to any one of Notes 9 to 14, wherein the information generation step collects the information generated by a plurality of selected proxy existence models and corrects the generated information so that the information output with respect to the input information is consistent. (Note 16) The output step is the proxy existence model management method according to any one of Notes 9 to 15, wherein the information generated by the proxy existence model calculates the degree of fitness with respect to the input information and outputs the degree of fitness linked to the information.(Note 17) A proxy existence model management program for causing a computer to execute each of the following procedures: a model pool management procedure, an input information acquisition procedure, an input analysis procedure, a model selection procedure, an information generation procedure, and an output procedure, wherein the model pool management procedure manages a proxy existence model pool by linking at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in a manner that mimics the model target for a large-scale learning model, the input information acquisition procedure acquires user input information, the input analysis procedure analyzes the input information, the model selection procedure selects at least one proxy existence model that mimics the target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation procedure provides the selected proxy existence model with the input information to generate information, and the output procedure outputs the information generated by the proxy existence model. (Note 18) The surrogate existence model management program as described in Note 17, wherein the model pool management procedure manages the occupation of the model subject in each surrogate existence model and the tasks corresponding to the occupation as model feature information, linked to the surrogate existence model, the input analysis procedure analyzes the input information and extracts at least one task included in the input information, and the model selection procedure selects at least one surrogate existence model that mimics a person belonging to the corresponding occupation based on the task. (Note 19) The surrogate existence model management program as described in Note 18, wherein the input analysis procedure extracts a plurality of tasks included in the input information, and the model selection procedure selects a surrogate existence model that mimics a person belonging to the occupation corresponding to each of the plurality of tasks. (Note 20) The model pool management procedure manages the knowledge systems handled by each surrogate existence model as model feature information, the input analysis procedure analyzes the input information, extracts at least one task contained in the input information, analyzes the necessary knowledge system for each extracted task, and the model selection procedure selects a surrogate existence model for each task based on the necessary knowledge system, as described in Note 17, for the surrogate existence model management program.(Note 21) The input analysis procedure further divides the task into stepwise divided tasks, analyzes the necessary knowledge system for each divided task, and the model selection procedure selects a surrogate existence model for each divided task based on the necessary knowledge system, as described in Note 20, as a surrogate existence model management program. (Note 22) The information generation procedure manages the order in which the selected surrogate existence models generate information based on the input information, as described in any of Notes 17 to 21, as a surrogate existence model management program. (Note 23) The information generation procedure aggregates the information generated by the selected surrogate existence models and corrects the generated information so that the information output with respect to the input information is consistent, as described in any of Notes 17 to 22, as a surrogate existence model management program. (Note 24) The output procedure calculates the fitness of the information generated by the surrogate existence models with respect to the input information and outputs the fitness in association with the information, as described in any of Notes 17 to 23, as a surrogate existence model management program. (Note 25) A computer-readable recording medium that records a proxy existence model management program for causing a computer to execute each of the following procedures: a model pool management procedure, an input information acquisition procedure, an input analysis procedure, a model selection procedure, an information generation procedure, and an output procedure, wherein the model pool management procedure manages a proxy existence model pool by linking at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition procedure acquires user input information, the input analysis procedure analyzes the input information, the model selection procedure selects at least one proxy existence model that imitates a target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation procedure provides the selected proxy existence model with the input information to generate information, and the output procedure outputs the information generated by the proxy existence model.(Note 26) The recording medium according to Note 25, wherein the model pool management procedure manages the occupation of the model subject in each surrogate existence model and the tasks corresponding to the occupation as model feature information, linked to the surrogate existence model, the input analysis procedure analyzes the input information and extracts at least one task included in the input information, and the model selection procedure selects at least one surrogate existence model that imitates a person belonging to the corresponding occupation based on the task. (Note 27) The recording medium according to Note 26, wherein the input analysis procedure extracts a plurality of tasks included in the input information, and the model selection procedure selects a surrogate existence model that imitates a person belonging to the occupation corresponding to each of the plurality of tasks. (Note 28) The recording medium described in Note 25, wherein the model pool management procedure manages the knowledge systems handled by each surrogate existence model as model feature information, the input analysis procedure analyzes the input information, extracts at least one task contained in the input information, analyzes the necessary knowledge system for each extracted task, and the model selection procedure selects a surrogate existence model for each task based on the necessary knowledge system. (Note 29) The recording medium described in Note 28, wherein the input analysis procedure further divides the task into stepwise divided tasks, analyzes the necessary knowledge system for each divided task, and the model selection procedure selects a surrogate existence model for each divided task based on the necessary knowledge system. (Note 30) The recording medium described in any of Notes 25 to 29, wherein the information generation procedure manages the order in which the selected surrogate existence models generate information based on the input information. (Note 31) The recording medium according to any one of Notes 25 to 30, wherein the information generation procedure aggregates the information generated by a selection of the proxy existence models and corrects the generated information so that the information output is consistent with respect to the input information. (Note 32) The recording medium according to any one of Notes 25 to 31, wherein the output procedure calculates the fitness of the information generated by the proxy existence model with respect to the input information and outputs the fitness in association with the information.

[0076] According to this disclosure, users can interact with a surrogate existence model that mimics a designated person. Unlike typical large-scale language models, the surrogate existence model learns the personality information of the individual person, enabling it to act as a surrogate for that person. As a result, according to this disclosure, by utilizing a surrogate existence model that acts as a proxy for personnel in personal work environments, it is possible to address the shortage of personnel in those personal work environments. Therefore, this disclosure is useful, for example, in a wide range of industries where personal work environments exist.

[0077] 10 Proxy existence model management device 11 Model pool management unit 12 Input information acquisition unit 13 Input analysis unit 14 Model selection unit 15 Information generation unit 16 Output unit 101 Central processing unit 102 Memory 103 Bus 104 Storage device 105 Input device 106 Output device 107 Communication device 20 Proxy existence model manufacturing device 21 Knowledge information acquisition unit 22 Construction information extraction unit 23 Model construction unit 201 Central processing unit 202 Memory 203 Bus 204 Storage device 205 Input device 206 Output device 207 Communication device

Claims

1. A proxy existence model management device comprising a model pool management unit, an input information acquisition unit, an input analysis unit, a model selection unit, an information generation unit, and an output unit, wherein the model pool management unit manages a proxy existence model pool by linking at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition unit acquires user input information, the input analysis unit analyzes the input information, the model selection unit selects at least one proxy existence model that imitates a target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation unit provides the selected proxy existence model with the input information to generate information, and the output unit outputs the information generated by the proxy existence model.

2. The surrogate existence model management device according to claim 1, wherein the model pool management unit manages the occupation of the model subject in each surrogate existence model and the tasks corresponding to the occupation as model feature information, linking them with the surrogate existence model; the input analysis unit analyzes the input information and extracts at least one task included in the input information; and the model selection unit selects at least one surrogate existence model that imitates a person belonging to the corresponding occupation based on the task.

3. The proxy existence model management device according to claim 2, wherein the input analysis unit extracts a plurality of tasks contained in the input information, and the model selection unit selects a proxy existence model that mimics a person belonging to the occupation corresponding to each of the plurality of tasks.

4. The proxy existence model management device according to claim 1, wherein the model pool management unit manages the knowledge system handled by each proxy existence model as model feature information, the input analysis unit analyzes the input information, extracts at least one task included in the input information, analyzes the necessary knowledge system for each extracted task, and the model selection unit selects a proxy existence model for each task based on the necessary knowledge system.

5. The proxy existence model management device according to claim 4, wherein the input analysis unit further divides the task into stepwise divided tasks, analyzes the necessary knowledge system for each divided task, and the model selection unit selects a proxy existence model for each divided task based on the necessary knowledge system.

6. The proxy existence model management device according to any one of claims 1 to 5, wherein the information generation unit manages the order in which the selected proxy existence models generate information based on the input information.

7. The proxy existence model management device according to any one of claims 1 to 6, wherein the information generation unit collects information generated by a plurality of selected proxy existence models and corrects the generated information so that the information output with respect to the input information is consistent.

8. The proxy existence model management device according to any one of claims 1 to 7, wherein the output unit calculates the degree of fitness for the information generated by the proxy existence model with respect to the input information and outputs the degree of fitness in association with the information.

9. A method for managing a proxy model, comprising a model pool management step, an input information acquisition step, an input analysis step, a model selection step, an information generation step, and an output step, wherein the model pool management step manages a proxy model pool by associating at least one proxy model with the model feature information of the proxy model, the proxy model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition step acquires user input information, the input analysis step analyzes the input information, the model selection step selects at least one proxy model that imitates a target from the proxy model pool based on the analysis results of the input information and the model feature information, the information generation step provides the selected proxy model with the input information to generate information, and the output step outputs the information generated by the proxy model, each step being executed by a computer.

10. The method for managing surrogate existence models according to claim 9, wherein the model pool management step manages the occupation of the model subject in each surrogate existence model and the tasks corresponding to the occupation as model feature information, linking them with the surrogate existence model; the input analysis step analyzes the input information and extracts at least one task included in the input information; and the model selection step selects at least one surrogate existence model that imitates a person belonging to the corresponding occupation based on the task.

11. The method for managing a surrogate existence model according to claim 10, wherein the input analysis step extracts a plurality of tasks contained in the input information, and the model selection step selects a surrogate existence model that mimics a person belonging to the occupation corresponding to each of the plurality of tasks.

12. The method for managing surrogate existence models according to claim 9, wherein the model pool management step manages the knowledge systems handled by each surrogate existence model as model feature information, the input analysis step analyzes the input information, extracts at least one task included in the input information, analyzes the necessary knowledge system for each extracted task, and the model selection step selects a surrogate existence model for each task based on the necessary knowledge system.

13. The method for managing a surrogate existence model according to claim 12, wherein the input analysis step further divides the task into stepwise divided tasks, analyzes the necessary knowledge system for each divided task, and the model selection step selects a surrogate existence model for each divided task based on the necessary knowledge system.

14. The method for managing proxy existence models according to any one of claims 9 to 13, wherein the information generation step manages the order in which the selected proxy existence models generate information based on the input information.

15. The proxy existence model management method according to any one of claims 9 to 14, wherein the information generation step involves aggregating information generated by a selected plurality of proxy existence models and correcting the generated information so that the information output with respect to the input information is consistent.

16. The method for managing a surrogate existence model according to any one of claims 9 to 15, wherein the output step calculates the degree of fitness for the information generated by the surrogate existence model with respect to the input information and outputs the degree of fitness in association with the information.

17. A proxy existence model management program for causing a computer to execute each of the following procedures: a model pool management procedure, an input information acquisition procedure, an input analysis procedure, a model selection procedure, an information generation procedure, and an output procedure, wherein the model pool management procedure manages a proxy existence model pool by associating at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in a manner that mimics the model target for a large-scale learning model, the input information acquisition procedure acquires user input information, the input analysis procedure analyzes the input information, the model selection procedure selects at least one proxy existence model that mimics the target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation procedure provides the selected proxy existence model with the input information to generate information, and the output procedure outputs the information generated by the proxy existence model.

18. The surrogate existence model management program according to claim 17, wherein the model pool management procedure manages the occupation of the model subject in each surrogate existence model and the tasks corresponding to the occupation as model feature information, linking them with the surrogate existence model; the input analysis procedure analyzes the input information and extracts at least one task included in the input information; and the model selection procedure selects at least one surrogate existence model that mimics a person belonging to the corresponding occupation based on the task.

19. The surrogate existence model management program according to claim 18, wherein the input analysis procedure extracts a plurality of tasks contained in the input information, and the model selection procedure selects a surrogate existence model that mimics a person belonging to the occupation corresponding to each of the plurality of tasks.

20. A computer-readable recording medium that records a proxy existence model management program for causing a computer to execute each of the following procedures: a model pool management procedure, an input information acquisition procedure, an input analysis procedure, a model selection procedure, an information generation procedure, and an output procedure, wherein the model pool management procedure manages a proxy existence model pool by associating at least one proxy existence model with the model feature information of the proxy existence model, the proxy existence model is a model constructed to behave in imitation of a model target for a large-scale learning model, the input information acquisition procedure acquires user input information, the input analysis procedure analyzes the input information, the model selection procedure selects at least one proxy existence model that imitates a target from the proxy existence model pool based on the analysis results of the input information and the model feature information, the information generation procedure provides the selected proxy existence model with the input information to generate information, and the output procedure outputs the information generated by the proxy existence model.

Images