Information provision device, information provision method, and information provision program

Abstract

This information provision device (10) is provided with a discussion unit (33) and a decision unit (34). A discussion unit (33) discusses a legal case by means of a plurality of lawyer accounts that are set in the language model and that perform different claims. The language model is generated by learning data stored in a database (20a) about legal cases and is used for generating answers to input cues. A decision unit (34) decides the proposal quality on the basis of the discussion result of the discussion unit (33) by means of a decision account set in the language model.

Description

[0001] Technology Field The disclosed embodiments relate to an information providing device, an information providing method, and an information providing procedure. Background Technology

[0002] Previously, a system was known that used a language model to generate answer text to a question text input by a user (e.g., see Patent Document 1).

[0003] Patent Document 1: Japanese Patent Publication No. 2022-503838.

[0004] In existing technologies, although response text is generated based on news reports, the application of domain-specific language models is not taken into account. Summary of the Invention

[0005] The present invention was made in view of the above circumstances, and its purpose is to effectively apply a language model specific to a particular domain.

[0006] One embodiment of the information providing device includes: a discussion unit that conducts discussions about legal cases through multiple lawyer accounts that make different claims, each set in a language model, wherein the language model is a language model generated by learning data stored in a database about legal cases and used to generate responses to input prompts; and an adjudication unit that, through adjudication accounts set in the language model, adjudicates the merits of claims based on the discussion results of the discussion unit.

[0007] According to one implementation scheme, a domain-specific language model can be effectively applied. Attached Figure Description

[0008] Figure 1 This is a diagram illustrating the outline of the information providing device according to the first embodiment.

[0009] Figure 2 This is a functional block diagram illustrating a structural example of the information providing device according to the first embodiment.

[0010] Figure 3 This is a flowchart illustrating the processing steps performed by the information providing device involved in the implementation method.

[0011] Figure 4 This is a flowchart illustrating the processing steps performed by the information providing device involved in the implementation method.

[0012] Figure 5 This is a diagram that roughly illustrates an example of a computer hardware structure functioning as an information providing device.

[0013] Figure 6This is a diagram illustrating the processing flow of the information providing device according to the second embodiment.

[0014] Figure 7 This is a diagram showing the structure of an information providing system that includes the information providing device according to the third embodiment.

[0015] Figure 8 This is a diagram illustrating the discussion.

[0016] Figure 9 This is a diagram illustrating the outline of the information providing device according to the fourth embodiment.

[0017] Figure 10 This is a diagram illustrating the outline of the information providing device according to the fifth embodiment.

[0018] Figure 11 This is a diagram illustrating the discussion.

[0019] Figure 12 This is a flowchart representing the processing steps of the discussion.

[0020] Figure 13 This is a diagram illustrating the outline of the information providing device according to the sixth embodiment.

[0021] Figure 14 This is a diagram showing an outline of the discussion related to the sixth embodiment.

[0022] Figure 15 This is a diagram illustrating an example of the prompt words involved in the sixth embodiment.

[0023] Figure 16 This is a flowchart illustrating the processing steps performed by the information providing device according to the sixth embodiment.

[0024] Explanation of reference numerals in the attached figures 1. Information Provision System 10 Information providing device 11 Ministry of Communications 12 Storage Department 13 Control Department 20 databases 20a Legal Case Database 20b Accounting Case Database 20c Medical Case Database 20d Policy Information Database 20e Investment Information Database 31 Acquisition Department 32 Generation Department 33 Discussion Section 34. Adjudication Department 121 Language model information. Detailed Implementation

[0025] The present invention will now be described through embodiments, but these embodiments do not limit the invention as defined in the claims. Furthermore, not all combinations of features described in the embodiments are indispensable solutions to the invention.

[0026] 1. First Implementation Method use Figure 1 The processing flow of the information providing device according to the first embodiment is explained. Figure 1 This is a diagram illustrating the outline of the information providing device according to the first embodiment. Figure 1 This describes the structure of the information providing system 1, which includes the information providing device 10 according to the first embodiment.

[0027] like Figure 1 As shown, the information providing system 1 includes an information providing device 10 and a legal case database 20a. The legal case database 20a is a database that stores data about various laws. For example, the legal case database 20a stores various laws concerning civil and criminal cases, as well as data about previous precedents. Furthermore, the legal case database 20a corresponds to an example of the database involved in the implementation method.

[0028] Information providing device 10 is a device that connects to various devices containing legal case database 20a via plug-ins. These devices include various servers, vehicles, smartphones, and all other devices. In information providing system 1, information providing device 10 connects to these devices via plug-ins to achieve the highest level of artificial intelligence.

[0029] For example, in information providing system 1, the information providing device 10 and each device containing the legal case database 20a have built-in semiconductor chips from the same manufacturer. These semiconductor chips can each perform machine learning or deep learning.

[0030] Furthermore, the semiconductor chips used in the information providing device 10 and the legal case database 20a are chips that conform to their respective housing sizes. For example, assuming the chip size of the information providing device 10 is XL, the server and vehicle use L size, and the user terminal uses S or M size. Additionally, for terminals with housing sizes smaller than the user terminal, SS size is used, and they are placed in a System on a Chip (SoC) to achieve single-chip integration. The XL size semiconductor chip has 200 blocks, the L size has 50 blocks, the M size has 20 blocks, the S size has 10 blocks, and the SS size has 2 blocks; however, the number of blocks for each size is only one example and is not limited to the above numbers. If accessories such as cameras or microphones are to be installed on the user terminal, dedicated accessory chips, different from the semiconductor chips, can be configured in the free space near the semiconductor chips in the SoC.

[0031] Thus, by using semiconductor chips from the same manufacturer in various devices connected to the information providing device 10, the information providing system 1 can be kept secure. That is, the information providing system 1 can effectively evade hacking attacks, virus infections, and deepfakes, and can ensure the privacy between the information providing device 10 and the data source provider containing the legal case database 20a.

[0032] In this embodiment, the information providing device 10 enables the model to learn based on a data source obtained from the legal case database 20a. The model is a language model. For example, ChatGPT from OpenAI is known as a language model (reference: https: / / openai.com / blog / chatgpt). The language model can also be a language model utilizing neural networks such as GAN (Generative Adversarial Networks) or VAE (Variational Autoencoder).

[0033] In this implementation, the language model generates response text to the user's input query text (hereinafter referred to as prompt words). Furthermore, the language model undergoes pre-learning using data sources. The language model can be learned using known machine learning methods.

[0034] Thus, the language model becomes capable of generating responses based on the data source. In this embodiment, the information providing device 10 is connected to the legal case database 20a via a plug-in (step S1), enabling the language model to learn based on the data stored in the legal case database 20a (step S2).

[0035] In this way, the information providing device 10 enables the language model to learn various data stored in the legal case database 20a. Consequently, the language model generates an answer based on the learned legal cases in response to input prompts. That is, the information providing device 10 is capable of generating a language model specifically for the legal field.

[0036] In this implementation, a virtual law firm is designed using a language model specific to the legal field. Specifically, in this implementation, accounts (roles) of "Lawyer A to C," "Prosecutor," and "Adjudication System" are pre-set via system prompts. "Lawyer A to C" are accounts that present different claims. The "Prosecutor" is the account that determines the topics for discussion by Lawyers A to C. Additionally, the "Prosecutor" can also assume the responsibilities of a plaintiff and present the plaintiff's claims. The "Adjudication System" is the account that adjudicates the merits of Lawyers A to C's claims.

[0037] In this embodiment, the system prompts are set as follows: the claims of lawyers A through C are input into the language model as prompts, so that other lawyers can discuss the input prompts.

[0038] This setup allows lawyers A through C to actively engage in discussions using a language model. For example, as... Figure 1 As shown, the "prosecutor" sets the topic of discussion (step S3). This topic is equivalent to the consultation content for a corporate lawyer. Furthermore, the topic can be related to current litigation cases facing the company, or it can be fictional.

[0039] Lawyers A through C will discuss the topics set by the prosecutor (step S4). For example, lawyers A through C will each present different arguments and discuss the logic supporting those arguments. For example, lawyers A through C will discuss each other by developing the logic supporting their own arguments based on relevant laws and previous precedents.

[0040] Then, the "adjudication system" makes a ruling based on the content of the discussion (step S5). For example, if the discussion is sufficient and the prescribed termination conditions are met, the "adjudication system" decides which adjudication system's argument is superior. For example, the "adjudication system" decides the merits of each lawyer's argument by referring to relevant laws and previous precedents.

[0041] Then, the "adjudication system" considers previous precedents and other factors to adjudicate the arguments and logic of each lawyer, selecting the best argument. In this way, the information providing device 10 enables the language model to learn based on data obtained from the legal case database 20a.

[0042] Furthermore, the information providing device 10 uses a language model to allow pre-set accounts to engage in discussions and derives conclusions from these discussions. For example, compared to discussions conducted by actual lawyers, the information providing device 10 can conduct such discussions multiple times. For instance, it can also conduct multiple discussions on the same topic, allowing the arguments of the winner of the first discussion to be discussed in relation to the arguments of the winner of the second discussion.

[0043] That is, by repeatedly engaging in such discussions, the information providing device 10 is able to improve the various claims. Thus, compared to an actual lawyer, the information providing device 10 is able to make meaningful claims.

[0044] Therefore, the information providing device 10 according to the implementation method can effectively apply a language model specific to a particular field.

[0045] use Figure 2 Explain the structure of the information providing device 10. Figure 2 This is a functional block diagram illustrating a structural example of the information providing device according to the first embodiment.

[0046] like Figure 2 As shown, the information providing device 10 includes a communication unit 11, a storage unit 12, and a control unit 13.

[0047] The communications department 11 sends and receives information with legal case database 20a and other entities via the network.

[0048] The storage unit 12 is implemented, for example, using semiconductor storage elements such as RAM (Random Access Memory) and flash memory, or storage devices such as HDD (Hard Disk Drive), SSD (Solid State Drive), and optical disc. The storage unit 12 stores various programs and data. The storage unit 12 stores language model information 121.

[0049] Language model information 121 is information about the language model. For example, language model information 121 is information about the parameters used to construct the language model. These parameters are, for example, the weights and biases of the neural network.

[0050] The control unit 13 is a controller, which may include, for example, a microcomputer with a CPU (Central Processing Unit), ROM (Read Only Memory), RAM, input / output ports, and various circuits. Alternatively, the control unit 13 may be constructed from hardware such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array). The control unit 13 includes an acquisition unit 31, a generation unit 32, a discussion unit 33, and a decision unit 34.

[0051] The Acquisition Department 31 acquires data from the data source provider. For example, the Acquisition Department 31 acquires data from the legal case database 20a (see...). Figure 1 ( ) Obtain data about legal cases. For example, obtain data from the legal case database 20a each time it is updated.

[0052] The generation unit 32 generates a language model. Specifically, the generation unit 32 enables the language model to learn based on data sources about legal cases. Furthermore, the generation unit 32 updates the model's parameters through learning. The generation unit 32 can regenerate the model on each learning iteration, or it can incorporate the difference in data sources between the previous learning iteration and the current one into the model.

[0053] Discussion section 33 uses a language model generated by learning from data stored in a database about legal cases to generate responses to input prompts, enabling discussions to take place between multiple lawyer accounts that are distinct from each other.

[0054] First, the discussion section 33 generates multiple lawyer accounts, prosecutor accounts, and adjudication system accounts using system prompts from the language model. For example, each lawyer account is configured to make different claims and then to participate in discussions supporting those claims.

[0055] For the prosecutor's account, it can be set to set the topic of discussion. Additionally, the prosecutor's account can also be set to act as the plaintiff and participate in discussions with the lawyers' accounts. Furthermore, the adjudication system account can be set to adjudicate the discussion results of the lawyers' accounts, or the discussion results of the lawyers' accounts and the prosecutor's account.

[0056] Discussion section 33, based on the setup of these accounts, sets the topics for discussion. For example, the topics for discussion can be arbitrarily set by the user. For instance, in this case, data from when a user hires a lawyer can be input into the language model as the topic.

[0057] Furthermore, the discussion section 33 can also use user-defined consultation topics as discussion topics. For example, consultation topics here may include: whether a client company should pursue litigation, or consultation on the content of a contract that is favorable to the client company.

[0058] The adjudication department 34 uses a language model to adjudicate claims based on the discussion results of the discussion department 33. The adjudication department 34 corresponds to the adjudication system account. The adjudication department 34 adjudicates the claims of each lawyer's account during a stage where the discussion within the lawyer's account is sufficient.

[0059] For example, the adjudication department 34 inputs the claims of each lawyer's account as prompts into the language model to adjudicate which claim is superior.

[0060] More specifically, the adjudication department 34 uses a linguistic model, comparing it with relevant laws and previous precedents, to determine the best argument. Furthermore, when the subject of discussion is whether litigation should proceed, the adjudication department 34 determines its conclusion based on the discussions held by each lawyer's account.

[0061] Furthermore, when the topic of discussion is the content of the contract, the adjudication department 34 determines the draft contract based on the discussion results of each lawyer's account. The result of the adjudication department 34 is then output to the user terminal of a user not shown in the figure.

[0062] Next, using Figure 3 and Figure 4 The following describes the processing steps performed by the information providing device 10 involved in the implementation method. Figure 3 and Figure 4 This is a flowchart illustrating the processing steps performed by the information providing device 10 involved in the implementation method.

[0063] First, using Figure 3 Explain the steps involved in generating a language model. For example... Figure 3 As shown, the information providing device 10 first obtains data about legal cases from the legal case database 20a (step S101). Next, the information providing device 10 generates a language model based on the data obtained from the legal case database 20a (step S102), and the processing ends.

[0064] Next, using Figure 4 Explain the discussion process using language models. For example... Figure 4 As shown, the information providing device 10 first determines the topic of discussion (step S111). Then, the information providing device 10 begins the discussion of each lawyer's account (step S112).

[0065] Next, the information providing device 10 determines whether the discussion of each lawyer's account has ended (step S113). If it is determined that the discussion has not ended (step S113; no), the processing of step S113 is repeated until the discussion ends.

[0066] Furthermore, if the information providing device 10 determines that the discussion of each lawyer's account has ended (step S113; Yes), it makes a ruling on each claim based on the discussion results (step S114) and ends the process.

[0067] According to this embodiment, by using a language model learned from data stored in the legal case database 20a to set up a virtual law firm, a language model specific to a particular field can be effectively applied.

[0068] Figure 5 This diagram is a schematic representation of an example of a computer hardware structure functioning as an information providing device. A program installed in computer 1200 enables computer 1200 to function as one or more "units" of the apparatus according to this embodiment, or to perform operations associated with the apparatus according to this embodiment or those one or more "units," and / or to perform processes according to this embodiment or stages of those processes. Such a program, in order to enable computer 1200 to perform specific operations associated with several or all of the blocks in the flowcharts and block diagrams described in this specification, can be executed by CPU 1212.

[0069] The computer 1200 of this embodiment includes a CPU 1212, RAM 1214, and a graphics controller 1216, which are interconnected via a host controller 1210. The computer 1200 also includes input / output units such as a communication interface 1222, a storage device 1224, a DVD drive, and an IC card driver, which are connected to the host controller 1210 via an input / output controller 1220. The DVD drive can be a DVD-ROM drive or a DVD-RAM drive, etc. The storage device 1224 can be a hard disk drive or a solid-state drive, etc. The computer 1200 also includes input / output units such as a ROM 1230 and a keyboard, which are connected to the input / output controller 1220 via an input / output chip 1240.

[0070] The CPU 1212 operates according to the program stored in the ROM 1230 and RAM 1214, thereby controlling each unit. The graphics controller 1216 acquires the image data generated by the CPU 1212 into the frame buffer provided in RAM 1214 or into the graphics controller 1216 itself, so that the image data is displayed on the display device 1218.

[0071] Communication interface 1222 communicates with other electronic devices via a network. Storage device 1224 stores programs and data used by the CPU 1212 within computer 1200. DVD drive reads programs or data from DVD-ROM, etc., and provides them to storage device 1224. IC card driver reads programs and data from IC card, and / or writes programs and data to IC card.

[0072] ROM 1230 stores boot programs and / or programs that depend on the hardware of computer 1200, which are executed by computer 1200 when activated. Input / output chip 1240 can also connect various input / output units to input / output controller 1220 via USB port, parallel port, serial port, keyboard port, mouse port, etc.

[0073] The program is provided by a computer-readable storage medium such as a DVD-ROM or an IC card. The program is read from the computer-readable storage medium, installed in a storage device 1224, RAM 1214, or ROM 1230, which is also an example of a computer-readable storage medium, and executed by the CPU 1212. The information processing described within these programs is read by the computer 1200, resulting in cooperation between the program and the aforementioned various types of hardware resources. The apparatus or method can also be configured to perform information manipulation or processing using the computer 1200.

[0074] For example, when communication is performed between computer 1200 and external devices, CPU 1212 can execute a communication program loaded into RAM 1214, and perform communication processing on communication interface 1222 based on the processing described in the communication program. Under the control of CPU 1212, communication interface 1222 reads transmission data stored in a transmission buffer area provided in recording media such as RAM 1214, storage device 1224, DVD-ROM, or IC card, and sends the read transmission data to the network, or writes received data received from the network into a receive buffer area provided on the recording medium, etc.

[0075] Furthermore, the CPU 1212 can read all or necessary portions of files or databases stored in external recording media such as storage device 1224, DVD drive (DVD-ROM), and IC card into RAM 1214, and perform various types of processing on the data in RAM 1214. The CPU 1212 can then write the processed data back to the external recording medium.

[0076] Various types of information, such as programs, data, tables, and databases, can be stored in the recording medium and processed. The CPU 1212 can execute various types of processing on data read from RAM 1214, as described throughout this disclosure and specified by a sequence of program instructions, including various types of operations, information processing, conditional judgments, conditional branches, unconditional branches, and information retrieval / replacement, and write the results back to RAM 1214. Furthermore, the CPU 1212 can retrieve information from files, databases, etc., within the recording medium. For example, if multiple entries are stored in the recording medium, and each entry has an attribute value for a first attribute associated with the attribute value of a second attribute, the CPU 1212 can retrieve from these multiple entries the entry whose attribute value of the first attribute matches a specified condition, read the attribute value of the second attribute stored in that entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

[0077] The aforementioned programs or software modules may be stored on or near the computer 1200 in a computer-readable storage medium. Furthermore, a recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet may be used as a computer-readable storage medium, thereby providing programs to the computer 1200 via the network.

[0078] In this embodiment, the boxes in the flowcharts and block diagrams may represent stages of a process for performing an operation or "parts" of a device that performs the operation. Specific stages and "parts" may be implemented by dedicated circuitry, programmable circuitry supplied along with computer-readable instructions stored on a computer-readable storage medium, and / or a processor supplied along with computer-readable instructions stored on a computer-readable storage medium. Dedicated circuitry may include digital and / or analog hardware circuitry, and may include integrated circuits (ICs) and / or discrete circuitry. Programmable circuitry may include reconfigurable hardware circuitry such as field-programmable gate arrays (FPGAs) and programmable logic arrays (PLAs), containing logical AND, OR, XOR, NAND, NOR, and other logical operations, flip-flops, registers, and storage elements.

[0079] A computer-readable storage medium can include any tangible device capable of storing instructions executable by a suitable device. As a result, a computer-readable storage medium having instructions stored therein comprises an article of products containing instructions executable for creating means for performing operations specified in a flowchart or block diagram. Examples of computer-readable storage media include electronic storage media, magnetic storage media, optical storage media, electromagnetic storage media, semiconductor storage media, etc. More specific examples of computer-readable storage media include floppy disks (registered trademark), floppy disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), electrically erasable programmable read-only memory (EEPROM), static random access memory (SRAM), optical disc read-only memory (CD-ROM), digital versatile optical disc (DVD), Blu-ray disc (registered trademark), memory sticks, integrated circuit cards, etc.

[0080] Computer-readable instructions may include any of the following: assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code described by any combination of one or more programming languages ​​including object-oriented programming languages ​​such as Smalltalk (registered trademark), JAVA (registered trademark), C++, and traditional procedural programming languages ​​such as the "C" programming language or similar programming languages.

[0081] Computer-readable instructions are provided locally or via a wide area network (WAN) such as a local area network (LAN), the Internet, or the processor or programmable circuitry of a general-purpose computer, special-purpose computer, or other programmable data processing device to generate means for performing the operations specified in a flowchart or block diagram, so that the computer-readable instructions can be executed to produce means for performing the operations specified in the flowchart or block diagram. Examples of processors include computer processors, processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc.

[0082] 2. Second Implementation Method Next, the second embodiment will be described. Furthermore, descriptions that overlap with the first embodiment will be omitted below. Figure 6 The processing flow of the information providing device according to the second embodiment is explained. For example... Figure 6 As shown, the information providing system 1 according to the second embodiment includes an information providing device 10 and an accounting case database 20b. The accounting case database 20b is a database that stores data on various laws. For example, the accounting case database 20b stores data on various laws related to accounting, as well as data on previous case precedents.

[0083] The information providing device 10 is a device that connects to various devices containing the accounting case database 20b via plug-ins. These devices include various servers, vehicles, smartphones, and all other devices.

[0084] In this embodiment, the information providing device 10 enables the model to learn based on a data source obtained from the accounting case database 20b. The information providing device 10 is connected to the accounting case database 20b via a plug-in (step S1-1), allowing the language model to learn based on data stored in the accounting case database 20b (step S2-1).

[0085] In this way, the information providing device 10 enables the language model to learn various data stored in the accounting case database 20b. Consequently, the language model generates responses based on the learned legal cases in response to input prompts. That is, the information providing device 10 is capable of generating a language model specifically for the legal field.

[0086] A virtual accounting firm is designed using a language model specific to the legal field. Specifically, accounts (roles) for "Accountants A-C," "IRS," and "Adjudication System" are pre-set via system prompts. "Accountants A-C" are accounts that make differing claims. "IRS" is the account that participates in discussions from the perspective of Accountants A-C. Additionally, as described later, "IRS" can also determine the topics of discussion for Accountants A-C. "Adjudication System" is the account that adjudicates the merits of Accountants A-C's claims.

[0087] The system prompts are set as follows: The claims of accountants A through C are input into the language model as prompts, which are then discussed by other accountant accounts.

[0088] This setup allows "Accountants A through C" to actively engage in discussions using a language model. For example, ... Figure 6 As shown, the discussion topic is set by the IRS (step S3-1). This topic is equivalent to the content of an accountant's consultation with the company. Furthermore, the topic can be about current accounting cases facing the company, or it can be fictional.

[0089] Accountants A through C will discuss the topics set by the IRS (step S4-1). For example, Accountants A through C will each make different claims and discuss the logic supporting those claims. For example, Accountants A through C will discuss each other by developing the logic supporting their claims based on relevant laws and previous case law.

[0090] Then, the "adjudication system" makes a ruling based on the content of the discussion (step S5-1). For example, if the discussion is sufficient and the prescribed termination conditions are met, the "adjudication system" decides which accountant's claim is superior. For example, the "adjudication system" decides the merits of each accountant's claim by referring to relevant laws and previous precedents.

[0091] Then, the "adjudication system" considers previous precedents and other precedents to adjudicate the claims and logic of each accountant, selecting the best claim. In this way, the information providing device 10 enables the language model to learn based on data obtained from the accounting case database 20b.

[0092] Furthermore, the information providing device 10 uses a language model to allow pre-set accounts to engage in discussions and derives conclusions from these discussions. For example, compared to discussions conducted by actual accountants, the information providing device 10 can conduct such discussions multiple times. For instance, it can also conduct multiple discussions on the same topic, allowing the claims of the winners of the first discussion to be compared with those of the winners of the second discussion.

[0093] More specifically, the information providing device 10 uses a language model generated by learning from data stored in a database about accounting cases to generate responses to input prompts, which are discussed by multiple accountant accounts that are different from each other.

[0094] First, the information providing device 10 generates multiple accountant accounts, IRS accounts, and adjudication system accounts using system prompts from a language model. For example, each accountant account is configured to make different claims and then to discuss in support of its own claims.

[0095] For the IRS account, set it to be the topic of discussion. Additionally, the IRS account can be set to participate in discussions between the accountants' accounts from the IRS's perspective. Furthermore, for the adjudication system account, set it to adjudicate the discussion results of the accountants' accounts, or the discussion results of the accountants' accounts and the IRS account.

[0096] The information providing device 10 sets the discussion topics based on these accounts. For example, the discussion topics can be arbitrarily set by the user. For instance, in this case, data from when the user hires an accountant can be input into the language model as the topic.

[0097] In addition, the information providing device 10 can also use consultation content set by the user as the topic of discussion.

[0098] Information providing device 10 uses a language model to adjudicate claims based on the results of discussions. Information providing device 10 adjudicates claims for each accountant's account at a stage where discussions of the accountant's account are sufficient.

[0099] For example, the information providing device 10 inputs the claims of each accountant's account as prompts into the language model to determine which claim is superior.

[0100] More specifically, the information providing device 10 uses a language model to compare the argument with relevant laws and previous precedents, and then determines the best claim. The decision of the information providing device 10 is then output to a user terminal (not shown) or similar device.

[0101] That is, by repeatedly conducting such discussions, the information providing device 10 is able to improve the claims. Thus, compared with an actual accountant, the information providing device 10 is able to make meaningful claims.

[0102] Therefore, the information providing device 10 according to the second embodiment can effectively apply a language model specific to a particular domain.

[0103] 3. Third Implementation Method Next, the third embodiment will be described. Figure 7 This is a diagram illustrating the outline of the information providing device 10 according to the third embodiment. Figure 7 This describes the structure of the information providing system 1, which includes the information providing device 10 according to the third embodiment.

[0104] like Figure 7 As shown, the information providing system 1 includes an information providing device 10 and a medical case database 20c. The medical case database 20c is a database that stores various data related to medicine. For example, the medical case database 20c stores papers, knowledge, and various medical records related to medicine, and further stores various data related to pharmaceuticals, insurance, and medical equipment, etc. The information providing device 10 is connected to the medical case database 20c via a plug-in (steps S1-2), allowing the language model to learn based on the data stored in the medical case database 20c (steps S2-2).

[0105] In this way, the information providing device 10 enables the language model to learn various data stored in the medical case database 20c. Consequently, the language model generates a response based on the learned medical data in response to input prompts. That is, the information providing device 10 is capable of generating a language model specifically for the medical field.

[0106] In this embodiment, a language model specific to the medical field is used to provide suggestions on medical policies for patients. Specifically, in this embodiment, accounts (roles) for "doctor," "patient," "family member," "experts A to C," and "adjudication system" are pre-set using system prompts from the language model.

[0107] Here, it is assumed that the accounts in this embodiment include robot accounts and human accounts. The robot account is a language model capable of setting system prompts. The robot account uses the language model to output responses. On the other hand, the human account is a human. That is, the human account outputs responses corresponding to human input.

[0108] Figure 1 The "Doctor," "Experts A-C," "Adjudication System," "Patient," and "Family Member" are the superior accounts. Each superior account must contain at least one robot account or a human account. Figure 7 The white circles represent bot accounts. Additionally, Figure 1 The black circles represent human accounts.

[0109] Hereinafter, robot accounts or human accounts contained within a superior account will sometimes be referred to as subordinate accounts. Furthermore, where it is obvious from the preceding and following descriptions and accompanying diagrams, and where there is no distinction between superior and subordinate accounts, "superior" and "subordinate" will be omitted, and the term "account" will be used alone.

[0110] The following explanation assumes all accounts are bot accounts. This is achieved by appropriately replacing bot accounts with human accounts, or by adding human accounts as subordinate accounts within a higher-level account. Figure 7 The structure.

[0111] "Doctor" is, for example, the superior account of the patient's attending physician. Figure 7 The example shown here represents a single "doctor" account, but multiple accounts can actually be set up. Furthermore, multiple accounts can also represent doctors specializing in different fields (such as internal medicine or surgery).

[0112] "Patient" refers to the patient's account. "Family Member" refers to the patient's family member's account. "Experts A through C" are the accounts of the experts involved in the patient's treatment.

[0113] Specific examples of expert accounts include medical researchers, hospitals, nurses, pharmacists, pharmaceutical manufacturers, and medical device manufacturers. The "adjudication system" is the account that adjudicates the results of the discussions among the various accounts. For example, the adjudication system decides the merits of each account's claims and provides recommendations to the actual physician (e.g., the attending physician).

[0114] In this embodiment, the system prompts are set as follows: the opinions of multiple doctor accounts are input into the language model as prompts, so that other doctor accounts can discuss the input prompts.

[0115] This setup allows doctors' accounts to proactively engage in discussions using language models. For example... Figure 7 As shown, the information providing device 10 uses a language model to conduct discussions about the patient's treatment strategy through various accounts (step S3-2).

[0116] For example, initial input of various data such as the patient's age, gender, and medical records as prompts initiates a discussion. More specifically, the doctor account is further divided into multiple accounts such as Doctor 1, Doctor 2, and Doctor 3. Doctor 1 is categorized by personality type, Doctor 2 by personality type, and Doctor 3 by personality type, etc., and each account is discussed using a language model. Furthermore, Doctor 1 or Doctor 2 can further subdivide into sub-accounts such as Doctor 1-1, 1-2, 1-3, or Doctor 2-1, Doctor 2-2, Doctor 2-3 to discuss medical policies, deriving conclusions from Doctor 1 or Doctor 2. Virtual experts engage in billions of debates per second. This allows for verification and problem-solving from all angles, instantly providing highly accurate diagnoses and treatment recommendations based on prior medical data. Furthermore, Doctor 1 conducts multimodal data analysis and suggests: integrating and analyzing various types of medical data, such as genetic data, blood data, MRI, and X-ray images, to obtain the results; Doctor 2 uses a real-time learning model to acquire new data from medical settings worldwide in real time and suggests: always incorporating the latest medical information and research findings into treatment; Doctor 3 possesses advanced simulation capabilities and suggests: virtually simulating the progression of the disease and the effects of treatments, predicting the most appropriate treatment and its outcomes in advance. Then, the doctor account can also use the combined judgment derived from the outputs of Doctors 1, 2, and 3 to determine the treatment strategy. Other considerations for the doctor account could include: promoting personalized medicine by individually optimizing treatment plans based on each patient's genetic information and lifestyle; or collaborating with 3D bioprinting to determine the optimal organ and tissue structure for the patient and using 3D bioprinting technology to generate individually customized artificial organs and tissues.

[0117] For example, the family member account derives conclusions regarding medical policy from multiple accounts set up within the family member's account. Each expert account, within its respective field, conducts discussions through multiple accounts, deriving conclusions regarding medical policy for each field. Then, the physician account, based on these discussions across different fields, conducts a discussion regarding the patient's treatment plan. Alternatively, a final discussion can also be held through a unified account containing all the individual accounts.

[0118] For example, the physician's account, as the patient's treatment plan, involves discussions regarding treatment methods, costs, risks, cure rates, and the methods and timing of informing the patient and their family. Then, the "adjudication system" makes a decision based on the content of the discussion (step S4-2). For example, the "adjudication system" consists of multiple accounts. These multiple accounts discuss the content of the discussions by the physician's account, etc., to determine the final medical policy. For example, if the discussion is sufficient and the prescribed termination conditions are met, the adjudication system determines the most appropriate medical policy by comparing the claims of each account with the data stored in the medical case database 20c.

[0119] Then, the "adjudication system" will determine the medical policy, for example, and recommend it to the patient's attending physician. That is, in information provision system 1, by using a language model, it is possible to simulate extensive discussions about medical policies by attending physicians, etc. Furthermore, since the language model is pre-learned from a large amount of data about medical cases, it is possible to recommend medical policies based on various prior knowledge.

[0120] Thus, the information providing device 10 involved in the implementation generates a language model based on various knowledge entered in the medical case database 20c, and makes recommendations on medical policies after discussing using the language model.

[0121] Therefore, the information providing device 10 according to the implementation method can effectively apply a language model specific to a particular field.

[0122] The information providing device 10 uses a language model to conduct a discussion through the following steps 1 to 5.

[0123] Step 1: Set up a multi-functional virtual expert There are hundreds of combinations of virtual experts in different medical fields. For example, there are specific roles such as "Neurologist A", "Neurologist B", and "Neurologist C" for neurology, and "Cardiac Expert A", "Cardiac Expert B", and "Cardiac Expert C" for cardiology.

[0124] Step 2: High-speed information exchange These experts exchange information and engage in billions of debates per second, based on patient data. This allows them to instantly assess a patient's condition, symptoms, and treatment options.

[0125] Step 3: Data Processing Virtual experts advanced the discussion by sharing various types of information, including MRI data, genetic data, and blood data. This allowed them to quickly derive the optimal solution from multiple perspectives.

[0126] Step 4: Self-learning and updating When new knowledge, information, or treatments are discovered during discussions, they are immediately reflected in the database and applied to subsequent discussions. This allows for diagnosis and treatment based on the latest medical information.

[0127] Step 5: Consensus Formation Discussions, through the clash of differing opinions and viewpoints, culminate in a final "consensus." Consequently, when making recommendations for patient diagnosis and treatment, a comprehensive approach incorporating information and knowledge from multiple sources is adopted. Multiple treatment proposals can also be proposed.

[0128] The above explains the doctor's account, but roles are also set for the patient account, family account, and expert A to C accounts. By further dividing each account into various roles, discussions among the various roles can be held to form a consensus among patients, family members, and experts A to C.

[0129] The information providing device 10, through an adjudication account set in the language model, adjudicates the discussion results of the information providing device 10 and suggests treatment strategies. For example, the information providing device 10 makes an adjudication when the prescribed termination conditions, such as the discussion of the information providing device 10 being sufficient, are met.

[0130] For example, the information providing device 10 refers to various data entered in the medical case database 20c, and, among the medical policies discussed by each doctor's account, determines the most appropriate medical policy and recommends that policy. Furthermore, it can also incorporate various consensuses from the doctors' consensus results, such as the consensus of patients (avoiding hospitalization if possible), the consensus of family members (avoiding scarring if possible), and the consensus of expert A (how much insurance can cover), to recommend a medical policy.

[0131] Discussion in cases involving human accounts Previously, a discussion was conducted assuming the subordinate account was a robot account. Here, a discussion will be conducted in the case where the subordinate account includes a human account. In this case, the processing content of steps 1 and 2 of the information providing device 10 changes.

[0132] First, in step 1, the information providing device 10 alters its structure so that at least some of the higher-level accounts include human accounts as lower-level accounts. For example, the information providing device 10 replaces robot accounts with human accounts, or adds new human accounts. The human accounts correspond to users, i.e., real, existing humans.

[0133] Next, in step 2, the information providing device 10 provides text prompts, output as responses by each account, to the user corresponding to the human account. Then, the information providing device 10 accepts text input from the user. Furthermore, the actions of each account outputting text responses, and the actions of outputting text responses, are referred to as "speaking." "Speaking" includes the text input by the user.

[0134] The chatbot account processes human accounts' statements in the same way as its own. For example, the chatbot account inputs statements from other chatbot accounts and human accounts as prompts into its own language model and then outputs a response.

[0135] use Figure 8 Explanation of the processing of information providing device 10. Figure 8 This is a diagram illustrating the discussion. For example... Figure 8 As shown, the information providing device 10 will be discussed as the stages progress. Here, it is assumed that Doctor 1, as the superior account, consists of three subordinate accounts: Doctor 1-1, Doctor 1-2, and Doctor 1-3.

[0136] Doctor 1-1 is a robot account for surgeons. For example, Doctor 1-1's language model output prioritizes text such as surgical treatment methods.

[0137] Doctor 1-2 is a robot account for internal medicine doctors. For example, Doctor 1-2's language model output prioritizes text such as internal medicine treatment methods.

[0138] Furthermore, the entity of the robot account is a language model, and discussion is initiated by repeatedly inputting prompts into the language model through the information providing device 10. These prompts may be, for example, the language model's statements from the previous stage. The type of data emphasized in the statements can be set through the system prompts. As a result, the response tendencies of each robot account differ.

[0139] Doctors 1-3 are human accounts. Doctors 1-3 correspond to users. Furthermore, the personality traits (what they value) of Doctors 1-3 can be arbitrary.

[0140] like Figure 8 As shown, in the first stage, Doctor 1-1 states, "I think it's best to perform open surgery as soon as possible." Here, the information providing device 10 causes the stage to continue.

[0141] In the second stage, based on the results of inputting the statements of Doctor 1-1 from the first stage as cue words into the language model, Doctor 1-2 said, "Considering the patient's physical condition, wouldn't it be better to first observe the situation through drug treatment?"

[0142] Next, in the third stage, the human account says, "The blood test results aren't back yet." The human account's statement is user-inputted content. For example, the information providing device 10 can also display the statements from each account on the user terminal via a chat UI and accept user input from the chat UI's input field.

[0143] Furthermore, each account, including both bot and human accounts, is not required to speak at all stages. The information providing device 10 only needs to allow at least one account to speak at each stage.

[0144] Furthermore, when the accounts participating in the discussion include human accounts, it is difficult to get users to speak every single time across the sometimes billions of stages. Therefore, for example, the information providing device 10 may also accept user speech only in specific stages, such as once out of every 10,000 stages.

[0145] Furthermore, the information providing device 10 can also accept user comments under certain conditions. These conditions include the bot account repeating the same content a certain number of times. By having the bot account use the user's comments as prompts, it is hoped that a stalled discussion will begin to move towards a conclusion.

[0146] In this way, if the answer meets the conditions, the information providing device 10 prompts the user with the answer. Then, the information providing device 10 collects the text entered based on the prompt's answer.

[0147] Furthermore, sometimes users possess knowledge that was not used when the language model of the bot account was learning. In such cases, including a human account may yield more useful conclusions in the discussion compared to cases with only a bot account.

[0148] 4. Fourth Implementation Method Next, the fourth embodiment will be described. Figure 9 This is a diagram illustrating the outline of the information providing device according to the fourth embodiment. Figure 9 This describes the structure of the information providing system 1, which includes the information providing device 10 according to the fourth embodiment.

[0149] like Figure 9 As shown, the information providing system 1 includes an information providing device 10 and a policy information database 20d. The policy information database 20d is a database that stores various data about various policies of the state and various municipalities.

[0150] The information providing device 10 is a device that connects to various devices containing the policy information database 20d via plug-ins. These devices include various servers, vehicles, smartphones, and all other devices.

[0151] In this embodiment, the information providing device 10 learns the model based on the data source obtained from the policy information database 20d. In this embodiment, the information providing device 10 is connected to the policy information database 20d via a plug-in (steps S1-3), so that the language model learns based on the data stored in the policy information database 20d (steps S2-3).

[0152] In this way, the information providing device 10 enables the language model to learn various data stored in the policy information database 20d. Consequently, the language model generates a response based on the learned policy-related data in response to input prompts. That is, the information providing device 10 is capable of generating a language model specifically for the policy domain.

[0153] In this implementation, a language model specific to the policy domain is used to make recommendations regarding national or local government policies. Specifically, in this implementation, the accounts (roles) of "Speaker A to C", "Expert A to C", and "Adjudication System" are pre-set using system prompts from the language model.

[0154] "Speakers A through C" are accounts corresponding to the heads of the state or local governments. That is, Speakers A through C correspond to the accounts of the Prime Minister, prefectural governor, mayor, town mayor, village head, etc. In this implementation, multiple accounts such as Speaker A-1, Speaker A-2, and Speaker A-3 within Speaker A, and Speaker B-1, Speaker B-2, and Speaker B-3 within Speaker B, discuss and derive the policy guidelines for Speaker A or Speaker B.

[0155] "Experts A through C" are expert accounts in various fields that participate in the operation of national or local autonomous bodies.

[0156] Specific examples of expert accounts include those in the financial, security, administrative, energy, and healthcare sectors. The "adjudication system" is the account that adjudicates the results of the discussions among the various accounts. For instance, the adjudication system decides the merits of each account's claims and makes recommendations. Multiple accounts, such as expert accounts A-1, A-2, and A-3 in expert account A, and expert accounts B-1, B-2, and B-3 in expert account B, engage in discussions, thereby deriving conclusions on policy guidelines for each sector.

[0157] In this embodiment, the system prompts are set as follows: the claims of multiple speaker accounts are input into the language model as prompts, so that other speaker accounts can discuss the input prompts.

[0158] This setup allows each speaker's account to proactively engage in discussions using a language model. For example... Figure 9As shown, the information providing device 10 uses a language model to conduct political discussions through various accounts (step S3-3).

[0159] For example, Speaker accounts A through C represent different political ideologies such as conservatives, reformers, right-wingers, and left-wingers, and each uses a language model to conduct discussions. Virtual experts engage in billions of debates per second. This enables verification and problem-solving from all angles, and instantly provides policy recommendations.

[0160] Similarly, each expert account discusses their respective field through multiple accounts, deriving conclusions on policy guidelines for each field. Then, the speaker's accounts A through C conduct their own discussions on policy guidelines based on these field-specific discussions. Alternatively, the final discussion can also be conducted through all accounts that include all the participating experts.

[0161] For example, the speaker's account conducts discussions on policy, and the "adjudication system" makes a ruling based on the content of the discussion (step S4-3). For example, the "adjudication system" consists of multiple accounts. These multiple accounts discuss the content of the discussion by the speaker's account, etc., and adjudicate the final policy. For example, if the prescribed termination conditions are met, such as sufficient discussion, the adjudication system determines the most appropriate policy direction by comparing the claims of each account with the data stored in the policy information database 20d.

[0162] Then, the "adjudication system" recommends the determined policy direction to the actual speaker. That is, in information provision system 1, extensive discussions on policy can be simulated by using a language model.

[0163] Thus, the information providing device 10 involved in the implementation method generates language models based on various knowledge entered in the policy information database 20d, and proposes suggestions on policy guidelines after discussing using the language models.

[0164] Therefore, the information providing device 10 according to the implementation method can effectively apply a language model specific to a particular field.

[0165] Furthermore, the above discussion is conducted through the following steps 1 to 5.

[0166] Step 1: Set up a multi-functional virtual expert There are hundreds of combinations of virtual experts in different policy fields. For example, for political science, there are specific roles such as "Political Science Expert A", "Political Science Expert B", and "Political Science Expert C", and for economics, there are specific roles such as "Economics Expert A", "Economics Expert B", and "Economics Expert C".

[0167] Step 2: High-speed information exchange These experts exchange information and engage in billions of debates per second, based on various data. This allows them to instantly determine the best policies and economic countermeasures.

[0168] Step 3: Data Processing Virtual experts advance the discussion by sharing various types of information, such as stock prices and birth rates. This allows for the rapid derivation of the optimal solution from multiple perspectives.

[0169] Step 4: Self-learning and updating When new knowledge or information is discovered during discussions, it is immediately reflected in the database and applied to subsequent discussions. This allows for diagnosis and treatment based on the latest policy information.

[0170] Step 5: Consensus Formation The discussion, through the clash of different opinions and viewpoints, eventually reached a "consensus." Therefore, when making policy recommendations, a comprehensive approach was taken, incorporating information and knowledge from multiple perspectives.

[0171] The above explains the Speaker's account, but roles are also set for Experts A through C. By further dividing each account into various roles, discussions among these roles can be held to form a consensus among Experts A through C.

[0172] 5. Fifth Implementation Method Next, the fifth embodiment will be described. Figure 10 This is a diagram illustrating the outline of the information providing device involved in the implementation method. Figure 10 This describes the structure of the information providing system 1, which includes the information providing device 10 involved in the implementation.

[0173] like Figure 10 As shown, the information providing system 1 includes an information providing device 10 and an investment information database 20e. The investment information database 20e is a database that stores various data about investments.

[0174] In information provision system 1, investment information database 20e stores information published in various news media (Nikkei average, long and short-term interest rates, employment statistics, etc.) as well as information acquired independently by information provision system 1.

[0175] The investment information database 20e, for example, analyzes close-range satellite images to independently acquire and store information such as crop maturity, the number of vehicles parked in the parking lots of large supermarkets, POS data, individual warehouse information, credit card information, and residential sales market data. In this embodiment, the information providing device 10 is connected to the investment information database 20e via a plug-in (steps S1-4), allowing the language model to learn based on the data stored in the investment information database 20e (steps S2-4).

[0176] In this way, the information providing device 10 enables the language model to learn various data stored in the investment information database 20e. Consequently, the language model generates a response based on the learned investment-related data in response to input prompts. That is, the information providing device 10 is capable of generating a language model specifically for the investment field.

[0177] In this embodiment, a language model specific to the investment field is used to provide investment advice. Specifically, in this embodiment, the system prompts of the language model are used to pre-set accounts (roles) such as "Investor A to C", "Expert A to C", "Adjudication System", and "Risk Management".

[0178] Here, it is assumed that the accounts in this embodiment include robot accounts and human accounts. The robot account is a language model capable of setting system prompts. The robot account uses the language model to output responses. On the other hand, the human account is a human. That is, the human account outputs responses corresponding to human input.

[0179] Figure 10 "Investors A-C", "Experts A-C", "Judgment System", and "Risk Management" are the superior accounts. Each superior account must contain at least one of the following: a bot account or a human account. Figure 10 The white circles represent bot accounts. Additionally, Figure 10 The black circles represent human accounts.

[0180] Hereinafter, robot accounts or human accounts contained within a superior account will sometimes be referred to as subordinate accounts. Furthermore, where it is obvious from the preceding and following descriptions and accompanying diagrams, and where there is no distinction between superior and subordinate accounts, "superior" and "subordinate" will be omitted, and the term "account" will be used alone.

[0181] The following explanation assumes all accounts are bot accounts. This is achieved by appropriately replacing bot accounts with human accounts, or by adding human accounts as subordinate accounts within a higher-level account. Figure 10 The structure.

[0182] "Investors A through C" are, for example, the superior accounts corresponding to institutional or individual investors operating so-called funds, etc. In this embodiment, Investor A consists of Investor A-1, Investor A-2, and Investor A-3; Investor B consists of Investor B-1, Investor B-2, and Investor B-3, etc. Each Investor A through C consists of multiple subordinate accounts. Investor A, through discussions with Investors A-1, A-2, and A-3, derives its own conclusions regarding investment strategies. For example, Investor A-1 might prioritize inflation rates, Investor A-2 might prioritize employment statistics, and Investor A-3 might prioritize industrial production indices. Through separate discussions, Investor A's conclusions are derived. Thus, there are: 1. Inflation rate: This indicates the rate at which prices rise.

[0183] 2. Unemployment rate: The proportion of unemployed persons in the labor force.

[0184] 3. GDP growth rate: This indicates the growth of the gross domestic product.

[0185] 4. Consumer Confidence Index: Consumers' level of trust in the economy.

[0186] 5. Industry Production Index: Represents production activities in manufacturing, mining, and public utilities.

[0187] 6. Job creation: The number of new employees hired.

[0188] 7. Real wages: Wage trends that take into account price changes.

[0189] 8. Residential Sales Data: Sales status of newly built and existing residential properties.

[0190] 9. Consumer Price Index (CPI): Represents price changes for the average consumer.

[0191] 10. Producer Price Index (PPI): Price changes during the production stage.

[0192] 11. Retail sales: The sales performance of the retail industry.

[0193] 12. Commercial inventory: The amount of goods and materials held by a company.

[0194] 13. International trade balance: the difference between exports and imports.

[0195] 14. Manufacturing Order Data: Order status of the manufacturing industry.

[0196] 15. Non-manufacturing activity index: Represents economic activity in the service sector and other sectors.

[0197] 16. Trends in the financial markets: stock prices, bond yields, currency prices, etc.

[0198] 17. Trends in bank lending: The status of bank lending activities.

[0199] 18. Core inflation rate: The inflation rate excluding food and energy.

[0200] 19. Wage growth rate: This indicates the growth of average wages.

[0201] 20. Trends in the world economy: economic growth rates and policy trends in major countries.

[0202] It features various indicators and is characterized by discussions among investors who value different indicators. "Expert A-C" are expert accounts covering various areas of investment.

[0203] Examples of investment areas include commodities, corporate bonds, real estate, FX, and portfolio management. The "adjudication system" is an account that adjudicates the results of discussions among various accounts. For example, the adjudication system decides the merits of each account's claims and provides recommendations.

[0204] In this embodiment, the system prompts are set as follows: the claims of multiple accounts are input into the language model as prompts, so that other accounts can discuss the input prompts.

[0205] This setup enables each account to actively engage in discussions using a language model. The information providing device 10 uses the language model to facilitate discussions about investment strategies among the accounts (steps S3-4).

[0206] For example, investor accounts A through C represent accounts with different investment philosophies, and each is discussed using a language model. Virtual experts engage in billions of debates per second. This enables verification and problem-solving from all angles, instantly providing investment strategy recommendations.

[0207] Similarly, each expert account conducts discussions within its respective field, using multiple accounts for each field, deriving conclusions regarding investment strategies for each field. Then, investor accounts discuss their investment strategies based on these field-specific discussions. Alternatively, the final discussion may include all accounts from each field.

[0208] As a result of these discussions, the "adjudication system" makes a decision based on the content of the discussions (step S4-4). For example, the "adjudication system" consists of multiple accounts. The multiple accounts constituting the adjudication system discuss the content of discussions regarding investor accounts, etc., to determine the final investment policy. For example, if the prescribed termination conditions are met, such as sufficient discussion, the adjudication system determines the most appropriate investment policy by comparing the claims of each account with the data stored in the investment information database 20e.

[0209] Then, the "adjudication system" advises the actual investors on the determined investment strategy. That is, in information provision system 1, a large-scale discussion on the investment strategy can be simulated by using a language model.

[0210] Furthermore, "Risk Management" discusses the risks in the current investment portfolio based on the latest information entered into the Investment Information Database 20e. "Risk Management" consists of multiple accounts, which discuss investment risks. As a result, if the risk of an investment exceeds a threshold, "Risk Management" recommends terminating the investment (step S5-4).

[0211] That is, the information providing device 10 can avoid losses in investment in advance or control them to a minimum through the function of "risk management".

[0212] Thus, the information providing device 10 involved in the implementation method generates language models based on various knowledge entered in the investment information database 20e, and proposes suggestions on investment policies after discussing using the language models.

[0213] Therefore, the information providing device 10 according to the implementation method can effectively apply a language model specific to a particular field.

[0214] Furthermore, the above discussion is achieved through steps 1 to 5 below.

[0215] Step 1: Set up a multi-functional virtual expert There are hundreds of virtual experts in different investment fields. For example, there are specific roles such as "Real Estate Expert A", "Real Estate Expert B", and "Real Estate Expert C" for the real estate field, and "FX Expert A", "FX Expert B", and "FX Expert C" for the FX field.

[0216] Step 2: High-speed information exchange These experts exchange information and engage in billions of debates per second, based on various data. This allows them to instantly determine the best investment strategies and other appropriate decisions.

[0217] Step 3: Data Processing Virtual experts advance the discussion by sharing various types of information, such as stock prices. This allows for the rapid derivation of optimal solutions from multiple perspectives.

[0218] Step 4: Self-learning and updating When new knowledge or information is discovered during discussions, it is immediately reflected in the database and applied to subsequent discussions. This allows for the provision of optimal investment strategies based on the latest investment information.

[0219] Step 5: Consensus Formation The discussion, through the clash of different opinions and viewpoints, culminated in a final "consensus." Therefore, when recommending investment strategies, a comprehensive approach incorporating information and knowledge from multiple sources is adopted.

[0220] The above explains the investor account, but roles are also set for expert accounts A through C. By further dividing each account into various roles, discussions among these roles can be held to form a consensus among experts A through C.

[0221] The information providing device 10, through an adjudication account set in the language model, adjudicates the discussion results of the information providing device 10 and recommends investment strategies. For example, the information providing device 10 makes an adjudication when the prescribed termination conditions, such as the discussion of the information providing device 10 being sufficient, are met.

[0222] For example, the information providing device 10 refers to various data entered in the investment information database 20e, determines the most appropriate investment strategy from the investment strategies discussed by various investor accounts, and recommends that investment strategy. Alternatively, it can incorporate various consensuses, such as the consensus of expert A, into the consensus results of investor accounts to recommend an investment strategy.

[0223] Furthermore, if the risk management account determines that the risk is high, the information providing device 10 suggests that the investor withdraw their investment. This helps mitigate the risks associated with the investment.

[0224] Discussion in cases involving human accounts Previously, a discussion was conducted assuming the subordinate account was a robot account. Here, a discussion will be conducted in the case where the subordinate account includes a human account. In this case, the processing content of steps 1 and 2 of the information providing device 10 changes.

[0225] First, in step 1, the information providing device 10 alters its structure so that at least some of the higher-level accounts include human accounts as lower-level accounts. For example, the information providing device 10 replaces robot accounts with human accounts, or adds new human accounts. The human accounts correspond to users, i.e., real, existing humans.

[0226] Next, in step 2, the information providing device 10 provides text prompts, output as responses by each account, to the user corresponding to the human account. Then, the information providing device 10 accepts text input from the user. Furthermore, the act of outputting the text responses by each account, and the action of outputting the text responses, is referred to as "speaking." "Speaking" includes the text input by the user.

[0227] The chatbot account processes human accounts' statements in the same way as its own. For example, the chatbot account inputs statements from other chatbot accounts and human accounts as prompts into its own language model and then outputs a response.

[0228] use Figure 11 Explanation of the processing of information providing device 10. Figure 11 This is a diagram illustrating the discussion. For example... Figure 11 As shown, the information providing device 10 will be discussed as the stages progress. Here, it is assumed that investor A, as the superior account, consists of three subordinate accounts: investor A-1, investor A-2, and investor A-3.

[0229] Investor A-1 is a bot account that prioritizes residential sales data. For example, Investor A-1's language model relies on residential sales data to output text about investment.

[0230] Investor A-2 is a bot account that prioritizes employment statistics. For example, Investor A-2's language model relies on employment statistics to output text about investing.

[0231] Furthermore, the entity of the robot account is a language model, and discussion is initiated by repeatedly inputting prompts into the language model through the information providing device 10. These prompts may be, for example, the language model's statements from the previous stage. The type of data emphasized in the statements can be set through the system prompts. As a result, the response tendencies of each robot account differ.

[0232] Investor A-3 is a human account. Investor A-3 corresponds to a user. Furthermore, Investor A-3's personality (what they value) can be arbitrary.

[0233] like Figure 11 As shown, in the first stage, investor A-1 states, "Because Tokyo's housing market is booming, let's invest in Tokyo real estate." At this point, information providing device 10 continues the stage.

[0234] In the second phase, based on the results of inputting investor A-1's statement from the first phase as cue words into the language model, investor A-2 stated, "But employment in Tokyo seems to be decreasing, so there may be risks."

[0235] Next, in the third stage, a human account says, "I heard the population living in Osaka is increasing." The human account's statement is content entered by the user. For example, the information providing device 10 can also display the statements of each account on the user terminal through the chat UI and accept user input from the input field of the chat UI.

[0236] Furthermore, each account, including both bot and human accounts, is not required to speak at all stages. The information providing device 10 only needs to allow at least one account to speak at each stage.

[0237] Furthermore, when the accounts participating in the discussion include human accounts, it is difficult to get users to speak every single time across the sometimes billions of stages. Therefore, for example, the information providing device 10 may also accept user speech only in specific stages, such as once out of every 10,000 stages.

[0238] Furthermore, the information providing device 10 can also accept user comments under certain conditions. These conditions include the bot account repeating the same content a certain number of times. By having the bot account use the user's comments as prompts, it is hoped that a stalled discussion will begin to move towards a conclusion.

[0239] In this way, if the answer meets the conditions, the information providing device 10 prompts the user with the answer. Then, the information providing device 10 collects the text entered in response to the prompt.

[0240] Furthermore, sometimes users possess knowledge that was not used when the language model of the bot account was learning. In such cases, including a human account may yield more useful conclusions in the discussion compared to cases with only a bot account.

[0241] use Figure 12 Explain the steps involved in the discussion. Figure 12 This is a flowchart illustrating the processing steps involved in the discussion. For example... Figure 12 As shown, at the start of a phase, the information providing device 10 determines whether to allow a human account to speak (step S201). The conditions for allowing a human account to speak are that a predetermined number of phases have passed, or that the robot account has repeated the same statement a certain number of times.

[0242] If the information providing device 10 does not allow a human account to speak (step S202, No), it proceeds to step S204. On the other hand, if the information providing device 10 allows a human account to speak (step S202, Yes), it collects the human account's speech (step S203). For example, the information providing device 10 collects the text entered by the user into the input field of the chat UI as the human account's speech. Furthermore, the information providing device 10 may also prompt the user about each account's speech before step S203.

[0243] Next, the information providing device 10 collects the messages from the robot account (step S204). Here, if the discussion has ended (step S205, yes), the information providing device 10 performs discussion ending processing (step S208). For example, the information providing device 10 may determine that the discussion has ended after a predetermined number of stages or after a ruling account has made a ruling. Discussion ending processing refers to transferring the discussion to a higher-level account (e.g., ...). Figure 1 Step S3).

[0244] If the discussion has not ended (step S205, no), the information providing device 10 continues the phase (step S206). Then, based on the collected statements, prompt words are input into the language models of each robot account (step S207), and then the process returns to step S201 to proceed with the entered phase.

[0245] 6. Sixth Implementation Method Next, the sixth embodiment will be described. Figure 13 This is a diagram illustrating the outline of the information providing device according to the sixth embodiment.

[0246] like Figure 13 As shown, the information providing system 1 includes an information providing device 10 and a database 20. The database 20 is a database that stores various types of data. Furthermore, the database 20 can also consist of multiple databases, and can also be configured within the information providing device 10.

[0247] The information providing device 10 is a device that connects to various devices containing the database 20 via plug-ins. These devices include various servers, vehicles, smartphones, and all other similar devices. That is, data is sent to the database 20 in real time from all these devices, including servers, vehicles, and smartphones, and this data is stored.

[0248] In this embodiment, the information providing device 10 is connected to the database 20 via a plug-in (steps S1-5), enabling the language model to learn based on the data stored in the database 20 (steps S2-5).

[0249] As mentioned above, since the data stored in database 20 is frequently updated, language models that learn using the various data stored in database 20 are able to generate answers based on the latest data.

[0250] In this embodiment, suggestions are made regarding the use of language models. Specifically, such as... Figure 13 As shown, the information providing device 10 inputs prompt words into the language model, sets up multiple accounts in the language model, and then conducts discussions through the multiple accounts (steps S3-5).

[0251] existFigure 13 The example shown represents a scenario where three accounts, A through C, are set up. Furthermore, for example, as... Figure 13 As shown, accounts A through C can each be composed of multiple accounts. After discussing each account, the conclusions for accounts A through C are derived.

[0252] Next, using Figure 14 Provide a summary of the discussions for each account. Figure 14 This diagram represents a summary of the discussions involved in the implementation method. In this implementation, a method for conducting discussions among accounts using a language model more efficiently is proposed.

[0253] Specifically, such as Figure 14 As shown, first, Account A states its claim as Account A on a specific issue and seeks the opinion of Account B (step S11). Account B, having been consulted by Account A, states its claim as Account B and seeks the opinion of Account C (step S12).

[0254] At this point, Account B agrees with the parts of Account A's claim that it can agree with, after stating its reasons, and states its claim as Account B from a different perspective than Account A.

[0255] Account C, having been consulted by Account B, then seeks the opinion of Account A after stating its claim as Account C (step S13). At this point, Account C agrees with the parts of the claims made by Account A and Account B that it can agree with, stating its reasons, and presents its claim as Account C from a viewpoint different from that of Account A or Account B.

[0256] For example, account C develops logic and expresses agreement with account A or account B, or both, regarding the parts that account A or B can agree on. In cases where there is a claim from a completely different perspective than that which can be agreed on, account C presents its own claim.

[0257] Furthermore, without asserting a different perspective, Account C explicitly stated which opinion it agreed with, that of Account A or Account B.

[0258] Subsequently, Account A, after stating its claims regarding Account B and Account C, seeks Account B's opinion. At this point, Account A agrees with the parts of Account B or Account C's claims that it can agree with, that is, it provides further logical explanations of Account A's original claims, or presents its claims from a different perspective than Account B or Account C.

[0259] Through this cycle of discussion, accounts A through C, in the process of making logically explainable claims from their respective different viewpoints, agree with the claims of other accounts when they no longer have any logically explainable claims.

[0260] Then, ultimately, all accounts agree on the claim of one account. Furthermore, in this embodiment, the state where all accounts agree is referred to as a state of consensus. Then, the information providing device 10 provides users, etc., with information about such a discussion (articles, etc., regarding the responses of each account). For example, by determining their opinions based on information about such a discussion, users can avoid conflict with others due to differing claims.

[0261] In this way, in the information providing device 10 involved in the implementation method, each account presents its opinions on the issue from multiple perspectives and continues to discuss until a consensus is finally reached.

[0262] Therefore, the information providing device 10 involved in the implementation can, for example, simulate discussions for reaching a consensus in any group holding different opinions, and thus can effectively apply language models.

[0263] Next, using Figure 15 This illustrates an example of prompt words input to a language model. Figure 15 This diagram illustrates an example of the prompt words involved in the sixth embodiment. For example... Figure 15 As shown, the prompts are textual information related to the preconditions of a discussion involving multiple accounts (here, 3 brilliant scientists).

[0264] like Figure 15 As shown, for example, Mr. A, a brilliant scientist, after stating his opinion, seeks Mr. B's opinion, saying, "This is what I think. What do you think, Mr. B?" Similarly, Mr. B, also a brilliant scientist, seeks Mr. C's opinion, saying, "This is what I think. What do you think, Mr. C?"

[0265] Similar to Mr. A and Mr. B, Mr. C also sought Mr. A's opinion after stating his own, saying, "This is what I think. What does Mr. A think?" At this point, Mr. A through Mr. C either agreed with the other's opinion or stated an opinion different from the other's.

[0266] These discussions continue until no further disagreements arise, that is, until all accounts agree on one of Mr. A's to Mr. C's opinions, thus finally reaching a consensus.

[0267] Using these prompts, the various accounts set up in the language model express their respective claims and continue the discussion until a consensus is reached. Furthermore, in Figure 3The example illustrates a scenario where multiple accounts are designated as "genius scientists," but this is not a limitation. That is, arbitrary accounts can also be set up. In this case, the arbitrary accounts can be accounts with the same attributes or accounts with different attributes. Specifically, for example, assuming a discussion of a patient's medical policy, the accounts could be designated as doctors, patients, family members, etc. In this case, doctors could be further subdivided into surgeons, internists, dermatologists, etc.

[0268] In the information providing device 10, after multiple accounts in the language model that generates responses to input prompts each express their opinions, opinions are solicited from other accounts, and this process is repeated until a consensus is reached among the multiple accounts.

[0269] For example, the information providing device 10 inputs such as language model Figure 15 The prompts shown indicate that multiple accounts are set up in the language model, and discussions among these accounts can begin. These accounts can be set up by the user or by the language model. Similarly, the discussion topics can also be set by the user or by the language model.

[0270] Furthermore, for example, when setting up multiple accounts for the language model, the following discussion will cover steps 1 through 5. Additionally, the following explanation addresses the case where multiple doctor accounts are used in relation to actual patient medical policies.

[0271] Step 1: Set up a multi-functional virtual expert There are hundreds of combinations of virtual experts, for example, specific roles such as "Neurologist A", "Neurologist B", and "Neurologist C" for neurology, and "Cardiac Expert A", "Cardiac Expert B", and "Cardiac Expert C" for cardiology.

[0272] Step 2: High-speed information exchange These experts exchange information and engage in billions of debates per second, based on patient data. This allows them to instantly assess a patient's condition, symptoms, and treatment options.

[0273] Step 3: Data Processing Virtual experts advanced the discussion by sharing various types of information, including MRI data, genetic data, and blood data. This allowed them to quickly derive the optimal solution from multiple perspectives.

[0274] Step 4: Self-learning and updating When new knowledge, information, or treatments are discovered during the discussions, they are immediately reflected in database 20 and applied to subsequent discussions. This enables diagnoses and treatments to be provided based on the latest medical information.

[0275] Step 5: Consensus Formation Discussions, through the clash of differing opinions and viewpoints, culminate in a final "consensus." Consequently, when making recommendations for patient diagnosis and treatment, a comprehensive approach incorporating information and knowledge from multiple sources is adopted. Multiple treatment proposals can also be proposed.

[0276] These discussions can, for example, lead to a consensus among multiple physician accounts regarding patient care policies. Additionally, they can allow multiple accounts involved in patient care, such as patient accounts and family accounts, to participate in the discussions.

[0277] The information providing device 10 provides information about the above discussion to users and others. For example, the information providing device 10 provides answers output by a language model on behalf of each account as information about the discussion.

[0278] This allows users to confirm the content of the discussion and determine their opinions based on that content.

[0279] Next, using Figure 16 The processing steps performed by the information providing device 10 according to the sixth embodiment will be explained. Figure 16 This is a flowchart illustrating the processing steps performed by the information providing device 10 according to the sixth embodiment.

[0280] like Figure 16 As shown, the information providing device 10 first inputs prompt words into the language model (step S301). Then, the information providing device 10 uses the language model to conduct a discussion (step S302).

[0281] Next, the information providing device 10 determines whether a consensus has been reached (step S303). If the information providing device 10 determines that a consensus has been reached (step S303; yes), the process ends; if the information providing device 10 determines that a consensus has not been reached (step S303; no), it returns to the process in step S302 and repeats the discussion.

[0282] According to this embodiment, since a language model learned from data stored in a database is used for discussion until a consensus is reached, the language model can be applied effectively.

[0283] The present invention has been described above using embodiments, but the technical scope of the present invention is not limited to the scope described in the above embodiments. Those skilled in the art will clearly understand that various changes or modifications can be made to the above embodiments. It is evident from the claims that such changes or modifications can also be included within the technical scope of the present invention.

[0284] It should be noted that the execution order of actions, processes, steps, and stages in the apparatus, system, program, and method shown in the claims, specification, and drawings can be implemented in any order, as long as it is not specifically stated as "before" or "prior to," and as long as the output of the previous process is not used in subsequent processes. Regarding the flow of actions in the claims, specification, and drawings, even if "firstly," "nextly," etc., are used for convenience, it does not mean that they must be implemented in that order.

Claims

1. An information providing device, characterized in that, include: A discussion section, which facilitates discussions of legal cases through multiple lawyer accounts making different claims, configured within a language model. This language model is generated by learning from data stored in a database of legal cases and is used to produce responses to input prompts. The adjudication department, through an adjudication account set up in the language model, adjudicates the merits of the claims based on the discussion results of the discussion department.

2. The information providing device according to claim 1, wherein, The discussion section conducts the discussion in a manner that allows prosecutor accounts set up in the language model to participate in the discussion.

3. The information providing device according to claim 1, wherein, The adjudication department makes its rulings by referring to relevant laws or previous precedents stored in the database.

4. The information providing device according to claim 1, wherein, The discussion section addresses the necessity of litigation against the client company.

5. The information providing device according to claim 1, wherein, The discussion section discusses anticipated litigation between the client company and other companies, as well as contractual provisions for avoiding or winning litigation. The adjudication department prepares a draft of the contract content based on the adjudication results.

6. An information providing method, performed by an information providing device, characterized in that it comprises: The discussion process involves multiple lawyer accounts, each making different claims, set up within a language model to discuss legal cases. This language model is generated by learning from data stored in a database of legal cases and is used to generate responses to input prompts. The adjudication step, through an adjudication account set in the language model, adjudicates the merits of the claims based on the discussion results of the discussion step.

7. An information provider, characterized in that, It causes the computer to perform the following steps: The discussion process involves multiple lawyer accounts, each making different claims, set up within a language model to discuss legal cases. This language model is generated by learning from data stored in a database of legal cases and is used to generate responses to input prompts. The adjudication step, through an adjudication account set in the language model, adjudicates the merits of the claims based on the discussion results of the discussion step.

8. An information providing device, characterized in that, include: A discussion section, which facilitates discussions about accounting through multiple accountant accounts making different claims, set up in a language model. This language model is generated by learning from data stored in a database of accounting cases and is used to generate responses to input prompts. The adjudication department, through an adjudication account set up in the language model, adjudicates the merits of the claims based on the discussion results of the discussion department.

9. An information providing device, characterized in that, include: A discussion section, which facilitates discussions about patient treatment strategies through multiple doctor accounts set up in a language model, wherein the language model is generated by learning from data stored in a medical database and is used to generate responses to input prompts; and The proposal department, through the adjudication account set in the language model, adjudicates the discussion results of the discussion department and recommends the treatment policy.

10. An information providing device, characterized in that, include: A discussion section, which facilitates discussions about patient treatment strategies through multiple doctor accounts set up in a language model, wherein the language model is generated by learning from data stored in a medical database and is used to generate responses to input prompts; and The proposal department, through the adjudication account set up in this language model, adjudicates the discussion results of the discussion department and recommends the treatment strategy. The proposal department generates minutes of the discussion that led to the treatment policy, and provides the minutes together with the treatment policy.

11. An information providing device, characterized in that, include: A discussion section, which facilitates discussions about patient treatment strategies through multiple doctor accounts set up in a language model, wherein the language model is generated by learning from data stored in a medical database and is used to generate responses to input prompts; and The proposal department, through the adjudication account set up in this language model, adjudicates the discussion results of the discussion department and recommends the treatment strategy. The discussion section restricts the language model's access to the data within the discussion.

12. An information providing device, characterized in that, include: The collection unit collects responses from multiple language models that output answers about the patient's treatment strategy based on input prompts, as well as text input by the user, wherein the multiple language models are configured to have different response tendencies. The prompting unit will display the collected answers to the user; and The output control unit inputs the collected answers and text as prompt words into the multiple language models, so that the multiple language models output the answers again.

13. An information providing device, characterized in that, include: A discussion section, which facilitates discussions about the company's business policies through multiple accounts of responsible business personnel set up in a language model, wherein the language model is generated by learning from data stored in a database about the company's activities and is used to generate responses to input prompts; and The proposal department, through the adjudication account set up in the language model, adjudicates the discussion results of the discussion department and makes recommendations on the business policy.

14. An information providing device, characterized in that, include: A discussion forum, which uses multiple speaker accounts set up in a language model for discussing policy guidelines, wherein the language model is generated by learning from data stored in a database about politics and is used to generate responses to input prompts; and The Proposal Department, through the adjudication account set up in the language model, adjudicates the discussion results of the Discussion Department and recommends the policy guidelines.

15. An information providing device, characterized in that, include: A discussion forum, which facilitates discussions about investment strategies through multiple investor accounts set up in a language model, wherein the language model is generated by learning from data stored in a database about investments and is used to generate responses to input prompts; and The proposal department, through the adjudication account set up in the language model, adjudicates the discussion results of the discussion department and proposes the investment policy.

16. An information providing device, characterized in that, include: A discussion forum, which facilitates discussions about investment strategies through multiple investor accounts set up in a language model, wherein the language model is generated by learning from data stored in a database about investments and is used to generate responses to input prompts; and The proposal department, through the adjudication account set up in this language model, adjudicates the discussion results of the discussion department and recommends the investment strategy. The proposal department generates a minutes of discussion that yielded the results of the investment policy, and provides the minutes together with the investment policy.

17. An information providing device, characterized in that, include: A discussion forum, which facilitates discussions about investment strategies through multiple investor accounts set up in a language model, wherein the language model is generated by learning from data stored in a database about investments and is used to generate responses to input prompts; and The proposal department, through the adjudication account set up in this language model, adjudicates the discussion results of the discussion department and recommends the investment strategy. The discussion section restricts the language model's access to the data within the discussion.

18. An information providing device, characterized in that, include: The discussion section, which, after multiple accounts in a language model used to generate responses to input prompts have expressed their opinions, solicits opinions from other accounts, and repeats this discussion until a consensus is reached among the multiple accounts; and The provision department provides information about the content of the discussions conducted in the discussion department.

Images