Program, information processing device, and method

Abstract

We provide technology that uses AI agents to support patent searches. [Solution] The program causes the computer to perform the following processes: acquiring technical information including technical ideas; acquiring the purpose of the patent search; inputting the technical information and the purpose of the search into a search design AI agent that has been instructed in advance to output search design information that serves as a guideline for the patent search; and acquiring search design information related to the technical information based on the results obtained from the search design AI agent by inputting the technical information and the purpose of the search into the search design AI agent.

Description

Technical Field

[0001] This disclosure relates to a program, an information processing apparatus, and a method.

Background Art

[0002] In recent years, technologies related to AI (Artificial Intelligence) agents have become widespread. An AI agent is software that autonomously selects the optimal means to achieve a specific goal and performs a task.

[0003] In recent years, various open sources for developing AI agents have been publicly disclosed. Examples of such open sources include AutoGen, CrewAI, and Google Agent Development Kit (ADK).

[0004] Non-Patent Document 1 discloses how to use Google ADK and implementation examples of AI agents.

Prior Art Documents

Non-Patent Documents

[0005]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] Although AI agents can be applied to various services, there are still few examples of being realized as specific services. In this regard, a technology for assisting patent investigations using AI agents is desired.

Means for Solving the Problems

[0007] In one example of this disclosure, a program is provided. The program causes a computer to perform the following processes: acquiring technical information including technical ideas; acquiring search objectives related to patent searches; inputting the technical information and the search objectives into a search design AI agent that has been pre-specified to output search design information that serves as a guideline for patent searches; and acquiring the search design information related to the technical information based on the results obtained from the search design AI agent by inputting the technical information and the search objectives into the search design AI agent.

[0008] In one example of this disclosure, the instructions specified for the above-mentioned research and design AI agent include instructions to have the input invention content analyzed from a specific viewpoint, and instructions to have the analysis results output as the above-mentioned research and design information.

[0009] In one example of this disclosure, the particular viewpoint described above includes at least one of the following: the technical viewpoint, the problem viewpoint, the effect viewpoint, the feature viewpoint of the invention, the component viewpoint, and the patent classification viewpoint.

[0010] In one example of this disclosure, the instructions specified for the above-mentioned survey design AI agent include instructions to output survey design information in a predetermined format.

[0011] In one example of this disclosure, the type of large-scale language model used by the above-mentioned survey design AI agent is specified.

[0012] In one example of this disclosure, the above-mentioned survey and design AI agent has a program function defined as a tool. This program function is a function that acquires the above-mentioned technical information.

[0013] In one example of this disclosure, the program further inputs the search design information to a patent search AI agent that has been pre-specified to perform processing related to patent searches, and obtains the results related to the patent search based on the results obtained from the patent search AI agent by inputting the search design information to the patent search AI agent.

[0014] In one example of this disclosure, the patent search AI agent includes a patent search AI agent that is pre-specified to perform a patent search. The program further inputs the search design information to the computer and to the patent search AI agent, and obtains a set of patent documents to compare with the technical information based on the results obtained from the patent search AI agent by inputting the search design information to the patent search AI agent.

[0015] In one example of this disclosure, the program further causes the computer to perform a process to obtain a set of patent documents that are the subject of the patent search. The patent search AI agent includes a screening AI agent which is pre-specified to remove less relevant patent documents. The program further causes the computer to perform a process to input the search design information and the set of patent documents to the screening AI agent, and a process to obtain a set of patent documents after removing those less relevant to the technical information, based on the results obtained from the screening AI agent by inputting the search design information and the set of patent documents to the screening AI agent.

[0016] In one example of this disclosure, the program further causes the computer to perform a process to obtain a set of patent documents that are the subject of the patent search. The patent search AI agent includes a relevance evaluation AI agent which is instructed in advance to evaluate relevance. The program further causes the computer to perform a process to input the search design information and the set of patent documents to the relevance evaluation AI agent, and a process to obtain an evaluation result showing the relevance between each of the patent documents and the technical information based on the results obtained from the relevance evaluation AI agent by inputting the search design information and the set of patent documents to the relevance evaluation AI agent.

[0017] In one example of this disclosure, the program further causes the computer to perform the following processes: inputting the above-mentioned survey design information and the above-mentioned evaluation results into a report generation AI agent that has been instructed in advance to output data in a predetermined format; and obtaining a patent search report relating to the above-mentioned technical information based on the results obtained from the report generation AI agent by inputting the above-mentioned survey design information and the above-mentioned evaluation results into the report generation AI agent.

[0018] In one example of this disclosure, the patent search AI agent consists of a plurality of AI agents, each instructed to perform a different process related to patent searching. The process involves inputting the technical information and the search objective to the search design AI agent and a root AI agent, which is instructed to perform a process to at least one of the plurality of AI agents. Based on the results obtained from the root AI agent by inputting the technical information and the search objective to the root AI agent, the search design AI agent and at least one of the plurality of AI agents are instructed to perform a process.

[0019] In one example of the present disclosure, the instructions specified for the root AI agent include instructions to select an AI agent according to the above investigation purpose and to execute processing on the AI agent.

[0020] In one example of the present disclosure, the program further causes the computer to execute processing to receive the target number of cases for the patent investigation and processing to repeatedly perform a patent search on the patent search AI agent until the number of hits of patent documents by the patent search AI agent falls within a target range based on the target number of cases.

[0021] In one example of the present disclosure, the patent search AI agent is specified with an instruction to change the search conditions so as to narrow the search range when the number of hits exceeds the upper limit of the target range.

[0022] In one example of the present disclosure, the patent search AI agent is specified with an instruction to change the search conditions so as to widen the search range when the number of hits is below the lower limit of the target range.

[0023] In one example of the present disclosure, the program further causes the computer to execute processing to receive a setting regarding whether approval of the output result of the investigation design AI agent is required and, when the setting indicates that approval is required, display an approval request screen for receiving whether to approve the investigation design information based on the generation of the investigation design information.

[0024] In one example of the present disclosure, the approval request screen is configured to receive an operation instructing approval and an operation instructing redoing. The program further causes the computer to output the investigation design information obtained in the obtaining process when an operation instructing approval is received, and execute a process of causing the investigation design AI agent to regenerate the investigation design information again when an operation instructing redoing is received.

[0025] In an example of the present disclosure, the approval request screen is further configured to receive an operation for instructing editing. When the computer receives an operation for instructing the editing, the program further causes the computer to receive an editing operation of the survey design information and execute a process of outputting the edited survey design information.

[0026] In an example of the present disclosure, the program further causes the computer to receive a setting regarding whether approval of the output result of the patent survey AI agent is required, and when the setting indicates that approval is required, display an approval request screen for receiving whether to approve the result based on the fact that the result related to the patent survey has been generated.

[0027] In another example of the present disclosure, an information processing apparatus is provided. The information processing apparatus includes a control unit. The control unit performs a process of acquiring technical information including a technical idea, a process of acquiring a survey purpose related to a patent survey, a process of inputting the technical information and the survey purpose to a survey design AI agent that is instructed in advance to output survey design information serving as a guideline for the patent survey, and a process of acquiring the survey design information related to the technical information based on the result obtained from the survey design AI agent by inputting the technical information and the survey purpose to the survey design AI agent.

[0028] In another example of the present disclosure, a method executed by a computer is provided. The method includes steps of acquiring technical information including a technical idea, acquiring a survey purpose related to a patent survey, inputting the technical information and the survey purpose to a survey design AI agent that is instructed in advance to output survey design information serving as a guideline for the patent survey, and acquiring the survey design information related to the technical information based on the result obtained from the survey design AI agent by inputting the technical information and the survey purpose to the survey design AI agent.

[0029] The above and other objects, features, aspects and advantages of the present invention will become apparent from the following detailed description relating to the invention, which will be understood in conjunction with the accompanying drawings. [Brief explanation of the drawing]

[0030] [Figure 1] This figure shows an example of the device configuration of an information processing system. [Figure 2] This is a diagram illustrating the basic structure of an AI agent. [Figure 3] This is a diagram illustrating an example of AI agent coding. [Figure 4] This diagram illustrates the function of generating survey design information using an AI agent. [Figure 5] This is a schematic diagram showing an example of the hardware configuration of an information processing device. [Figure 6] This is a schematic diagram showing an example of the hardware configuration of a user terminal. [Figure 7] This figure shows an example of the application of the survey design agent. [Figure 8] This figure shows an example of a screen for setting technical information and research objectives. [Figure 9] This figure shows an example of survey design information. [Figure 10] This figure shows an example of the results of the relevance evaluation. [Figure 11] This is a diagram showing the first half of the report. [Figure 12] This is a diagram showing the latter half of the report. [Figure 13] This diagram shows the settings screen according to the modified example. [Figure 14] This figure shows an example of an execution screen that is displayed while an AI agent is performing a process. [Figure 15] This figure shows another example of an AI agent configuration. [Figure 16] This figure shows an example of a code definition related to the root agent. [Figure 17] This figure shows an example of code definition related to the initialization agent. [Figure 18] It is a diagram showing an example of code definition related to an investigation design agent. [Figure 19] It is a diagram showing an example of code definition related to a patent search agent. [Figure 20] It is a diagram showing an example of code definition related to an extraction agent. [Figure 21] It is a diagram showing an example of code definition related to a condition generation agent. [Figure 22] It is a diagram showing an example of code definition related to a search execution agent. [Figure 23] It is a diagram showing an example of code definition related to an optimization agent. [Figure 24] It is a diagram showing an example of code definition related to a screening agent. [Figure 25] It is a diagram showing an example of code definition related to a relevance evaluation agent. [Figure 26] It is a diagram showing an example of code definition related to an iterative review agent. [Figure 27] It is a diagram showing an example of code definition related to a quality evaluation agent. [Figure 28] It is a diagram showing an example of code definition related to a report generation agent. [Figure 29] It is a diagram showing another example of code definition related to a root agent.

Modes for Carrying Out the Invention

[0031] Hereinafter, each embodiment according to the present invention will be described with reference to the drawings. In the following description, the same parts and components are denoted by the same reference numerals. Their names and functions are also the same. Therefore, detailed descriptions thereof will not be repeated. In addition, each embodiment and each modification described below may be selectively combined as appropriate.

[0032] <A. Information Processing System 10> First, the device configuration of the information processing system 10 will be described with reference to Figure 1. Figure 1 is a diagram showing an example of the device configuration of the information processing system 10.

[0033] As shown in Figure 1, the information processing system 10 includes an information processing device 100, a user terminal 200, and a server 300. The information processing device 100, the user terminal 200, and the server 300 are configured to communicate with each other via a network NW (for example, the Internet).

[0034] The information processing device 100 is a notebook or desktop PC (Personal Computer), a tablet device, a smartphone, or another computer with communication capabilities. The number of information processing devices 100 constituting the information processing system 10 may be one or two or more. The information processing device 100 is operated, for example, by company "A".

[0035] The user terminal 200 is, for example, a notebook or desktop PC, a tablet device, a smartphone, or another computer with communication capabilities. The number of user terminals 200 constituting the information processing system 10 may be one or two or more. The user terminal 200 is owned, for example, by a general user, user "A".

[0036] Server 300 is a notebook or desktop PC (Personal Computer), a tablet device, a smartphone, or another computer with communication capabilities. The number of servers 300 constituting the information processing system 10 may be one or two or more. Server 300 is operated, for example, by company "B".

[0037] Server 300 stores the Large Language Model (MD). The Large Language Model (MD) is a language model trained on billions of text data points and is capable of processing a wide range of natural languages. The Large Language Model (MD) is also known as an LLM (Large Language Model). The Large Language Model (MD) is trained to generate output corresponding to a given prompt.

[0038] Examples of large-scale language models (MDs) include the GPT series such as GPT-4 (Generative Pretrained Transformer) and GPT-5, the Gemini series, LLaMA (Large Language Model Meta AI), and known LLMs. In addition to the GPT series, various other large-scale language models may be used, such as Transformer-based large-scale language models like BERT (Bidirectional Encoder Representations from Transformers). Large-scale language models may also include visual language models capable of processing images in addition to text, and further, language models capable of processing speech and video may also be used.

[0039] Company "B" has, for example, published a framework for utilizing the functionality of the large-scale language model (MD). This framework could include, for example, an API (Application Programming Interface). By using this framework, designers at company "A" can provide various functions using the large-scale language model (MD) to general users "A". For example, designers at company "A" can use this framework to design AI (Artificial Intelligence) agents.

[0040] In addition, the various processes described in this specification may be implemented in the information processing apparatus 100, may be implemented in the user terminal 200, may be implemented in the server 300, or may be implemented in other computers.

[0041] Also, in the above description, it was assumed that the large language model MD is provided with a framework by Company "B", but the API may be provided by Company "A" itself.

[0042] Also, in the above description, an example in which the information processing system 10 includes the server 300 has been described, but the information processing system 10 may not include the server 300. In this case, the information processing system 10 is composed of one or more information processing apparatuses 100 and one or more user terminals 200.

[0043] <Overview of AI Agent> Next, referring to FIG. 2, an overview of the AI agent will be described. FIG. 2 is a diagram for explaining the outline of the AI agent.

[0044] An AI agent is software that autonomously repeats and executes the collection, judgment, execution, and verification of necessary information. Hereinafter, for the sake of convenience of explanation, the AI agent is also simply referred to as an "agent".

[0045] An instruction is预先指定 for the agent AG. The "instruction" indicates the operation policy of the agent AG and is one of the arguments that can be specified when defining the agent AG using an open-source agent development framework program. The designer of Company "A" specifies an arbitrary instruction for the argument according to the service to be realized. Thereby, the agent AG autonomously uses the large language model MD and the tool TL according to the specified instruction and executes a task corresponding to a query from the user terminal 200.

[0046] The large language model MD provides the agent AG with language inference capabilities. The language inference capabilities here include natural language understanding, natural language summarization, natural language generation, hypothesis formulation, plan creation, complementation of ambiguous information, creation of explanatory texts, etc. The agent AG calls the large language model MD in situations where it decides "what the agent AG should do" and "what operation the agent AG should perform next" in order to achieve its goal. Then, the agent AG uses the output from the large language model MD for decision-making and input generation for the next step.

[0047] Note that the large language model MD used by the agent AG may be stored in a server 300 different from the information processing device 100, or may be stored in the information processing device 100 as a local LLM. Also, the large language model MD used by the agent AG may be one, or may be two or more.

[0048] The tool TL is a program function. The tool TL may be a local function, a library function, or an API that calls a function provided by an external company. By defining the tool TL for the agent AG, the agent AG autonomously selects the necessary tool TL and executes the selected tool TL. The tool TL to be executed is appropriately selected according to the query from the user.

[0049] Note that the tool TL does not necessarily have to be defined. Also, the tool TL defined for the agent AG may be one, or may be two or more.

[0050] <C. Framework Specification> Next, referring to FIG. 3, the framework specification for defining an AI agent will be described. FIG. 3 is a diagram for explaining a coding example of an AI agent.

[0051] Figure 3 shows an example of coding using the ADK (Agent Development Kit) provided by Google. The ADK is a framework made publicly available for the efficient development of AI agents.

[0052] The program code PR shown in Figure 3 defines agent AGX, which responds with the capital of the queried country. More specifically, program code PR has code 30, which is the definition part of agent AGX.

[0053] Code 30 has argument sections 31A to 31F. Argument sections 31A to 31F are defined as interface specifications in ADK. Designers can develop AI agents by defining codes for some or all of argument sections 31A to 31F.

[0054] The argument 31A, which is "name", accepts the definition of the name to be assigned to the agent AGX. In the example in Figure 3, the name "capital_agent" is specified in argument 31A.

[0055] The argument 31B, which is "model", accepts the setting for the large-scale language model MD used by agent AGX. In the example in Figure 3, the large-scale language model "gemini-2.0-flash" is specified in argument 31B.

[0056] The argument 31C, "description," accepts a definition of a description of the agent AGX's functions. This description is referenced, for example, when other agents determine the functions of agent AGX. Other agents refer to the description specified for agent AGX to decide whether or not to request a task from agent AGX. In the example in Figure 3, the function of agent AGX is specified as the function of answering the capital of the queried country.

[0057] The argument 31D, labeled "instruction," accepts a definition of instructions for agent AGX. For example, the argument 31D may specify the process to be executed, the execution order of that process, and the format of the output. The instructions specified in argument 31D do not necessarily have to be explicit; they may be abstract. Even if abstract instructions are specified in argument 31D, the AI ​​agent will autonomously interpret these instructions and repeatedly perform the necessary information gathering, judgment, execution, and verification.

[0058] The argument 31E for "tools" accepts the definition of a tool TL. As mentioned above, a tool TL is a program function. In the example in Figure 3, the local function "get_capital_city" is specified as tool TL1. Tool TL1 is a function that takes a country as an argument and returns the capital city.

[0059] The argument 31F for "sub_agents" accepts the definition of subordinate AI agents of agent AGX. In the example in Figure 3, "Agent A" and "Agent B" are specified as examples of subagents.

[0060] In addition to the ADK provided by Google, various other frameworks exist for developing AI agents. Other frameworks include AutoGen, OpenAI Agents SDK, LangChain, LangGraph, CrewAI, LlamaIndex, Semantic Kernel, and Haystack.

[0061] In various frameworks, arguments corresponding to the argument sections 31A to 31F described above are defined as interface specifications. Designers can develop AI agents by specifying parameters for each argument.

[0062] For example, regarding AutoGen, the API "AssistantAgent()" is publicly available as an interface specification for defining an AI agent. The API has an argument "name" corresponding to the argument part 31A, an argument "model_client" corresponding to the argument part 31B, an argument "system_message" corresponding to the argument part 31D, and an argument "tools" corresponding to the argument part 31E.

[0063] Regarding the OpenAI Agents SDK, the API "Agent()" is publicly available as an interface specification for defining an AI agent. The API has an argument "name" corresponding to the argument part 31A, an argument "model" corresponding to the argument part 31B, an argument "instructions" corresponding to the argument part 31D, and an argument "tools" corresponding to the argument part 31E.

[0064] <D. Overview> The information processing device 100 provides various analysis functions related to patent documents to the user "A". The analysis functions are realized using an AI agent.

[0065] As an example, the information processing device 100 provides a function of receiving an input of an investigation purpose related to a patent search and technical information of an investigation target, and generating investigation design information that serves as a guideline for the patent search.

[0066] The technical information is, for example, data including a technical idea. The technical idea means a technical means for solving a technical problem. As an example, the technical information is described by a character string of technical means. The technical information includes, for example, invention information in which invention-specific matters are described.

[0067] The technical information described above is, for example, patent documents. Patent documents include, for example, published patent gazettes, patent gazettes, published patent gazettes, republished patent gazettes, and utility model gazettes. As an example, a patent document consists of bibliographic information, a specification, claims, drawings, and an abstract. In addition to text, a patent document may also include drawings, tables, chemical structural formulas, mathematical formulas, gene sequences, etc. Bibliographic information may include, for example, the application number, publication number, patent registration number, application date, publication date, registration date, applicant, patentee, title of invention, agent, and country of application. Technical information may also be non-patent documents such as published technical reports. Furthermore, the technical information described above may also be, for example, text describing the components, steps, or features that constitute a product or service.

[0068] Figure 4 is a diagram illustrating the function of generating survey design information 130 using an AI agent. The overview of this generation function will be described below.

[0069] The information processing device 100 acquires technical information 123 from the user terminal 200. The technical information 123 may be a string entered by user "A" of the user terminal 200, or it may be data extracted from patent documents in response to user "A"'s operations.

[0070] Furthermore, the information processing device 100 obtains the search objectives 126 related to patent searches from the user terminal 200. Examples of search objectives 126 include prior art searches, invalidation searches, infringement prevention searches, and technology trend searches. The search objectives 126 may be strings entered by user "A" of the user terminal 200, or they may be search objectives selected from a predetermined list of options.

[0071] Subsequently, the information processing device 100 inputs the technical information 123 and the research objective 126 to the research and design agent AG10. The research and design agent AG10 is pre-specified with instruction 40. Instruction 40 is a parameter specified for the argument unit 31D (see Figure 3) described above. Instruction 40 is programmed to output research and design information that serves as a guideline for patent research to the argument unit 31D.

[0072] The survey design agent AG10 autonomously executes the optimal means to achieve the objective based on the instruction 40, based on the input of technical information 123 and survey objective 126. The information processing device 100 inputs the technical information 123 and survey objective 126 to the survey design agent AG10 and, based on the results obtained from the survey design agent AG10, acquires survey design information 130 related to the technical information 123.

[0073] The destination for the acquired survey design information 130 is arbitrary. For example, the destination is the user terminal 200. The survey design information 130 output to the user terminal 200 will be displayed on the user terminal 200's screen, for example. As another example, the destination may be an AI agent other than the survey design agent AG10.

[0074] As described above, the information processing device 100 generates search design information 130 related to the patent search of the technical information 123 by utilizing the search design agent AG10. In some cases, the user modifies the content of the search design information 130. When the search design information 130 is input to another AI agent that performs a patent search, if the search design information 130, which serves as a guideline for the patent search, is accurate, the accuracy of subsequent patent searches will be greatly improved. In this way, the information processing device 100 can support patent searches.

[0075] Preferably, instruction 40 specifies that the input invention content be analyzed from a predetermined viewpoint. This viewpoint includes, for example, a technical field viewpoint. As a result, the survey and design agent AG10 includes the technical field to which the invention described in technical information 123 belongs in the survey and design information 130.

[0076] As yet another example, the analytical perspective specified in instruction 40 includes the perspective of the problem. This causes the survey design agent AG10 to include the problem that the invention described in technical information 123 aims to solve in the survey design information 130.

[0077] As yet another example, the perspective of analysis specified in instruction 40 includes the perspective of effects. This causes the survey design agent AG10 to include the effects of the invention described in technical information 123 in the survey design information 130.

[0078] As yet another example, the perspective of analysis specified in instruction 40 includes the perspective of the features of the invention. This causes the survey design agent AG10 to include the features of the invention described in technical information 123 in the survey design information 130.

[0079] As yet another example, the perspective of analysis specified in instruction 40 includes the perspective of the components of the invention. The components represent the requirements that constitute the invention. This causes the research and design agent AG10 to divide the invention described in technical information 123 into components and include the divided components in research and design information 130.

[0080] As yet another example, the analytical perspectives specified in instruction 40 include patent classification perspectives. This causes the search design agent AG10 to include information regarding the patent classification to which the invention described in technical information 123 belongs in the search design information 130. Examples of such patent classifications include IPC (International Patent Classification), FI (File Index), and F-terms.

[0081] Preferably, the instruction 40 includes an instruction to include the input investigation purpose in the investigation design information 130. Thereby, the investigation design agent AG10 will include information regarding the investigation purpose 126 in the investigation design information 130.

[0082] Alternatively, the instruction 40 includes an instruction to output an investigation design instruction from the perspective of the investigation design corresponding to the input investigation purpose. Thereby, the investigation design agent AG10 will generate the investigation design information 130 corresponding to the investigation purpose 126. As an example, if the investigation purpose 126 is a validity investigation, the investigation design agent AG10 will autonomously judge the investigation purpose 126, for example, divide the invention described in the technical information 123 into constituent elements and include the constituent elements in the investigation design information 130. On the other hand, if the investigation purpose 126 is a technology trend investigation, the investigation design agent AG10 will autonomously judge the investigation purpose 126 and, for example, include the features of the invention described in the technical information 123 in the investigation design information 130.

[0083] Note that the analysis perspectives specified by the instruction 40 do not necessarily need to include all of the perspectives of summary, technical field, problem, effect, features of the invention, constituent elements, and patent classification. That is, the analysis perspectives specified by the instruction 40 include at least one of these perspectives.

[0084] Also, the investigation design agent AG10 does not necessarily need to be composed of one AI agent and may be composed of a plurality of AI agents. In this case, the plurality of AI agents will generate the investigation design information 130 by cooperating with each other.

[0085] Furthermore, the investigation design agent AG10 may be configured as a part of the functions included in the patent investigation agent AG20 described later.

[0086] <E. Hardware Configuration> Next, referring to FIGS. 5 and 6, the hardware configurations of the information processing apparatus 100 and the user terminal 200 shown in FIG. 1 above will be described in order.

[0087] The hardware configuration of the server 300 shown in Figure 1 is the same as that of the information processing device 100, so its explanation will not be repeated.

[0088] (E1. Information processing device 100) First, with reference to Figure 5, the hardware configuration of the information processing device 100 shown in Figure 1 will be explained. Figure 5 is a schematic diagram showing an example of the hardware configuration of the information processing device 100.

[0089] The information processing device 100 includes a control device 101, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, a communication interface 104, a display interface 105, an input interface 107, and an auxiliary storage device 120. These components are connected to a bus 110.

[0090] The control device 101 is comprised of, for example, at least one integrated circuit. The integrated circuit may consist of, for example, at least one CPU (Central Processing Unit), at least one GPU (Graphics Processing Unit), at least one ASIC (Application Specific Integrated Circuit), at least one FPGA (Field Programmable Gate Array), or a combination thereof.

[0091] The control device 101 controls the operation of the information processing device 100 by executing various programs such as the analysis program 122 and the operating system. Based on the receipt of execution commands for various programs, the control device 101 reads the program from the auxiliary storage device 120 or ROM 102 into the RAM 103. The RAM 103 functions as working memory and temporarily stores various data necessary for the execution of various programs.

[0092] The communication interface 104 is connected to a LAN (Local Area Network), an antenna, and other devices. The information processing device 100 exchanges data with external devices via the communication interface 104. These external devices include, for example, a user terminal 200, a server 300, and other communication devices.

[0093] A display 106 is connected to the display interface 105. The display interface 105 sends image signals to the display 106 for displaying images, in accordance with commands from the control device 101 or the like. The display 106 is, for example, a liquid crystal display, an organic EL (Electro-Luminescence) display, or other display device. The display 106 may be configured integrally with the information processing device 100, or it may be configured separately from the information processing device 100.

[0094] An input device 108 is connected to the input interface 107. The input device 108 is, for example, a mouse, keyboard, touch panel, or other device capable of receiving user input. The input device 108 may be configured integrally with the information processing device 100, or it may be configured separately from the information processing device 100.

[0095] The auxiliary storage device 120 is, for example, a hard disk, flash memory, SSD (Solid State Drive), or other storage medium. The auxiliary storage device 120 stores the analysis program 122 and the patent database 124, etc. The analysis program 122 includes AI agents such as the aforementioned research and design agent AG10. The patent database 124 includes multiple patent documents 125. The storage location of the analysis program 122 and the patent database 124 is not limited to the auxiliary storage device 120, but may also be stored in the storage area of ​​the control device 101 (for example, cache memory), ROM 102, RAM 103, external devices, etc.

[0096] Furthermore, the analysis program 122 may be provided not as a standalone program, but incorporated as part of any other program. In this case, the various processes defined in the analysis program 122 are realized in cooperation with the other program. Even a program that does not include such modules does not deviate from the intent of the analysis program 122 according to this embodiment. Moreover, some or all of the functions provided by the analysis program 122 may be realized by dedicated hardware. Furthermore, the information processing device 100 may be configured in a form similar to a so-called cloud service, where at least one server executes a portion of the processing of the analysis program 122.

[0097] (E2. User terminal 200) Next, with reference to Figure 6, the hardware configuration of the user terminal 200 shown in Figure 1 will be described. Figure 6 is a schematic diagram showing an example of the hardware configuration of the user terminal 200.

[0098] The user terminal 200 includes a control unit 201, a ROM 202, a RAM 203, a communication interface 204, a display interface 205, an input interface 207, and an auxiliary storage device 220. These components are connected to a bus 210.

[0099] The control device 201 is comprised of, for example, at least one integrated circuit. The integrated circuit may consist of, for example, at least one CPU, at least one GPU, at least one ASIC, at least one FPGA, or a combination thereof.

[0100] The control device 201 controls the operation of the user terminal 200 by executing various programs such as the analysis program 222 and the operating system. Based on the receipt of execution commands for various programs, the control device 201 reads the program from the auxiliary storage device 220 or ROM 202 into the RAM 203. The RAM 203 functions as working memory and temporarily stores various data necessary for the execution of the program.

[0101] The communication interface 204 is connected to a LAN, an antenna, and the like. The user terminal 200 exchanges data with external devices via the communication interface 204. These external devices include, for example, an information processing device 100, a server 300, and other communication devices. The user terminal 200 may be configured to download an analysis program 222 from the information processing device 100.

[0102] A display 206 is connected to the display interface 205. The display interface 205 sends image signals to the display 206 for displaying images, in accordance with commands from the control device 201 or the like. The display 206 is, for example, a liquid crystal display, an organic EL display, or other display device. The display 206 may be configured integrally with the user terminal 200, or it may be configured separately from the user terminal 200.

[0103] An input device 208 is connected to the input interface 207. The input device 208 may be, for example, a mouse, keyboard, touch panel, or other device capable of receiving user input. The input device 208 may be configured integrally with the user terminal 200 or separately from the user terminal 200.

[0104] The auxiliary storage device 220 is, for example, a hard disk, flash memory, SSD (Solid State Drive), or other storage medium. The auxiliary storage device 220 stores the analysis program 222, etc. The analysis program 222 includes AI agents such as the survey and design agent AG10 described above. The storage location of the analysis program 222 is not limited to the auxiliary storage device 220, but may also be stored in the storage area of ​​the control device 201 (e.g., cache memory), ROM 202, RAM 203, external devices (e.g., a server), etc.

[0105] Note that the analysis program 222 may be provided by being incorporated into a part of an arbitrary program instead of being provided as a single program. In this case, various processes defined in the analysis program 222 are realized in cooperation with an arbitrary program. Even a program that does not include such a part of the module does not deviate from the gist of the analysis program 222 according to the present embodiment. Furthermore, some or all of the functions provided by the analysis program 222 may be realized by dedicated hardware. Furthermore, the user terminal 200 may be configured in a form such as a so-called cloud service in which at least one server executes a part of the processing of the analysis program 222.

[0106] <F. Patent Investigation Function> Next, referring to FIG. 7, an application example of the above-described investigation design agent AG10 will be described. FIG. 7 is a diagram showing an application example of the investigation design agent AG10.

[0107] The information processing apparatus 100 provides various functions by causing the investigation design agent AG10 to cooperate with other AI agents. As an example, the information processing apparatus 100 provides a patent investigation function by causing the investigation design agent AG10 to cooperate with the patent investigation agent AG20.

[0108] An instruction is specified to the patent investigation agent AG20 to execute processing related to patent investigation. The processing includes at least one of a patent search process, a screening process, a relevance evaluation process, and a report generation process. The information processing apparatus 100 inputs the investigation design information 130 to the patent investigation agent AG20. The information processing apparatus 100 obtains a result related to the patent investigation based on the result obtained from the patent investigation agent AG20 by inputting the investigation design information 130 to the patent investigation agent AG20.

[0109] Preferably, an instruction is specified to the patent investigation agent AG20 so that the patent search process, the screening process, the relevance evaluation process, and the report generation process are executed in order.

[0110] The patent search agent AG20 may consist of a single AI agent that performs all processing related to patent searching, or it may consist of multiple AI agents, each performing a different processing of patent searching.

[0111] In the example shown in Figure 7, the patent search agent AG20 consists of a patent search agent AG21 that handles patent search processing, a screening agent AG22 that handles screening processing, a relevance evaluation agent AG23 that handles relevance evaluation processing, and a report generation agent AG24 that handles report generation processing.

[0112] The patent search agent AG20 may include at least one of the following: the patent search agent AG21, the screening agent AG22, the relevance evaluation agent AG23, and the report generation agent AG24. For example, the patent search agent AG20 may include only the patent search agent AG21, only the screening agent AG22, only the relevance evaluation agent AG23, or only the report generation agent AG24.

[0113] The following sections will describe the investigation design agent AG10, the patent search agent AG21, the screening agent AG22, the relevance evaluation agent AG23, and the report generation agent AG24 in order.

[0114] (F1. Survey Design Agent AG10) First, let's explain the survey design agent AG10 shown in Figure 7.

[0115] As described above, the survey design agent AG10 receives the technical information 123 and the survey objective 126 as input and generates the survey design information 130 related to the technical information 123. Figure 8 shows an example of the setting screen 400A for the technical information 123 and the survey objective 126.

[0116] The settings screen 400A is displayed, for example, on the user terminal 200. The settings screen 400A includes, for example, an input field 410, an input field 412, an input field 414, an input field 416, and a start button 420.

[0117] Input field 410 accepts input of technical information 123. Input field 410 accepts input of text that describes a technical idea, for example. As an example, the user enters the content of an invention made by themselves or another company into input field 410. As another example, the user enters the specifications of their own product or a product of another company into input field 410.

[0118] Note that the technical information 123 does not necessarily need to be entered by the user. As another example, the settings screen 400A may be configured to accept the patent document number (for example, application number, publication number, or patent number) and to accept settings regarding the parts to be retrieved from the said patent document. Examples of configurable retrieval parts include the abstract, claims, title of the invention, technical field, background art, problem to be solved by the invention, effect of the invention, means for solving the problem, and specification.

[0119] The method for specifying the acquisition location is arbitrary. For example, pressing the expand button (not shown) will display a list of configurable acquisition location options. The user can then apply the acquisition location by selecting one of these options.

[0120] The information processing device 100 retrieves patent documents corresponding to the entered number and extracts portions from those patent documents corresponding to the set extraction locations. The information processing device 100 then displays the extraction results as technical information 123 in the input field 410. The user can arbitrarily edit the content reflected in the input field 410.

[0121] Input field 412 accepts input for the purpose of the investigation 126. Input field 412 accepts input of text that includes the purpose of the investigation 126. In the example in Figure 8, the text "I would like to conduct an invalidation search regarding the invention under investigation" is entered in input field 412, and the purpose of the investigation 126 related to the invalidation search is entered.

[0122] Furthermore, the investigation objectives 126 that can be entered in input field 412 are not limited to invalid document searches. Other examples include prior art searches, infringement prevention searches, or technology trend surveys, which may be entered as investigation objectives 126.

[0123] Furthermore, the survey objective 126 does not necessarily have to be entered as text. As another example, the survey objective 126 may be selected from a predetermined list of options. In this case, pressing the expand button (not shown) will display a list of selectable survey objectives as options. The user selects the survey objective 126 by performing a selection operation on one of these options.

[0124] Input field 414 accepts input for a target number of patent documents to be searched. Entering this target number is optional.

[0125] Input field 416 accepts input for the maximum cost of a patent search. This cost represents the maximum cost when using the large-scale language model (MD). Entering this cost is optional.

[0126] The start button 420 is a button used to instruct the user to begin a patent search. Based on the user pressing the start button 420, the user terminal 200 transmits the information entered into the settings screen 400A shown in Figure 8 to the information processing device 100.

[0127] The information processing device 100 inputs the information received from the user terminal 200 to the research and design agent AG10. This information includes at least technical information 123 and research objectives 126. Preferably, this information further includes at least one of the upper limit number of patent documents to be searched and the upper limit cost of the patent search.

[0128] As described above, the investigation design agent AG10 has a pre-specified instruction 40 (see Figure 4). Instruction 40 specifies that the input invention content should be analyzed from predetermined perspectives. These perspectives include, for example, the abstract perspective, the technical field perspective, the problem perspective, the effect perspective, the features perspective of the invention, the component perspective, the patent classification perspective, and the points of review of patent information during a patent search. Preferably, instruction 40 includes an instruction to output the input investigation objective 126.

[0129] As a result, the information processing device 100 acquires the survey and design information 130 shown in Figure 9 from the survey and design agent AG10. Figure 9 is a diagram showing an example of the survey and design information 130.

[0130] In the example shown in Figure 9, the survey design information 130 includes information 302 indicating the entered survey objective 126, information 304 indicating a summary of the technical information 123, information 306 indicating the technical field to which the invention described in the technical information 123 belongs, information 308 indicating the problem that the invention aims to solve, information 310 indicating the effects of the invention, information 312 indicating the features of the invention, information 314 indicating the components of the invention, information 316 indicating the patent classification to which the invention belongs, and information 318 indicating the parts of the patent information reviewed during the patent search.

[0131] (F2. Patent Search Agent AG21) Next, we will describe the patent search agent AG21 shown in Figure 7.

[0132] The patent search agent AG21 is an AI agent responsible for patent searches. More specifically, the patent search agent AG21 is pre-configured with instructions to perform a patent search. These instructions are specified in the argument section 31D (see Figure 3) related to the patent search agent AG21. As an example, the patent search agent AG21 is configured to retrieve the set of patent documents 125 from the patent database 124.

[0133] The information processing device 100 inputs the survey and design information 130 generated by the survey and design agent AG10 into the patent search agent AG21. As a result, the patent search agent AG21 autonomously performs a patent search based on the survey and design information 130.

[0134] The information processing device 100 receives the survey and design information 130 as input to the patent search agent AG21, and based on the results obtained from the patent search agent AG21, it acquires a set of patent documents 125 to compare with the technical information 123.

[0135] Furthermore, since the patent search agent AG21 performs autonomous patent searches, the number of hits for patent document 125 may be too few or too many. Therefore, the argument unit 31D specifies that if the number of hits for patent document 125 falls outside the target range, the patent search agent AG21 should perform the patent search again.

[0136] The information processing device 100 causes the patent search agent AG21 to repeat the patent search until the number of patent document hits by the patent search agent AG21 falls within the target range. The target range can be determined, for example, based on the target number entered in the input field 414 (see Figure 8) mentioned above.

[0137] For example, the lower limit of the target range is calculated by multiplying the target number by a predetermined first ratio. This first ratio is, for example, a value greater than 0 and less than 1. For example, this first ratio is a value between 25% and 75% (for example, 50%).

[0138] The upper limit of the target range is calculated by multiplying the target number by a predetermined second ratio. This second ratio is, for example, a value greater than 1. As an example, this second ratio is a number greater than 150% (for example, 200%).

[0139] Preferably, the patent search agent AG21 is configured with an instruction to modify the search conditions to narrow the search range if the number of hits in patent document 125 exceeds the upper limit of the target range. This instruction is configured in the argument section 31D of the patent search agent AG21. The instruction to narrow the search range includes, for example, at least one of an instruction to add search keywords with an AND condition and an instruction to subdivide the search keywords.

[0140] Furthermore, as another example, if IPC is specified as a search criterion, it includes instructions to specify lower-level classifications. For example, if G06F "Classification of Electrical Digital Data Processing" is specified as the IPC subclass, it is instructed to specify G06F16 "Classification of Information Retrieval" as the lower-level IPC main group.

[0141] Furthermore, the patent search agent AG21 is instructed in the argument section 31D to broaden the search conditions if the number of hits falls below the lower limit of the target range. This instruction is specified in the argument section 31D of the patent search agent AG21. This instruction includes, for example, at least one of the following: an instruction to delete part of the search keywords, an instruction to add search keywords in an OR condition, and an instruction to broaden the search keywords to a higher-level concept.

[0142] Furthermore, as another example, if IPC is specified as a search criterion, it includes instructions to specify a higher-level classification. For example, if the classification G06F "Electrical Digital Data Processing" is specified as the IPC subclass, it is instructed to specify G06 "Computation or Counting Classification" as the higher-level IPC class.

[0143] The patent search agent AG21 outputs the set of patent documents 125 that it has searched to the screening agent AG22.

[0144] (F3. Screening Agent AG22) Next, we will explain the screening agent AG22 shown in Figure 7.

[0145] Screening agent AG22 is an AI agent responsible for screening patent documents 125 retrieved by patent search agent AG21. More specifically, screening agent AG22 is pre-configured to remove irrelevant patent documents 125. This configuration is specified in the argument section 31D (see Figure 3) of screening agent AG22.

[0146] Furthermore, the screening agent AG22 is instructed to accept the survey design information 130 generated by the survey design agent AG10 as input. As a result, the screening agent AG22 operates to autonomously accept not only the patent documents 125 retrieved by the patent search agent AG21, but also the survey design information 130 generated by the survey design agent AG10 as input.

[0147] The information processing device 100 receives the patent documents 125 retrieved by the patent search agent AG21 and the research and design information 130 generated by the research and design agent AG10 as input to the screening agent AG22. Based on the results obtained from the screening agent AG22, the device obtains a set of patent documents 125 after removing those with low relevance.

[0148] In the following, the set of patent documents 125 after removing less relevant ones will also be referred to as the "set of patent documents 125 after screening." The screening agent AG22 outputs the set of patent documents 125 after screening to the relevance evaluation agent AG23.

[0149] In the above description, an example was given in which the set of patent documents 125 retrieved by the patent search agent AG21 is input to the screening agent AG22. However, this set may be set by the user. That is, the information processing device 100 obtains the set of patent documents 125 that are the subject of the patent search from any source. The information processing device 100 then inputs the search design information 130 and the acquired set of patent documents 125 to the screening agent AG22. By inputting the search design information 130 and the set of patent documents 125 into the screening agent AG22, the information processing device 100 obtains the set of patent documents 125 after screening based on the results obtained from the screening agent AG22.

[0150] (F4. Relevance Evaluation Agent AG23) Next, we will explain the relevance evaluation agent AG23 shown in Figure 7.

[0151] The relevance evaluation agent AG23 is an AI agent responsible for evaluating the relevance between two input pieces of information. More specifically, the relevance evaluation agent AG23 is pre-configured with instructions to evaluate relevance. These instructions are specified in the argument section 31D (see Figure 3) related to the relevance evaluation agent AG23.

[0152] Furthermore, the relevance evaluation agent AG23 is instructed to accept the survey design information 130 generated by the survey design agent AG10 as input. As a result, the relevance evaluation agent AG23 operates autonomously to accept not only the set of patent documents 125 after screening, but also the survey design information 130 generated by the survey design agent AG10 as input.

[0153] The information processing device 100 receives the survey and design information 130 and the set of patent documents 125 after screening as input to the relationship evaluation agent AG23. Based on the results obtained from the relationship evaluation agent AG23, the device acquires an evaluation result showing the relationship between each of the patent documents 125 after screening and the technical information 123.

[0154] Figure 10 shows an example of the relevance evaluation results 326. In the example in Figure 10, the relevance evaluation results 326 are shown in tabular format, but the data format of the relevance evaluation results 326 is arbitrary.

[0155] The output format of the relevance evaluation result 326 is specified in advance as an instruction in the argument section 31D of the relevance evaluation agent AG23, for example. The relevance evaluation agent AG23 outputs the relevance evaluation result 326 according to the said output format.

[0156] The relevance evaluation result 326 includes an evaluation result 327 for each of the patent documents 125 after screening. In the example in Figure 10, each of the evaluation results 327 is associated with a specific identifier for patent document 125. The identifier for patent document 125 is defined, for example, by the application number, publication number, or registration publication number.

[0157] Furthermore, the evaluation result 327 includes, in addition to the constituent elements "A" to "D" of the invention relating to technical information 123, "description text," "location of description," "reason for extraction," "relevance evaluation rank," and "overall evaluation rank per document." "Description text," "location of description," "reason for extraction," "relevance evaluation rank," and "overall evaluation rank per document" are results obtained, for example, by the relevance evaluation agent AG23 autonomously inputting instruction sentences (prompts) to the large-scale language model MD.

[0158] As mentioned above, the survey and design information 130 includes the analysis results of the technical information 123 (for example, technical field, problem, effect, components, patent classification, etc.). Therefore, the relationship between the technical information 123 and patent document 125 is synonymous with the relationship between the survey and design information 130 and patent document 125.

[0159] The relevance evaluation agent AG23 outputs the results of the relevance evaluation to the report generation agent AG24.

[0160] In the above description, an example was given in which the set of patent documents 125 after screening by the screening agent AG22 is input to the relevance evaluation agent AG23. However, this set may be retrieved by the patent search agent AG21 mentioned above, or it may be set by the user. That is, the information processing device 100 obtains the set of patent documents 125 that are the subject of the patent search from an arbitrary source. The information processing device 100 then inputs the search design information 130 and the acquired set of patent documents 125 to the relevance evaluation agent AG23. By inputting the search design information 130 and the set of patent documents 125 into the relevance evaluation agent AG23, the information processing device 100 obtains the set of patent documents 125 after screening based on the results obtained from the relevance evaluation agent AG23.

[0161] (F5. Report generation agent AG24) Next, we will explain the report generation agent AG24 shown in Figure 7.

[0162] The report generation agent AG24 is an AI agent responsible for summarizing the relevance evaluation results from the relevance evaluation agent AG23. The report generation agent AG24 is pre-configured with instructions to summarize the data in a predetermined format. These instructions are specified in the argument section 31D (see Figure 3) related to the report generation agent AG24. The specified output format may be, for example, JSON, Markdown, or any other format.

[0163] Furthermore, the report generation agent AG24 is instructed to accept the survey design information 130 generated by the survey design agent AG10 as input. As a result, the report generation agent AG24 operates autonomously to accept not only the evaluation results output from the relevance evaluation agent AG23, but also the survey design information 130 generated by the survey design agent AG10 as input.

[0164] The information processing device 100 receives the survey and design information 130 and the aforementioned relevance evaluation results 326 as input to the report generation agent AG24, and based on the results obtained from the report generation agent AG24, it obtains a patent search report 140 relating to the technical information 123.

[0165] Figures 11 and 12 show an example of report 140. Figure 11 shows the first half of report 140. Figure 12 shows the second half of report 140. Report 140 is displayed, for example, on the display 206 of the user terminal 200.

[0166] Report 140 includes Information 142. Information 142 corresponds to the research objective entered in input field 412 (see Figure 8) of the settings screen 400A described above.

[0167] Report 140 also includes Information 144, which provides an overview of the patent search results. For example, Information 144 includes the date and time the patent search was conducted, the initial number of hits for Patent Document 125, the number of Patent Document 125 removed by screening, the number of Patent Document 125 reviewed in detail, the number of Patent Document 125 reviewed for relevance assessment, and the number of Patent Document 125 reviewed again.

[0168] Report 140 also includes Information 146, which provides a summary of the technical information 123 that was investigated.

[0169] Furthermore, report 140 includes information 148. Information 148 shows the contents of the survey design information 130 generated by the survey design agent AG10 described above.

[0170] Report 140 also includes Information 150, which shows the search criteria generated by the patent search agent AG21 described above. Information 150 also includes the number of patent documents 125 that were removed by screening by the screening agent AG22.

[0171] Furthermore, Report 140 includes the relevance evaluation results 152 relating to Technical Information 123 and Patent Document 125. The relevance evaluation results 152 are included in Report 140 for each Patent Document 125 whose relevance was evaluated. The relevance evaluation results 152 are arranged, for example, in descending order of relevance to Patent Document 125.

[0172] The relevance evaluation result 152 shows the relevance evaluation result performed by the relevance evaluation agent AG23 described above. The relevance evaluation result 152 also includes information to identify the patent document 125 that was the subject of the relevance evaluation (for example, application number and publication number).

[0173] Furthermore, the relevance evaluation result 152 includes results indicating the degree of relevance between technical information 123 and patent document 125. For example, the degree of relevance between technical information 123 and patent document 125 is indicated by an evaluation rank. The evaluation rank is displayed separately for each constituent element "A" to "D" related to technical information 123.

[0174] Furthermore, the relevance evaluation result 152 includes the location of the description in Patent Document 125 relating to constituent elements "A" to "D," and the content described in that location in Patent Document 125. The relevance evaluation result 152 may also include comments related to the relevance evaluation.

[0175] Report 140 also includes an overall evaluation result 158, which includes, for example, comments that provide an overall assessment of the patent search.

[0176] Further, the report 140 includes information 160. The information 160 shows a list of patent documents 125 whose relevance to the technical information 123 has been evaluated. Also, the information 160 includes the above evaluation ranks for each of the patent documents 125.

[0177] Further, the report 140 includes information 162. The information 162 shows the patent documents 125 removed by the above-described screening agent AG22. Preferably, the information 162 includes the reasons for removal from the evaluation target.

[0178] <G. Modified Example of the Setting Screen 400A> Next, referring to FIG. 13, a modified example of the above-described setting screen 400A will be described. FIG. 13 is a diagram showing the setting screen 400A according to the modified example.

[0179] The setting screen 400A shown in FIG. 13 is different from the setting screen 400A shown in FIG. 8 in that it includes a setting column 430. Since the other points are as described above, the description thereof will not be repeated.

[0180] As described above, the process of generating the investigation design information 130 by the investigation design agent AG10, the patent search process by the patent search agent AG21, the screening process by the screening agent AG22, the relevance evaluation process by the relevance evaluation agent AG23, and the report generation process by the report generation agent AG24 are executed in order.

[0181] The setting column 430 is configured to receive settings regarding whether to obtain approval from the user for the output result of the AI agent. As an example, the setting screen 400A includes setting buttons 432, 434, 436. The setting buttons 432, 434, 436 are configured to be selectable one by one.

[0182] If setting button 432 is selected, the entire process is left to the AI ​​agent. In other words, when setting button 432 is selected, each AI agent will have subsequent AI agents continue processing without obtaining user approval for the output results.

[0183] If setting button 434 is selected, each AI agent will obtain user approval for all processes. In other words, if setting button 434 is selected, each AI agent will obtain user approval for the output results each time its process is completed.

[0184] If the setting button 436 is selected, whether or not approval is required for the output results of each AI agent is set individually. Examples of processes for which approval requirements can be set include the patent search process by the patent search agent AG20, the patent search process by the patent search agent AG21, the screening process by the screening agent AG22, the evaluation process by the relevance evaluation agent AG23, and the report generation process by the report generation agent AG24.

[0185] The user makes various settings on the settings screen 400A and then presses the start button 420. Based on this, the information processing device 100 causes various AI agents to execute processing.

[0186] Figure 14 shows an example of the execution screen 400B that is displayed while the AI ​​agent is performing a process.

[0187] The execution screen 400B displays the processing status of the AI ​​agent in real time. For example, the tool (function) that the AI ​​agent is currently executing, the result indicating whether the execution of that tool has completed successfully, and the output results of the AI ​​agent are displayed on the execution screen 400B in real time as the process progresses.

[0188] In the example shown in Figure 14, the survey design information 130 generated by the survey design agent AG10 is displayed on the execution screen 400B. As an example, suppose the user has set the settings field 430 (see Figure 13) to require approval for the output results of the survey design agent AG10. In this case, the information processing device 100 temporarily suspends the execution of the survey design agent AG10 before it starts subsequent processing.

[0189] Next, the information processing device 100 displays the survey and design information 130 and an approval request screen 450 that accepts whether or not to approve the contents of the survey and design information 130. The approval request screen 450 may be configured integrally with the execution screen 400B, or it may be configured as a separate screen from the execution screen 400B. The approval request screen 450 includes, for example, buttons 452, 454, and 456.

[0190] Button 452 accepts an operation to instruct approval. When the information processing device 100 receives an operation to instruct approval, it inputs the survey and design information 130 as is to the subsequent patent search agent AG21 following the survey and design agent AG10.

[0191] Button 454 accepts an operation to instruct editing. When the information processing device 100 receives an operation to instruct editing, it accepts the editing of the survey design information 130. In this case, the survey design information 130 becomes editable. Based on the user's completion of editing the survey design information 130, the information processing device 100 inputs the edited survey design information 130 to the subsequent patent search agent AG21 following the survey design agent AG10.

[0192] Button 456 accepts an operation to instruct a redo. When the information processing device 100 accepts an operation to instruct a redo, it causes the survey design information 130 to be regenerated by the survey design agent AG10.

[0193] If a rework order is to be given, a form may be provided to set instructions for informing the survey design agent AG10 why the survey design information 130 was inappropriate. For example, if "the patent classification is incorrect" or "the point of the invention is incorrect" is specified, the survey design agent AG10 will regenerate the survey design information 130 while considering the reason for the error.

[0194] In the example shown in Figure 14, buttons 452, 454, and 456 are displayed on the approval request screen 450. However, the approval request screen 450 only needs to be configured to accept at least one of the following: whether or not to approve the survey design information 130. In other words, it is not necessary for both button 454, which instructs editing, and button 456, which instructs redo, to be displayed on the approval request screen 450. Only one of buttons 454 or 456 may be displayed on the approval request screen 450.

[0195] As described above, the information processing device 100 accepts a setting in advance regarding whether or not approval is required for the output results of the survey and design agent AG10. Then, if the setting indicates that approval is required, the information processing device 100 displays an approval request screen 450 that accepts whether or not to approve the survey and design information 130, based on the fact that the survey and design agent AG10 has generated the survey and design information 130.

[0196] The search design information 130 significantly impacts the accuracy of subsequent patent searches. By obtaining approval for the content of the search design information 130, the information processing device 100 ensures that subsequent patent searches align with the user's intentions.

[0197] Similarly, the information processing device 100 may accept a setting regarding whether or not approval is required for the output results of the patent search agent AG20. This setting can be accepted, for example, on the setting screen 400A. Suppose the user has set the setting screen 400A to require approval for the output results of the patent search agent AG20. In this case, the information processing device 100 displays an approval request screen that accepts whether or not to approve the results based on the fact that the patent search results of the patent search agent AG20 have been generated. This approval request screen is configured to accept, for example, an operation to instruct approval, an operation to instruct editing, and an operation to instruct redo.

[0198] When the information processing device 100 receives an operation to instruct approval, it outputs the output results of the patent search agent AG20 as they are. Furthermore, when the information processing device 100 receives an operation to instruct editing, it accepts the editing of the output results of the patent search agent AG20. Additionally, when the information processing device 100 receives an operation to instruct redo, it causes the patent search agent AG20 to perform the patent search again.

[0199] The patent search agent AG20 is composed of multiple AI agents: a patent search agent AG21, a screening agent AG22, a relevance evaluation agent AG23, and a report generation agent AG24. Preferably, the information processing device 100 may obtain approval for the output result of at least one of the multiple AI agents.

[0200] As an example, suppose the user has configured the settings screen 400A described above to require approval for the output results of the patent search agent AG21. In this case, the information processing device 100 temporarily suspends the execution of the patent search agent AG21 before starting subsequent processing.

[0201] Next, the information processing device 100 displays a list of patent documents 125 retrieved by the patent search agent AG21 and accepts at least whether to approve the contents of the list. This operation is accepted, for example, on the approval request screen. The approval request screen is configured to accept, for example, an operation to instruct approval, an operation to instruct editing, and an operation to instruct redo.

[0202] When the information processing device 100 receives an operation to instruct approval, it inputs the set of patent documents 125 as is to the subsequent screening agent AG22 following the patent search agent AG21.

[0203] Furthermore, when the information processing device 100 receives an operation to instruct editing, it accepts the editing of the set of patent documents 125 to be evaluated. Subsequently, the information processing device 100 inputs the edited set of patent documents 125 to the subsequent screening agent AG22 following the patent search agent AG21.

[0204] Furthermore, if the information processing device 100 receives an operation to instruct the user to redo the process, it causes the patent search agent AG21 to search for patent document 125 again.

[0205] Similarly, suppose the user has configured the settings screen 400A described above to require approval for the output results of the screening agent AG22. In this case, the information processing device 100 pauses the execution of the screening agent AG22 before starting subsequent processing.

[0206] Next, the information processing device 100 displays a list of the patent documents 125 after screening and accepts at least whether to approve the contents of the list. This operation is accepted, for example, on the approval request screen. The approval request screen is configured to accept, for example, an operation to instruct approval, an operation to instruct editing, and an operation to instruct redo.

[0207] When the information processing device 100 receives an operation to instruct approval, it inputs the set of patent documents 125 after screening directly to the subsequent relevance evaluation agent AG23 following the screening agent AG22.

[0208] Furthermore, when the information processing device 100 receives an operation to instruct editing, it accepts the editing of the set of patent documents 125 after screening. Subsequently, the information processing device 100 inputs the set of edited patent documents 125 to the subsequent relevance evaluation agent AG23 following the screening agent AG22.

[0209] Furthermore, if the information processing device 100 receives an operation to instruct the user to redo the process, it causes the patent search agent AG21 to perform the search for patent document 125 again.

[0210] Similarly, suppose the user has configured the settings screen 400A described above to require approval for the output results of the relevance evaluation agent AG23. In this case, the information processing device 100 pauses the execution of the relevance evaluation agent AG23 before it starts subsequent processing.

[0211] Next, the information processing device 100 displays the relevance evaluation results from the relevance evaluation agent AG23 and accepts at least whether to approve the relevance evaluation results. This operation is accepted, for example, on the approval request screen. The approval request screen is configured to accept, for example, an operation to instruct approval, an operation to instruct editing, and an operation to instruct redo.

[0212] When the information processing device 100 receives an operation to instruct approval, it inputs the relevance evaluation result directly to the subsequent report generation agent AG24 following the relevance evaluation agent AG23.

[0213] Furthermore, when the information processing device 100 receives an operation to instruct editing, it accepts the editing of the relevance evaluation results. Subsequently, the information processing device 100 inputs the edited relevance evaluation results to the subsequent report generation agent AG24 following the relevance evaluation agent AG23.

[0214] Furthermore, if the information processing device 100 receives an operation instructing a redo, it causes the relevance evaluation agent AG23 to execute the relevance evaluation process again.

[0215] Similarly, suppose the user has configured the settings screen 400A described above to require approval for the output results of the report generation agent AG24. In this case, the information processing device 100 temporarily suspends the execution of the process based on the completion of the processing by the report generation agent AG24.

[0216] Next, the information processing device 100 displays the report 140 generated by the report generation agent AG24 and accepts at least whether to approve the report 140. This operation is accepted, for example, on the approval request screen. The approval request screen is configured to accept, for example, an operation to instruct approval, an operation to instruct editing, and an operation to instruct redo.

[0217] If the information processing device 100 receives an operation to instruct approval, it terminates the patent search process. If the information processing device 100 receives an operation to instruct editing, it accepts the editing of the report 140 generated by the report generation agent AG24. After that, the information processing device 100 terminates the patent search process. Furthermore, if the information processing device 100 receives an operation to instruct redo, it causes the report generation agent AG24 to execute the report 140 generation process again.

[0218] This allows users to conduct patent searches while reviewing, correcting, and approving processes that the AI ​​agent is not good at or that are high-risk.

[0219] <Another Example of the Configuration of the H.AI Agent> In the example of FIG. 7 described above, the investigation design agent AG10 was responsible for generating the investigation design information 130, and the patent investigation agent AG20 was responsible for the patent investigation based on the investigation design information 130. And the patent investigation was realized by the patent search agent AG21, the screening agent AG22, the relevance evaluation agent AG23, and the report generation agent AG24.

[0220] However, the configuration of the AI agent is not limited to the example shown in FIG. 7. Below, referring to FIG. 15, another example of the configuration of the AI agent will be described. FIG. 15 is a diagram showing another example of the configuration of the AI agent.

[0221] In this example, the initialization agent AG1, the above-described investigation design agent AG10, and the above-described patent investigation agent AG20 are implemented in the information processing apparatus 100. The patent investigation agent AG20 is composed of a patent search agent AG21, an extraction agent AG21_1, a condition generation agent AG21_2, a search execution agent AG21_3, an optimization agent AG21_4, a screening agent AG22, a relevance evaluation agent AG23, an iterative proofreading agent AG23_1, a quality evaluation agent AG23_2, and a report generation agent AG24.

[0222] In the example of FIG. 15, the root agent AG0 is defined as the top-level AI agent. As the sub-agents of the root agent AG0, the initialization agent AG1, the above-described investigation design agent AG10, the patent search agent AG21, the screening agent AG22, the relevance evaluation agent AG23, and the report generation agent AG24 are defined.

[0223] Furthermore, as sub-agents of the patent search agent AG21, an extraction agent AG21_1, a condition generation agent AG21_2, a search execution agent AG21_3, and an optimization agent AG21_4 are defined.

[0224] Furthermore, as sub-agents of the relevance evaluation agent AG23, a repetitive proofreading agent AG23_1 and a quality evaluation agent AG23_2 are defined.

[0225] <I. Example of Agent Definition> Hereinafter, with reference to FIGS. 16 to 28, an example of the definition of each AI agent shown in FIG. 15 will be described.

[0226] (I1. Root Agent AG0) First, with reference to FIG. 16, the definition of the root agent AG0 shown in FIG. 15 will be described. FIG. 16 is a diagram showing an example of the code definition related to the root agent AG0.

[0227] Regarding the root agent AG0, the code shown in FIG. 16 is defined for the above-described argument parts 31A to 31D, 31F (see FIG. 3).

[0228] More specifically, the name of the root agent AG0 is specified in the argument part 31A. In the example of FIG. 16, the name "PatentResearchWorkflow" is specified in the argument part 31A.

[0229] The type of large language model used by the root agent AG0 is specified in the argument part 31B. In the example of FIG. 16, "large language model α" is specified in the argument part 31B. Note that the same large language model as other AI agents may be specified in the argument part 31B, or a different large language model from other AI agents may be specified.

[0230] The argument section 31C specifies a description of the function of the root agent AG0. In the example in Figure 16, the argument section 31C specifies a function description of managing all processes related to patent searching.

[0231] The argument section 31D contains instructions for the root agent AG0. In the example in Figure 16, the argument section 31D contains instructions to sequentially execute the patent search process. More specifically, the instructions specify that the patent search process should be executed sequentially as initialization, search design, patent search, screening, peer review / quality evaluation, and report generation.

[0232] The argument section 31F specifies the subagents of the root agent AG0. In the example in Figure 16, the subagents specified in argument section 31F are "initializer_agent", "investigation_design_agent", "patent_discovery_agent", "primary_screening_agent", "review_quality_agent", and "report_generator_agent".

[0233] "initializer_agent" corresponds to the initialization agent AG1 described later. "investigation_design_agent" corresponds to the investigation design agent AG10 described later. "patent_discovery_agent" corresponds to the patent search agent AG21 described later. "primary_screening_agent" corresponds to the screening agent AG22 described later. "review_quality_agent" corresponds to the relevance evaluation agent AG23 described later. "report_generator_agent" corresponds to the report generation agent AG24 described later.

[0234] (I2. Initialization Agent AG1) Next, with reference to Figure 17, the definition of the initialization agent AG1 shown in Figure 15 will be explained. Figure 17 is a diagram showing an example of the code definition related to the initialization agent AG1.

[0235] For the initialization agent AG1, the code shown in Figure 17 is defined for the argument parts 31A to 31E (see Figure 3) mentioned above.

[0236] More specifically, the argument 31A specifies the name of the initialization agent AG1. In the example in Figure 17, the name "initializer_agent" is specified in the argument 31A.

[0237] The argument section 31B specifies the model that the initialization agent AG1 can use. In the example in Figure 17, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a large-scale language model different from those of other AI agents.

[0238] The argument section 31C specifies a description of the function of the initialization agent AG1. In the example in Figure 17, the argument section 31C specifies a function that initializes memory and acquires the above-mentioned technical information 123 as the invention content.

[0239] The argument section 31D specifies instructions for the initialization agent AG1. In the example in Figure 17, instructions 1A to 1D are specified in the argument section 31D.

[0240] More specifically, instruction 1A is an instruction to retrieve the invention details (i.e., the technical information 123 mentioned above). Instruction 1B is an instruction to initialize the memory that will hold the retrieved invention details (i.e., the technical information 123 mentioned above). Instruction 1C is an instruction that indicates the conditions for performing the initialization. Instruction 1D is an instruction that indicates the content to be displayed when the initialization process is completed successfully.

[0241] Note that the argument section 31D does not need to specify all of the instructions 1A to 1D; it may specify only some of the instructions 1A to 1D.

[0242] The argument section 31E specifies a tool that the initialization agent AG1 can use. In the example in Figure 17, the function "initialize_review_memory_tool" is specified in the argument section 31E. This function is a program function for initializing memory. This function is called, for example, when instruction 1B is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0243] (I3. Survey Design Agent AG10) Next, with reference to Figure 18, the definition of the survey design agent AG10 shown in Figure 15 will be explained. Figure 18 is a diagram showing an example of the code definition related to the survey design agent AG10.

[0244] For the survey design agent AG10, the codes shown in Figure 18 are defined for the argument parts 31A to 31E (see Figure 3) mentioned above.

[0245] More specifically, the argument 31A specifies the name of the investigation design agent AG10. In the example in Figure 18, the name "investigation_design_agent" is specified in the argument 31A.

[0246] The argument section 31B specifies the type of large-scale language model used by the survey and design agent AG10. In the example in Figure 18, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a large-scale language model different from those used by other AI agents.

[0247] The argument section 31C specifies a functional description of the investigation and design agent AG10. In the example shown in Figure 18, the functional description specified in the argument section 31C is that the agent technically analyzes the input invention content and generates investigation and design information 130 that subsequent AI agents will refer to.

[0248] The argument section 31D specifies instructions for the survey and design agent AG10. In the example in Figure 18, instructions 10A to 10D are specified in the argument section 31D.

[0249] Instruction 10A is an instruction to retrieve the invention details (i.e., the technical information 123 described above) stored in memory by the initialization agent AG1 described above.

[0250] Instruction 10B is an instruction to analyze the acquired inventive content (i.e., technical information 123). Preferably, instruction 10B specifies that the inventive content (technical information 123) be analyzed from a predetermined viewpoint. This predetermined viewpoint includes at least one of the following: the viewpoint of the search content, the viewpoint of the technical field, the viewpoint of the problem, the viewpoint of the effect of the invention, the viewpoint of the features of each claim, the viewpoint of the components, and the viewpoint of patent classification. This patent classification includes, for example, at least one of IPC, FI, and F-term. Instruction 10B may also include a target number of patents to search.

[0251] Instruction 10C is an instruction to save the survey design information 130 in a predetermined format. This ensures that the survey design information 130 is saved in the specified data format. For example, the data format could be JSON.

[0252] Instruction 10D is an instruction to output the survey design information 130 in a predetermined format. As a result, the survey design information 130 is displayed in the specified data format, and the user can check the contents of the survey design information 130 on the display.

[0253] Note that the argument section 31D does not need to specify all of the instructions 10A to 10D; it may specify only some of the instructions 10A to 10D.

[0254] The argument section 31E specifies the tools that the investigation and design agent AG10 can use. In the example in Figure 18, the function "get_invention_description_tool" is specified in the argument section 31E. This function is a program function for retrieving the invention details (i.e., the technical information 123) stored in memory by the initialization agent AG1 described above. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0255] (I4. Patent Search Agent AG21) Next, with reference to Figure 19, the definition of the patent search agent AG21 shown in Figure 15 will be explained. Figure 19 is a diagram showing an example of a code definition related to the patent search agent AG21.

[0256] For the patent search agent AG21, the codes shown in Figure 19 are defined for the argument parts 31A to 31D and 31F (see Figure 3) mentioned above.

[0257] More specifically, the argument 31A specifies the name of the patent search agent AG21. In the example in Figure 19, the name "patent_discovery_agent" is specified in the argument 31A.

[0258] The argument section 31B specifies the type of large-scale language model used by the patent search agent AG21. In the example in Figure 19, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a large-scale language model different from those used by other AI agents.

[0259] The argument section 31C specifies a description of the function of the patent search agent AG21. In the example in Figure 19, the function description "managing the patent search process" is specified in the argument section 31C.

[0260] The argument section 31D contains instructions for the patent search agent AG21. In the example in Figure 19, the argument section 31D contains instructions to sequentially execute the patent search process. More specifically, the instructions specify that the argument section 31D should sequentially execute the following steps as the patent search process: keyword extraction, search condition construction, patent search execution, and search optimization.

[0261] The argument section 31F specifies the subagents of the patent search agent AG21. In the example in Figure 19, the subagents specified in the argument section 31F are "keyword_extractor_agent", "search_command_builder_agent", "patent_search_agent", and "search_optimizer_agent".

[0262] More specifically, "keyword_extractor_agent" corresponds to the extraction agent AG21_1 described later. "search_command_builder_agent" corresponds to the condition generation agent AG21_2 described later. "patent_search_agent" corresponds to the search execution agent AG21_3 described later. "search_optimizer_agent" corresponds to the optimization agent AG21_4 described later.

[0263] (I5. Extraction Agent AG21_1) Next, with reference to Figure 20, the definition of the extraction agent AG21_1 shown in Figure 15 will be explained. Figure 20 is a diagram showing an example of the code definition related to the extraction agent AG21_1.

[0264] For the extraction agent AG21_1, the code shown in Figure 20 is defined for the argument parts 31A to 31E (see Figure 3) mentioned above.

[0265] More specifically, the name of the extraction agent AG21_1 is specified in the argument part 31A. In the example of FIG. 20, the name "keyword_extractor_agent" is specified in the argument part 31A.

[0266] In the argument part 31B, the type of the large language model used by the extraction agent AG21_1 is specified. In the example of FIG. 20, "large language model α" is specified in the argument part 31B. Note that the same large language model as other AI agents may be specified in the argument part 31B, or a large language model different from other AI agents may be specified.

[0267] In the argument part 31C, a function description of the extraction agent AG21_1 is specified. In the example of FIG. 20, a function description of generating keywords for patent search based on the components described in the survey design information 130 is specified in the argument part 31C.

[0268] In the argument part 31D, an instruction for the extraction agent AG21_1 is specified. In the example of FIG. 20, instructions 21_1A to 21_1C are specified in the argument part 31D.

[0269] Instruction 21_1A is an instruction to acquire the survey design information 130 generated by the above-mentioned survey design agent AG10.

[0270] Instruction 21_1B is an instruction to extract keywords for patent search based on the acquired survey design information 130. Preferably, instruction 21_1B includes rules for extracting keywords.

[0271] Instruction 21_1C is an instruction to save the extracted keywords in a predetermined output format. As an example, in this output format, it is specified that the extracted keywords and the reasons for extraction are to be output.

[0272] Note that the argument section 31D does not need to specify all of the instructions 21_1A to 21_1C; it may specify only some of the instructions 21_1A to 21_1C.

[0273] The argument section 31E specifies the tools that the extraction agent AG21_1 can use. In the example in Figure 20, the function "get_investigation_design_tool" is specified in the argument section 31E. This function is a program function for obtaining the investigation design information 130 described above. This function is called, for example, when instruction 21_1B is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0274] (I6. Condition Generation Agent AG21_2) Next, referring to Figure 21, we will explain the definition of the condition generation agent AG21_2 shown in Figure 15. Figure 21 is a diagram showing an example of the code definition related to the condition generation agent AG21_2.

[0275] For the condition generation agent AG21_2, the code shown in Figure 21 is defined for the argument parts 31A to 31E (see Figure 3) mentioned above.

[0276] More specifically, the argument 31A specifies the name of the condition generation agent AG21_2. In the example in Figure 21, the name "search_command_builder_agent" is specified in the argument 31A.

[0277] The argument section 31B specifies the type of large-scale language model used by the condition generation agent AG21_2. In the example in Figure 21, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a large-scale language model different from those used by other AI agents.

[0278] The argument section 31C specifies a function description for the condition generation agent AG21_2. In the example in Figure 21, the function description specified in the argument section 31C is that the agent automatically determines search conditions other than keywords from the technical field. Since the search keywords are generated by the extraction agent AG21_1 mentioned above, the function description specified in the argument section 31C is that the agent generates search conditions other than the search keywords.

[0279] The argument section 31D specifies instructions for the condition generation agent AG21_2. In the example in Figure 21, instructions 21_2A to 21_2D are specified in the argument section 31D.

[0280] Instruction 21_2A is an instruction to obtain keywords for patent search. These keywords correspond to those extracted from the search design information 130 by the extraction agent AG21_1 described above.

[0281] Instruction 21_2B is an instruction to analyze the keywords obtained by Instruction 21_2A. Instruction 21_2B analyzes, for example, the patent classifications related to the keywords. These patent classifications include, for example, at least one of IPC, FI, and F-term.

[0282] Instruction 21_2C is an instruction that determines the search criteria for patent documents. For example, instruction 21_2C determines the patent classification. Another example is that instruction 21_2C determines the date range for the publication date of the patent documents to be searched. Yet another example is that instruction 21_2C determines the field to be searched. This field indicates a part of the patent document. Examples of such parts include claims and abstracts.

[0283] Instruction 21_2D is an instruction that specifies the data format for the output data related to the search criteria. For example, the output data includes keywords extracted by the extraction agent AG21_1 described above, the patent classification determined by instruction 21_2C, the date range determined by instruction 21_2C, and the search target field determined by instruction 21_2C. The output data may also include a country code.

[0284] Note that the argument section 31D does not need to specify all of the instructions 21_2A to 21_2D; it may specify only some of the instructions 21_2A to 21_2D.

[0285] The argument section 31E specifies the tools that the condition generation agent AG21_2 can use. In the example in Figure 21, the functions "build_search_conditions_tool" and "save_search_conditions_tool" are specified in the argument section 31E.

[0286] The function "build_search_conditions_tool" generates search conditions (such as patent classifications and date ranges) from keywords. This function is called, for example, when instruction 21_2A is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0287] The function "save_search_conditions_tool" saves the generated search conditions to memory. This function is called, for example, when instruction 21_2D is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0288] (I7. Search execution agent AG21_3) Next, with reference to Figure 22, the definition of the search execution agent AG21_3 shown in Figure 15 will be explained. Figure 22 is a diagram showing an example of the code definition related to the search execution agent AG21_3.

[0289] Regarding the search execution agent AG21_3, the code shown in FIG. 22 is defined for the above-described argument parts 31A to 31E (see FIG. 3).

[0290] More specifically, the name of the search execution agent AG21_3 is specified in the argument part 31A. In the example of FIG. 22, the name "patent_search_agent" is specified in the argument part 31A.

[0291] The type of large language model used by the search execution agent AG21_3 is specified in the argument part 31B. In the example of FIG. 22, "large language model α" is specified in the argument part 31B. Note that the same large language model as other AI agents may be specified in the argument part 31B, or a different large language model from other AI agents may be specified.

[0292] The function description of the search execution agent AG21_3 is specified in the argument part 31C. In the example of FIG. 22, the function description of executing a patent search is specified in the argument part 31C.

[0293] Instructions for the search execution agent AG21_3 are specified in the argument part 31D. In the example of FIG. 22, instructions 21_3A to 21_3C are specified in the argument part 31D.

[0294] Instruction 21_3A is an instruction to generate a search command based on the search conditions generated by the condition generation agent AG21_2.

[0295] Instruction 21_3B is an instruction to search for the patent document ********** 125 to be investigated from the above-described patent database 124 based on the search command generated by instruction 21_3A. Instruction 21_3B is specified to be executed until the number of hits of the patent document 125 reaches a predetermined target number. The target number is the information input on the above-described setting screen 400A (see FIG. 8).

[0296] It should be noted that there is an unclear expression "**********" in the original text at the position of "特許文献125" in Instruction 21_3B. I have translated it as "********** 125" as accurately as possible according to the format. If there is an error, please provide more accurate information for correction.Instruction 21_3C is an instruction to output the search results for Patent Document 125. The output data includes the total number of Patent Documents 125 that were found as search results, the number of Patent Documents 125 that were actually retrieved from that total number, and information about the retrieved Patent Documents 125.

[0297] Note that the argument section 31D does not need to specify all of the instructions 21_3A to 21_3C; it may specify only some of the instructions 21_3A to 21_3C.

[0298] The argument section 31E specifies the tools that the search execution agent AG21_3 can use. In the example in Figure 22, the functions "generate_search_command_tool" and "search_patents_tool" are specified in the argument section 31E.

[0299] The function "generate_search_command_tool" is a function that generates a search command from the search conditions. This function is called, for example, when instruction 21_3A is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0300] The function "search_patents_tool" is a function that takes a search command as input and searches for patent document 125 from the patent database 124. This function is called, for example, when instruction 21_3B is executed. The code portion of this function is predetermined in, for example, the analysis program 122 described above.

[0301] (I8. Optimization Agent AG21_4) Next, with reference to Figure 23, the definition of the optimization agent AG21_4 shown in Figure 15 will be explained. Figure 23 is a diagram showing an example of the code definition related to the optimization agent AG21_4.

[0302] For the optimization agent AG21_4, the code shown in Figure 23 is defined for the argument parts 31A to 31E (see Figure 3) mentioned above.

[0303] More specifically, the argument 31A specifies the name of the optimization agent AG21_4. In the example in Figure 23, the name "search_optimizer_agent" is specified in the argument 31A.

[0304] The argument section 31B specifies the type of large-scale language model used by the optimization agent AG21_4. In the example in Figure 23, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a different large-scale language model from other AI agents.

[0305] The argument section 31C specifies a function description for the optimization agent AG21_4. In the example in Figure 23, the argument section 31C specifies a function description that automatically adjusts the search conditions for patent document 125 when the number of hits for patent document 125 deviates significantly from the target number.

[0306] The argument section 31D specifies instructions for the optimization agent AG21_4. In the example in Figure 23, instructions 21_4A to 21_4F are specified in the argument section 31D.

[0307] Instruction 21_4A is an instruction to obtain the current hit count related to Patent Document 125.

[0308] Instruction 21_4B indicates the number of re-searches to perform, depending on the number of hits obtained by instruction 21_4A. For example, instruction 21_4B is specified so that the larger the absolute difference between the current number of hits and the target number, the more re-searches will be performed. In other words, instruction 21_4B is specified so that the smaller the absolute difference between the current number of hits and the target number, the fewer re-searches will be performed.

[0309] Instruction 21_4C is an instruction relating to guidelines for changing search conditions in accordance with the number of hits obtained in Instruction 21_4A. For example, in a re-search, the search conditions are expanded. The method of expansion includes, for example, at least one of the following: raising the conceptual level of the search keyword, increasing the number of synonyms or related terms of the search keyword, and raising the patent classification in the search conditions to a higher level.

[0310] Instruction 21_4D is an instruction to execute the re-search process of Patent Document 125. This re-search process includes the process of changing the search conditions, the process of generating a search command from the changed search conditions, and the process of executing a search according to the search command.

[0311] Instruction 21_4E is an instruction to report the changes made to the search conditions by instruction 21_4C.

[0312] Instruction 21_4F is an instruction to output the re-search results for Patent Document 125. The output data includes the number of Patent Document 125 results found as re-search results and information about the hit Patent Document 125. In addition, the output data may also include results indicating whether or not a number of Patent Document 125 corresponding to the target number were found.

[0313] Note that the argument section 31D does not need to specify all of the instructions 21_4A to 21_4F; it may specify only some of the instructions 21_4A to 21_4F.

[0314] The argument section 31E specifies the tools that the optimization agent AG21_4 can use. In the example in Figure 23, the functions "get_keyword_optimization_context_tool", "set_optimized_keywords_tool", "generate_search_command_tool", and "search_patents_tool" are specified in the argument section 31E.

[0315] The function "get_keyword_optimization_context_tool" is a program function used to retrieve the current search result status (e.g., the number of results and the completion rate). This function is called, for example, when instruction 21_4A is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0316] The function "set_optimized_keywords_tool" is a program function for setting optimized search conditions. This function is called, for example, when instruction 21_4D is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0317] The function "generate_search_command_tool" is a function that generates a search command from the search conditions. This function is called, for example, when instruction 21_4D is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0318] The function "search_patents_tool" is a function that takes a search command as input and searches for patent document 125 from the patent database 124. This function is called, for example, when instruction 21_4D is executed. The code portion of this function is predetermined, for example, in the analysis program 122 described above.

[0319] (I9. Screening Agent AG22) Next, with reference to Figure 24, the definition of the screening agent AG22 shown in Figure 15 will be explained. Figure 24 is a diagram showing an example of a code definition related to the screening agent AG22.

[0320] For the screening agent AG22, the codes shown in Figure 24 are defined for the argument parts 31A to 31E (see Figure 3) mentioned above.

[0321] More specifically, the argument 31A specifies the name of the screening agent AG22. In the example in Figure 24, the name "primary_screening_agent" is specified in the argument 31A.

[0322] The argument section 31B specifies the type of large-scale language model used by the screening agent AG22. In the example in Figure 24, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a large-scale language model different from those used by other AI agents.

[0323] The argument section 31C specifies a description of the function of the screening agent AG22. In the example in Figure 24, the argument section 31C specifies a function description that removes duplicate patent documents and clearly unrelated patent documents.

[0324] The argument section 31D specifies instructions for the screening agent AG22. In the example in Figure 24, instructions 22A to 22E are specified in the argument section 31D.

[0325] Instruction 22A includes instructions to obtain a list of patent documents 125 to be screened and instructions to obtain the above-mentioned survey design information 130. The patent documents 125 to be screened are the patent documents 125 retrieved by the above-mentioned patent search agent AG21.

[0326] Instruction 22B is an instruction that specifies the conditions for excluding from the screening target.

[0327] Instruction 22C is an instruction to remove Patent Document 125, which has overlapping content. Criteria for determining overlapping content include, for example, that the cases are family cases and that the applicants are the same.

[0328] Instruction 22D is an instruction to perform a screening process on the set of Patent Documents 125. This screening process assigns an index to each of the Patent Documents 125 that indicates the screening result. This index includes a numerical value indicating retention and a numerical value indicating removal. Preferably, the index is numerically divided according to the reasons for retention and removal.

[0329] Instruction 22E is an instruction to output the screening results of the set of patent documents 125. The output data includes, for example, the number of patent documents 125 before screening, the number of patent documents 125 that were retained, and the number of patent documents 125 that were removed.

[0330] The argument section 31E specifies the tools that the screening agent AG22 can use. In the example in Figure 24, the functions "screen_patents_tool" and "apply_screening_results_tool" are specified in the argument section 31E.

[0331] The function "screen_patents_tool" is a program function for obtaining a list of the set of patent documents 125 to be screened. This function is called, for example, when instruction 22A is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0332] The function "apply_screening_results_tool" is a program function for executing a screening process. This screening process assigns an index indicating the screening result to each of the documents in Patent Document 125. This function is called, for example, when instruction 22D is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0333] (I10. Relevance Evaluation Agent AG23) Next, with reference to Figure 25, the definition of the relevance evaluation agent AG23 shown in Figure 15 will be explained. Figure 25 is a diagram showing an example of the code definition related to the relevance evaluation agent AG23.

[0334] For the relevance evaluation agent AG23, the codes shown in Figure 25 are defined for the argument parts 31A to 31D and 31F (see Figure 3) mentioned above.

[0335] More specifically, the argument 31A specifies the name of the relevance evaluation agent AG23. In the example in Figure 25, the name "review_quality_agent" is specified in the argument 31A.

[0336] The argument section 31B specifies the type of large-scale language model used by the relevance evaluation agent AG23. In the example in Figure 25, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a large-scale language model different from those used by other AI agents.

[0337] The argument section 31C specifies a functional description of the relevance evaluation agent AG23. In the example in Figure 25, the functional description specified in the argument section 31C is that iteratively performs peer review and quality evaluation. Preferably, the functional description includes a maximum number of iterations (for example, 3 times).

[0338] The argument section 31D contains instructions for the relevance evaluation agent AG23. In the example in Figure 25, the argument section 31D contains an instruction to repeatedly perform a review of Patent Document 125 until it meets a predetermined quality standard.

[0339] The argument section 31F specifies a subagent of the relevance evaluation agent AG23. In the example in Figure 25, "iterative_patent_reviewer_agent" and "quality_evaluator_for_loop" are specified as subagents in the argument section 31F.

[0340] More specifically, "iterative_patent_reviewer_agent" corresponds to the iterative review agent AG23_1 described later. "quality_evaluator_for_loop" corresponds to the quality evaluation agent AG23_2 described later.

[0341] (I11. Repeated Peer Review Agent AG23_1) Next, with reference to Figure 26, the definition of the iterative review agent AG23_1 shown in Figure 15 will be explained. Figure 26 is a diagram showing an example of the code definition related to the iterative review agent AG23_1.

[0342] For the iterative review agent AG23_1, the code shown in Figure 26 is defined for the argument parts 31A to 31E (see Figure 3) mentioned above.

[0343] More specifically, the argument 31A specifies the name of the iterative review agent AG23_1. In the example in Figure 26, the name "iterative_patent_reviewer_agent" is specified in the argument 31A.

[0344] The argument section 31B specifies the type of large-scale language model used by the iterative review agent AG23_1. In the example in Figure 26, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a large-scale language model different from those used by other AI agents.

[0345] The argument section 31C specifies a functional description of the iterative review agent AG23_1. In the example shown in Figure 26, the functional description specified in the argument section 31C is that the agent reviews each patent document 125 and evaluates its relationship to the survey and design information 130.

[0346] The argument section 31D specifies instructions for the iterative review agent AG23_1. In the example in Figure 26, instructions 23_1A to 23_1H are specified in the argument section 31D.

[0347] Instruction 23_1A is an instruction to obtain the survey design information 130 generated by the aforementioned survey design agent AG10. Furthermore, the instruction specifies that the acquisition of the survey design information 130 should only be performed the first time.

[0348] Instruction 23_1B is an instruction to obtain patent documents 125 for evaluation of their relevance to the survey design information 130. The patent documents 125 to be reviewed for evaluation of relevance are obtained sequentially from the patent documents 125 that have been screened by the screening agent AG22 described above.

[0349] Instruction 23_1C is an instruction to output the progress of the relevance assessment. This means that the progress of the peer review will be output while the iterative review agent AG23_1 is processing.

[0350] Instruction 23_1D is an instruction to obtain the detailed contents of the patent document 125 to be reviewed (for example, the contents of the claims).

[0351] Instruction 23_1E is an instruction to output a peer review evaluation. The review results are shown as an evaluation of the relationship between the survey and design information 130 and the patent document 125 being reviewed. As an example, the review results include commonality in the technical field, commonality in the problems, commonality in the solutions, commonality in each component, and an overall evaluation.

[0352] Instruction 23_1F is an instruction to save the results of the correlation evaluation between the survey and design information 130 and patent document 125 in a predetermined data format. For example, the data format may be JSON. The output data includes the above-mentioned review results.

[0353] Instruction 23_1G is an instruction to repeatedly perform the relevance evaluation. For example, the relevance evaluation is repeated until all of the patent documents 125, after screening by the screening agent AG22 described above, have been reviewed.

[0354] Note that the argument section 31D does not need to specify all of the instructions 23_1A to 23_1G; it may specify only some of the instructions 23_1A to 23_1G.

[0355] The argument section 31E specifies the tools that the iterative review agent AG23_1 can use. In the example in Figure 26, the functions "get_investigation_design_tool", "get_next_patent_for_review", "get_patent_details_tool", and "review_single_patent_tool" are specified in the argument section 31E.

[0356] The function "get_investigation_design_tool" is a program function for obtaining the investigation design information 130 described above. This function is called, for example, when instruction 23_1A is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0357] The function "get_next_patent_for_review" is a program function for sequentially retrieving patent documents 125 that are subject to relevance evaluation from the set of patent documents 125 after screening by the screening agent AG22 described above. This function is called, for example, when instruction 23_1B is executed. The code portion of this function is predetermined in, for example, the analysis program 122 described above.

[0358] The function "get_patent_details_tool" is a program function that takes the identifier of patent document 125 (for example, the application number) as input and retrieves the detailed contents of patent document 125 (for example, the contents of the claims). This function is called, for example, when instruction 23_1D is executed. The code portion of this function is predetermined, for example, in the analysis program 122 described above.

[0359] The function "review_single_patent_tool" is a program function for saving the review results of Patent Document 125 to memory. This function is called, for example, when instruction 23_1F is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0360] (I12. Quality Evaluation Agent AG23_2) Next, with reference to Figure 27, the definition of the quality evaluation agent AG23_2 shown in Figure 15 will be explained. Figure 27 is a diagram showing an example of a code definition related to the quality evaluation agent AG23_2.

[0361] For the quality evaluation agent AG23_2, the codes shown in Figure 27 are defined for the argument parts 31A to 31E (see Figure 3) mentioned above.

[0362] More specifically, the argument 31A specifies the name of the quality evaluation agent AG23_2. In the example in Figure 27, the name "quality_evaluator_for_loop" is specified in the argument 31A.

[0363] The argument section 31B specifies the type of large-scale language model used by the quality evaluation agent AG23_2. In the example in Figure 27, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a large-scale language model different from those used by other AI agents.

[0364] The argument section 31C specifies a functional description for the quality evaluation agent AG23_2. In the example in Figure 27, the functional description specified in the argument section 31C is that the agent evaluates the quality of the review results by the iterative review agent AG23_1.

[0365] The argument section 31D specifies instructions for the quality evaluation agent AG23_2. In the example in Figure 27, instructions 23_2A to 23_2E are specified in the argument section 31D.

[0366] Instruction 23_2A is an instruction to execute the function "evaluate_review_quality_tool" described below. This function evaluates the quality of the peer review results. This quality is indicated, for example, by a score for each Patent Document 125.

[0367] Instruction 23_2B is an instruction to output the review results from the aforementioned iterative review agent AG23_1, along with the quality of that review.

[0368] Instruction 23_2C indicates the conditions for ending the peer review and the conditions for performing a re-review. For example, if the peer review quality score is below a predetermined value, Patent Document 125 is counted as low quality. If this count value is zero, the conditions for ending the peer review are met. Otherwise, the conditions for performing a re-review are met. In this case, the function "prepare_re_review_tool" described later is executed, and the re-review process is performed.

[0369] Instruction 23_2D is an instruction to calculate the quality score of the re-review. If the score is below a predetermined value (for example, below 70 points), the process is repeated until the predetermined quality standard is met.

[0370] Instruction 23_2E is an instruction that specifies the maximum number of iterations for peer review.

[0371] The argument section 31E specifies the tools that the quality evaluation agent AG23_2 can use. In the example in Figure 27, the functions "evaluate_review_quality_tool", "prepare_re_review_tool", and "exit_review_loop_tool" are specified in the argument section 31E.

[0372] The function "evaluate_review_quality_tool" is used to evaluate the quality of the peer review results. This function is called, for example, when instruction 23_2A is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0373] The function "prepare_re_review_tool" is a program function for removing low-quality reviews and preparing them for re-review. This function is called, for example, when instruction 23_2C is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0374] The function "exit_review_loop_tool" is used to terminate the loop by the relevance evaluation agent AG23 when the quality criteria are met. This function is called, for example, when instruction 23_2C is executed. The code portion of this function is predefined, for example, in the analysis program 122 described above.

[0375] (I13. Report Generation Agent AG24) Next, with reference to Figure 28, the definition of the report generation agent AG24 shown in Figure 15 will be explained. Figure 28 is a diagram showing an example of the code definition related to the report generation agent AG24.

[0376] For the report generation agent AG24, the code shown in Figure 28 is defined for the argument parts 31A to 31E (see Figure 3) mentioned above.

[0377] More specifically, the argument 31A specifies the name of the report generation agent AG24. In the example in Figure 28, the name "report_generator_agent" is specified in the argument 31A.

[0378] The argument section 31B specifies the type of large-scale language model used by the report generation agent AG24. In the example in Figure 28, "Large-scale language model α" is specified in the argument section 31B. Note that the argument section 31B may specify the same large-scale language model as other AI agents, or it may specify a large-scale language model different from those used by other AI agents.

[0379] The argument section 31C specifies a description of the function of the report generation agent AG24. In the example in Figure 28, the function description specified in the argument section 31C is to generate a final structured report by integrating all data.

[0380] The argument section 31D specifies instructions for the report generation agent AG24. In the example in Figure 28, instructions 24A and 24B are specified in the argument section 31D.

[0381] Instruction 24A is an instruction to execute the function "generate_final_report_tool," which will be described later. This function retrieves data output from other AI agents.

[0382] Instruction 24B is an instruction to save the acquired data as a final report in a predetermined data format. For example, the data format may be Markdown. The final report may include, for example, a survey summary, a summary of the technical information 123 which is the invention under investigation, the aforementioned survey and design information 130, the search criteria for patent document 125, and the results of the relationship evaluation between technical information 123 and patent document 125.

[0383] In the argument section 31E, tools available to the report generation agent AG24 are specified. In the example of FIG. 28, the function "generate_final_report_tool" is specified in the argument section 31E. The said function is called, for example, when the instruction 24A is executed. The code portion of the said function is predefined, for example, in the above-described analysis program 122.

[0384] <J1. Others> Next, other examples of the above embodiment will be described.

[0385] In the above, the patent search process by the patent search agent AG21, the screening process by the screening agent AG22, the relevance evaluation process by the relevance evaluation agent AG23, and the report generation process by the report generation agent AG24 were executed in sequence. However, it is not necessary for all of the patent search agent AG21, the screening agent AG22, the relevance evaluation agent AG23, and the report generation agent AG24 to be implemented in the information processing apparatus 100.

[0386] As an example, there may be a case where a set of patent documents 125 to be investigated is determined. In such a case, the information processing apparatus 100 does not need to execute the patent search process by the patent search agent AG21 and the screening process by the screening agent AG22.

[0387] <J2. Others> Next, still other examples of the above embodiment will be described.

[0388] In the above, the root agent AG0 was specified to sequentially execute agents for initialization, investigation design, patent search, screening, review / quality evaluation, and report generation. However, depending on the investigation purpose 125, an instruction may be specified such that necessary agents are selectively configured according to the determination of the root agent AG0.

[0389] Specifically, the following instruction is specified in the argument part 31D: "When the investigation purpose is an invalid investigation, a rigorous evaluation of technical relevance is required, so it is desirable to execute agents for all processes. When the investigation purpose is a technology trend investigation, for macro analysis with cost reduction, it is desirable to skip the agents for scrutiny and quality evaluation." According to this instruction, the root agent AG0 can autonomously determine appropriate agents according to the investigation purpose and conduct a patent investigation.

[0390] Note that the instruction text for executing agents according to the above investigation purpose is not limited to invalid investigations and technology trend investigations, and instruction texts for executing agents according to various investigation purposes such as prior art investigations and infringement prevention investigations can be specified.

[0391] <J3. Others> Next, another example of the above embodiment will be described.

[0392] In the above, an example where the investigation purpose 125 is input by the user on the setting screen 400A has been described. However, the method for obtaining the investigation purpose 125 is not limited to this. As an example, when the investigation purpose is determined in advance to be one, the investigation purpose 125 may be stored in advance in a memory (for example, the auxiliary storage device 120) in the information processing apparatus 100. In this case, the instruction 40 of the investigation design agent AG10 is specified to obtain the investigation purpose 125 from the memory. Preferably, a function for obtaining the investigation purpose 125 is specified in the argument part 31E as a tool that the investigation design agent AG10 can use.

[0393] <J4. Others> Next, another example of the above embodiment will be described.

[0394] In the example of root agent AG0 shown in Figure 16 above, the argument section 31D specifies instructions to sequentially execute the patent search workflow. More specifically, the instructions specified to sequentially execute the patent search workflow as initialization, search design, patent search, screening, peer review / quality evaluation, and report generation. However, the instructions specified for root agent AG0 are not limited to this.

[0395] Figure 29 shows another example of the code definition for the root agent AG0. In the root agent AG0 according to this example, the argument section 31D specifies an instruction for the survey design agent AG10, which is responsible for generating the survey design information 130, to perform processing. In addition, in the root agent AG0 according to this example, the argument section 31D specifies an instruction for at least one of the following AI agents to perform processing: the patent search agent AG21, which is responsible for the patent search process; the screening agent AG22, which is responsible for the screening process; the relevance evaluation agent AG23, which is responsible for the relevance evaluation process; and the report generation agent AG24, which is responsible for the report generation process.

[0396] The instructions specified for root agent AG0 specify to the argument unit 31D that the processing related to patent search agent AG21, screening agent AG22, relevance evaluation agent AG23, and report generation agent AG24 should be executed as needed. Preferably, the instructions specified for root agent AG0 specify to the argument unit 31D that an AI agent corresponding to the input research objective should be selected from among patent search agent AG21, screening agent AG22, relevance evaluation agent AG23, and report generation agent AG24, and that the AI ​​agent should execute the processing.

[0397] The information processing device 100 inputs the technical information 123 and the research objective 125 to the root agent AG0. As a result, the root agent AG0 causes the research design agent AG10 to execute processing, and also causes at least one of the following AI agents to selectively execute processing: the patent search agent AG21, the screening agent AG22, the relevance evaluation agent AG23, and the report generation agent AG24.

[0398] In addition, in the root agent AG0 shown in Figure 16 above, instructions were given to have the initialization agent AG1, which is responsible for the initialization process, execute the processing. These instructions may or may not be given to the root agent AG0 shown in Figure 29.

[0399] Furthermore, the term "root agent" is not limited to a specific name, but is a general term for agents that have the function of overseeing the execution and management of other AI agents. For example, agents with names such as orchestrator agent, coordinator agent, master agent, management agent, or planner agent would qualify as "root agents" in this invention if they have equivalent functions. Similarly, other agents mentioned above, such as research and design agents and patent search agents, are not limited to specific names.

[0400] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the present invention is indicated by the claims rather than by the foregoing description, and all modifications within the meaning and scope equivalent to the claims are intended to be included. [Explanation of Symbols]

[0401] 1A instruction, 1B instruction, 1C instruction, 1D instruction, 10 information processing system, 10A instruction, 10B instruction, 10C instruction, 10D instruction, 21_1A instruction, 21_1B instruction, 21_1C instruction, 21_2A instruction, 21_2B instruction, 21_2C instruction, 21_2D instruction, 21_3A Instructions, 21_3B Instructions, 21_3C Instructions, 21_4A Instructions, 21_4B Instructions, 21_4C Instructions, 21_4D Instructions, 21_4E Instructions, 21_4F Instructions, 22A Instructions, 22B Instructions, 22C Instructions, 22D Instructions, 22E Instructions, 23_1A Instructions, 23_1B Instructions, 23_1C Instructions, 23_1D Instructions, 23_1E Instructions, 23_1F Instructions, 23_1G Instructions, 23_2A Instructions, 23_2B Instructions, 23_2C Instructions, 23_2D Instructions, 23_2E Instructions, 24A Instructions, 24B Instructions, 30 Code, 31A Argument Section, 31B Argument Section, 31C Argument Section, 31D Argument Section, 31E Argument Section, 31F Argument Section, 40 Instructions, 100 Information Processing Device, 101 Control Device, 102 ROM, 103 RAM, 104 Communication Interface, 105 Display Interface, 106 Display, 107 Input Interface, 108 Input Device, 110 Bus, 120 Auxiliary Storage Device, 122 Analysis Program, 123 Technical Information, 124 Patent Database, 125 Patent Documents, 126 Research Objectives, 130 Research and Design Information, 140 Report, 142 Information, 144 Information, 146 Information, 148 Information, 150 Information, 152 Relevance Evaluation Results, 158 Overall Evaluation Results, 160 Information, 162 Information, 200 User Terminal, 201 Control Device, 202 ROM, 203 RAM, 204 Communication Interface, 205 Display Interface, 206 Display, 207 Input Interface, 208 Input Device, 210 Bus, 220 Auxiliary Storage Device, 222 Analysis Program, 300 Server, 302 Information, 304 Information, 306 Information, 308 Information, 310 Information, 312 Information, 314 Information, 316 Information, 318 Information, 326 Relevance Evaluation Results, 327 Evaluation Results, 400A Settings Screen, 400B Execution Screen, 410 Input Field, 412 Input Field, 414 Input field, 416 Input field, 420 Start button, 430 Settings field, 432 Settings button, 434Settings button, 436 Settings button, 450 Approval request screen, 452 Button, 454 Button, 456 Button, AG Agent, AG0 Root Agent, AG1 Initialization Agent, AG10 Survey Design Agent, AG20 Patent Search Agent, AG21 Patent Search Agent, AG21_1 Extraction Agent, AG21_2 Condition Generation Agent, AG21_3 Search Execution Agent, AG21_4 Optimization Agent, AG22 Screening Agent, AG23 Relevance Assessment Agent, AG23_1 Iterative Review Agent, AG23_2 Quality Assessment Agent, AG24 Report Generation Agent, AGX Agent, MD Large-scale Language Model, NW Network, PR Program Code, TL Tool, TL1 Tool.

Claims

1. It is a program, The aforementioned program is installed on the computer. The process of acquiring technical information, including technical concepts, The process of obtaining the search objective related to the patent search, An AI (Artificial Intelligence) agent for patent search design, which is pre-instructed to output search design information that serves as a guideline for patent searches, is instructed to perform the process of inputting the aforementioned technical information and the aforementioned search objective. The aforementioned survey and design AI agent has a defined function description. The aforementioned function description stipulates that it generates search design information that other AI agents refer to in patent searches. The aforementioned program further provides the computer with: A program that inputs the aforementioned technical information and the aforementioned survey objectives into the survey design AI agent, and then executes a process to acquire the aforementioned survey design information related to the aforementioned technical information based on the results obtained from the survey design AI agent.

2. The instructions specified to the aforementioned survey design AI agent are: Instructions to analyze the entered invention from a specific perspective, The program according to claim 1, further comprising an instruction to output the analysis results as the survey design information.

3. The program according to claim 2, wherein the aforementioned specific viewpoint includes at least one of the following: a technical viewpoint, a problem viewpoint, an effect viewpoint, a feature viewpoint of the invention, a component viewpoint, and a patent classification viewpoint.

4. The program according to claim 1 or 2, wherein the instructions specified to the survey design AI agent include instructions to output survey design information in a predetermined format.

5. The program according to claim 1 or 2, wherein the type of large-scale language model to be used is specified for the survey design AI agent.

6. The aforementioned survey and design AI agent has program functions defined as tools. The program according to claim 1 or 2, wherein the program function is a function for acquiring the technical information.

7. The aforementioned program further provides the computer with: The process of inputting the aforementioned search design information to a patent search AI agent that has been instructed in advance to perform the process related to patent searches, The program according to claim 1 or 2, which inputs the aforementioned survey design information into the patent search AI agent and performs a process of obtaining results related to the patent search based on the results obtained from the patent search AI agent.

8. The aforementioned patent search AI agent includes a patent search AI agent that has been instructed in advance to perform a patent search, The aforementioned program further provides the computer with: The process of inputting the survey and design information to the aforementioned patent search AI agent, The program according to claim 7, which inputs the aforementioned survey and design information into the patent search AI agent and, based on the results obtained from the patent search AI agent, performs the process of obtaining a set of patent documents to be compared with the aforementioned technical information.

9. The aforementioned program further provides the computer with: The process is executed to obtain the set of patent documents that are the subject of the aforementioned patent search. The aforementioned patent search AI agent includes a screening AI agent that is pre-specified to remove less relevant patent documents. The aforementioned program further provides the computer with: The process of inputting the survey design information and the set of patent documents to the screening AI agent, The program according to claim 7, wherein the program inputs the aforementioned survey and design information and the set of patent documents to the screening AI agent, and based on the results obtained from the screening AI agent, it performs a process of obtaining a set of patent documents from which patent documents with low relevance to the technical information have been removed.

10. The aforementioned program further provides the computer with: The process is executed to obtain the set of patent documents that are the subject of the aforementioned patent search. The aforementioned patent search AI agent includes a relevance evaluation AI agent that has been pre-specified to evaluate relevance, The aforementioned program further provides the computer with: The process involves inputting the aforementioned research design information and the aforementioned set of patent documents to the aforementioned relevance evaluation AI agent, The program according to claim 7, wherein the program inputs the aforementioned survey and design information and the set of patent documents to the relationship evaluation AI agent, and then performs a process to obtain an evaluation result showing the relationship between each of the patent documents and the technical information based on the results obtained from the relationship evaluation AI agent.

11. The aforementioned program further provides the computer with: A process for inputting the survey design information and the evaluation results to a report generation AI agent that has been instructed in advance to output data in a predetermined format, The program according to claim 10, which inputs the aforementioned survey design information and the aforementioned evaluation results into the report generation AI agent, and then performs the process of obtaining a patent search report relating to the aforementioned technical information based on the results obtained from the report generation AI agent.

12. The aforementioned patent search AI agent is composed of multiple AI agents, each of which is instructed to perform a different process related to patent searching. The process involves inputting the technical information and the survey objective to the aforementioned survey design AI agent and a root AI agent that has been instructed to perform processing on at least one of the plurality of AI agents, The program according to claim 7, wherein the technical information and the investigation objective are input to the root AI agent, and based on the results obtained from the root AI agent, the investigation design AI agent and at least one of the plurality of AI agents are instructed to perform processing.

13. The program according to claim 12, wherein the instructions specified for the root AI agent include instructions to select an AI agent according to the purpose of the investigation and to cause the AI ​​agent to perform processing.

14. The aforementioned program further provides the computer with: The process of accepting the target number of patent searches, The program according to claim 8, which causes the patent search AI agent to repeat the patent search until the number of patent documents hit by the patent search AI agent falls within a target range based on the target number.

15. The program according to claim 14, wherein the patent search AI agent is instructed to change the search conditions to narrow the search range if the number of hits exceeds the upper limit of the target range.

16. The program according to claim 14, wherein the patent search AI agent is instructed to change the search conditions to broaden the search range if the number of hits falls below the lower limit of the target range.

17. The aforementioned program further provides the computer with: A process for receiving a setting regarding whether or not approval is required for the output results of the aforementioned survey design AI agent, The program according to claim 1 or 2, which, if the setting indicates that approval is required, performs a process of displaying an approval request screen that accepts whether or not to approve the survey design information based on the fact that the survey design information has been generated.

18. The aforementioned approval request screen is The operation to instruct approval, It is configured to accept commands to request a redo. The aforementioned program further provides the computer with: When the operation to instruct the aforementioned approval is received, the survey and design information acquired in the acquisition process is output. The program according to claim 17, which, upon receiving the operation to instruct the aforementioned redo, causes the survey and design AI agent to regenerate the survey and design information.

19. The aforementioned approval request screen is further configured to accept operations that instruct editing, The aforementioned program further provides the computer with: The program according to claim 18, which, upon receiving an operation to instruct the aforementioned editing, accepts an operation to edit the survey design information and executes a process to output the edited survey design information.

20. The aforementioned program further provides the computer with: A process for receiving a setting regarding whether or not approval is required for the output results of the aforementioned patent search AI agent, The program according to claim 7, which, if the setting indicates that approval is required, causes the program to execute a process that displays an approval request screen that accepts whether or not to approve the results based on the fact that the results relating to the patent search have been generated.

21. An information processing device, Equipped with a control unit, The control unit, The process of acquiring technical information, including technical concepts, The process of obtaining the search objective related to the patent search, The AI ​​agent for patent search design, which is instructed in advance to output search design information that serves as a guideline for patent searches, is given the task of inputting the aforementioned technical information and the aforementioned search objective. The aforementioned survey and design AI agent has a defined function description. The aforementioned function description stipulates that it generates search design information that other AI agents refer to in patent searches. The control unit further, An information processing device that inputs the aforementioned technical information and the aforementioned investigation objectives into the aforementioned investigation and design AI agent, and then performs a process to acquire the aforementioned investigation and design information related to the aforementioned technical information based on the results obtained from the aforementioned investigation and design AI agent.

22. A method performed by a computer, Steps to acquire technical information, including technical concepts, Steps to obtain the search objective related to the patent search, The system includes a step of inputting the technical information and the search objective to a search design AI agent that has been instructed in advance to output search design information that serves as a guideline for patent searches. The aforementioned survey and design AI agent has a defined function description. The aforementioned function description stipulates that it generates search design information that other AI agents refer to in patent searches. A method comprising the step of inputting the aforementioned technical information and the aforementioned investigation objective into the aforementioned investigation and design AI agent, and acquiring the aforementioned investigation and design information related to the aforementioned technical information based on the results obtained from the aforementioned investigation and design AI agent.

Images