Program, information processing device, method, and system

Abstract

To enable even a user with little knowledge about carbon credits to easily and accurately grasp the validity of a plan at the early stage of planning a carbon credit project. [Solution] A program to be executed on a computer having a processor and memory, the program causing the processor to perform the following steps: accepting input from a user of geographic information of a candidate site for a carbon credit project; obtaining information about other carbon credit projects from a knowledge base based on the accepted geographic information; inputting the geographic information and information about other carbon credit projects into a generative AI model and causing the generative AI model to output analytical information indicating the appropriateness of planning to implement a carbon credit project in the candidate site; and presenting the analytical information to the user.

Description

[Technical Field]

[0001] The present disclosure relates to a program, an information processing device, a method, and a system. [Background technology]

[0002] In recent years, research and development into technologies to support carbon credit projects has become more active. For example, Patent Document 1 discloses a technology in which a business plan for acquiring carbon credits and evaluation criteria conforming to carbon credit certification standards are given to a generation AI as prompts, an evaluation level for each evaluation criterion for the business plan is output, and a final evaluation is made based on the obtained evaluation levels. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7644302 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the technology disclosed in Patent Document 1 evaluates a business plan on the premise that the land to be used for the greenhouse gas emission reduction project has already been decided, and does not evaluate, for example, whether a candidate site for the project is suitable for implementation. Therefore, there is room for improvement in the technology disclosed in Patent Document 1 in terms of evaluating the validity of the plan at a stage before the business plan is created.

[0005] An object of the present disclosure is to enable even a user with little knowledge about carbon credits to easily and accurately grasp the validity of a plan in the early stages of planning a carbon credit project. [Means for solving the problem]

[0006] In order to solve the above-mentioned problem, a program according to one aspect of the present disclosure is a program to be executed by a computer having a processor and a memory, and the program causes the processor to perform the following steps: accepting input of geographic information of a candidate site for a carbon credit project from a user; obtaining information about other carbon credit projects from a knowledge base pre-constructed for evaluating the appropriateness of planning to implement the carbon credit project in the candidate site based on the accepted geographic information; inputting the geographic information and information about other carbon credit projects into a generative AI model and causing the generative AI model to output analytical information regarding the appropriateness of planning to implement the carbon credit project in the candidate site; and presenting the analytical information to the user. [Effects of the Invention]

[0007] According to the present disclosure, even a user with little knowledge about carbon credits can easily and accurately grasp the validity of a plan in the early stages of planning a carbon credit project. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a block diagram showing an example of the overall configuration of a system 1. FIG. [Figure 2] 1 is a block diagram showing an example of the configuration of a terminal device 10. FIG. [Figure 3] FIG. 2 is a block diagram showing an example of the configuration of a server 20. [Figure 4] FIG. 2 is a diagram showing the data structure of a related information database 2021. [Figure 5] 10 is a flowchart showing an example of the operation of the server 20. [Figure 6] FIG. 10 is a schematic diagram illustrating an example of a screen according to an embodiment of the present disclosure. [Figure 7] FIG. 2 is a block diagram showing the basic hardware configuration of a computer 90. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In all drawings describing the embodiments, common components are designated by the same reference numerals, and repeated explanations will be omitted. Note that the following embodiments do not unduly limit the content of the present disclosure described in the claims. Furthermore, not all components shown in the embodiments are necessarily essential components of the present disclosure. Furthermore, each drawing is a schematic diagram and is not necessarily a precise illustration.

[0010] [1. Overview] The system according to this embodiment accepts input of geographic information of candidate sites for carbon credit projects from a user. The geographic information includes, for example, address information of the candidate sites and polygon data of the candidate sites. The system according to this embodiment acquires related information from a pre-constructed knowledge base based on the accepted geographic information. The related information is information about other carbon credit projects related to the candidate sites. The system according to this embodiment inputs the geographic information and related information into a generative AI model such as a Large Language Model (LLM), and outputs validity information from the generative AI model. The validity information is analytical information regarding the validity of planning to implement a carbon credit project in the candidate sites. The system according to this embodiment presents the validity information to the user.

[0011] 2. First Embodiment <2-1 Overall system configuration> FIG. 1 is a block diagram showing an example of the overall configuration of system 1. System 1 is, for example, a system for providing a validity assessment service. The validity assessment service is, for example, a service that evaluates the validity of planning to implement a carbon credit project in a candidate site (hereinafter referred to as the validity of planning). A candidate site is, for example, a specific geographical area (e.g., land, location) that a user of the validity assessment service is considering as a target for implementing a carbon credit project or wants to know information about.

[0012] There are no particular limitations on the types of carbon credit projects and methodologies (hereinafter referred to as "target projects") that are eligible for the validation service. Methodologies, for example, stipulate how a carbon credit project calculates reductions or removals and how carbon credits can be certified and issued. Specifically, the scope of application, calculation methods for reductions / removals, monitoring methods, etc. are defined for each type of project (e.g., renewable energy, forest protection, energy conservation, etc.). The following are examples of methodologies that are eligible for the validation service. ARR (Afforestation, Reforestation and Revegetation): Afforestation, reforestation and revegetation REDD+ (Reducing Emissions from Deforestation and Forest Degradation, and the Role of Conservation, Sustainable Management of Forests, and Enhancement of Forest Carbon Stocks in Developing Countries): Reducing emissions from deforestation and forest degradation in developing countries, and conserving and sustainable forest management and enhancing forest carbon stocks. AWD (Alternate Wetting and Drying): Intermittent irrigation. WRC (Wetlands Restoration and Conservation): Wetlands restoration and conservation ·ALM(Agricultural Land Management): Farmland management

[0013] In the above example, for example, the type of carbon credit project to which ARR belongs is an absorption / removal project (nature-based). Also, for example, the type of carbon credit project to which REDD+ belongs is an emission reduction / avoidance project (forest conservation). Also, for example, the type of carbon credit project to which AWD belongs is an emission reduction / avoidance project (agriculture). Furthermore, the type of carbon credits (e.g., forest credits, etc.) planned to be generated from the candidate site will vary depending on, for example, the target project, etc.

[0014] The candidate sites subject to the validity assessment service are not limited to land in Japan, but may also be land in other countries. In this case, the target projects will conform to the types of carbon credits and certification standards established by, for example, Verra, Gold Standard, JCM (Joint Crediting Mechanism), etc.

[0015] The system 1 can provide various services as validity assessment services. Below, several examples of services that can be considered as validity assessment services are described. However, the validity assessment services are not limited to the following examples.

[0016] For example, the system 1 can evaluate the eligibility of a candidate site as part of the validity assessment service and present the evaluation results to the user. Eligibility is a concept that indicates, for example, the degree to which a candidate site is suitable as land for the creation of carbon credits. Eligibility is determined, for example, according to the degree to which the candidate site satisfies the eligibility conditions that indicate eligibility as land for the creation of carbon credits. There are no particular limitations on the content of the eligibility conditions, and they can be set flexibly depending on the content of the target project, etc. Below, the eligibility conditions are illustrated together with specific examples of methodology.

[0017] (1) ARR The conditions specified in this methodology are mainly related to the historical condition of the land and the nature of the project activities. For example, according to Verra's methodology VM0047, the following conditions are listed: Land history requirements: The proposed project area must not have been a "managed forest" at any time during the past 10 years prior to the project start date. The proposed area must not be a mangrove forest or a tidal wetland such as a salt marsh. · Existing Biomass Condition: Removal of existing woody biomass prior to project activities must not involve commercial timber harvesting or degradation of native ecosystems. Conditions regarding planting density and monitoring methods: for example, planting density must exceed 50 planting units per hectare. This is intended to exclude census-based monitoring methods that may be applied when planting densities are low. ·Conditions for elimination of overlap: The candidate area must not overlap with the project area of ​​any other VCS (Verified Carbon Standard) AFOLU (Agriculture, Forestry and Other Land Use) project.

[0018] (2) REDD+ The specified conditions in this methodology are mainly related to conservation activities of existing forests. Land classification criteria: The proposed area must be an area that has been classified as "forest" for the past 10 years. Conditions on the nature of the activity: Project activities may cover activities that are classified as Improved Forest Management (IFM) or REDD+ per se, and may be outside the scope of other specific methodologies, for example if the baseline scenario includes commercial forestry. Land history conditions: It must be confirmed through image data, etc., before the project is planned that there has been no deforestation or removal of woody biomass that would lead to forest degradation in the past 10 years. ·Conditions regarding elimination of overlap: It is required that the project does not overlap with the implementation area of ​​other carbon credit projects.

[0019] (3) AWD The conditions specified in this methodology include a wide range of technical requirements regarding land history, water management, cultivation activities, and additionality. For example, according to individual methodologies such as Verra, Gold Standard (GS), or JCM (Joint Crediting Mechanism), the following conditions may be included: Land history and classification criteria: Candidate areas must have been non-forested for the past 10 years and have a woody biomass coverage of less than 10%. Alternatively, candidate areas must be classified as "continuously used agricultural land," "residential land," or "other land" under the IPCC land use categories. In particular, rain-fed rice paddies, deep-water paddies, and non-irrigated lowland rice paddy systems such as farmland are excluded. Conditions for mandatory and optional activities: Mandatory activities include the introduction of improved irrigation management to reduce methane production, such as single-period drainage, intermittent irrigation (AWD), reduced flooding period, or direct seeding rice (DSR). Optional activities include the use of methanotrophs, avoidance of rice straw burning, improved nitrogen management (e.g., slow-release fertilizer), and biochar application. Water Management and Infrastructure Technical Requirements: Project rice fields must be equipped with controllable irrigation and drainage systems that can be designed to maintain flooded and non-flooded conditions. "Drainage" here is technically defined as "complete" when the water level reaches 15 cm below the soil surface, or when the water level remains between 0 cm and -15 cm below the soil surface for a total of 10 days, including at least three consecutive days. Re-irrigation may be required within two days of drainage completion to prevent yield loss. Conditions regarding additionality and regulations: The drainage implemented in the project must not be required by local laws, etc. Also, it must be proven that AWD is not already widespread in the local government (e.g., state level) that includes the project area (e.g., agricultural census, expert certification, remote sensing data, etc., confirming that the implementation rate is less than 20%). · Conditions on support to farmers: Project activities must include training and technical assistance to provide farmers with appropriate knowledge on field preparation, irrigation, drainage, and fertilization. Exclusion criteria: Activities that significantly reduce soil organic carbon (SOC) (e.g., increasing the amount of rice straw removed and introducing varieties with significantly less root mass) are not permitted. In addition, projects that change management practices outside of the rice-growing season (off-season) are not eligible, with the exception of avoiding post-harvest burning of crop residues.

[0020] (4) WRC This methodology is particularly applicable to restoration projects of tidal wetlands such as mangrove forests. For example, according to the Verra methodology VM0033 or GS methodology, the following conditions are set out: Land history and classification criteria: The candidate area must be a natural mangrove forest more than 10 years ago, but is now a non-mangrove area. It must also be a coastal area with an elevation of 0 to 50 meters, as confirmed using satellite data (e.g., SRTM DEM). Examples of eligible activities include "restoring hydrological conditions by removing tidal barriers," "modifying sediment supply by utilizing dredged material," "reintroducing native plant communities by sowing seeds or planting seedlings," and "removing invasive alien species." Land use conditions (leakage prevention): It must be proven that the land use prior to the start of the project has been abandoned for more than two years, or that commercial use is no longer profitable due to salt intrusion, etc., and that emissions from conversion of activities will not occur outside the area. - Ineligible conditions: Project activities such as "if they cause a decline in the groundwater level," "if nitrogen fertilizer is applied within the project area," or "if the baseline includes commercial forestry" are not covered by this methodology.

[0021] Eligibility criteria may consist of, for example, a first condition that indicates the status or condition of the candidate site at a given point in time, and a second condition whose fulfillment can only be determined once the site's status or condition changes over time are known. In the example above, the first condition would be "not overlapping with the implementation area of ​​other carbon credit projects," "having controllable irrigation and drainage facilities," "the candidate site's elevation must be between 0 and 50 meters," and "no nitrogen fertilizer application as part of the project activities." Meanwhile, the second condition would be "not being a 'managed forest' for the past 10 years" in the case of ARR, "not having multiple drainage events in the past two years prior to the project start" in the case of AWD, or "having been a natural mangrove forest for more than 10 years" in the case of WRC.

[0022] For example, among candidate sites, an area that satisfies both the first and second conditions during a specified period in the past is evaluated as an area eligible for the implementation of a carbon credit project. However, even if the specified period in the past includes a period in which the first condition is not satisfied, the area may also be evaluated as eligible if the period and degree to which the second condition is not satisfied is equal to or less than a specified threshold (which can be set arbitrarily).

[0023] Furthermore, for example, the system 1 can generate methodology information as part of the content of the validity assessment service and present it to the user. The methodology information is, for example, information on methodologies applicable to the candidate site. The methodology information includes, for example, a list of areas in the candidate site that have been assessed as eligible by the above-mentioned eligibility assessment (each area is associated with a type of applicable methodology), proposals regarding applicable methodologies (e.g., types of trees to be planted if ARR is applicable, specific forest conservation techniques if REDD+ is applicable, etc.), past application examples of methodologies applicable to the candidate site, etc.

[0024] Alternatively, the system 1 can generate forecast information as part of the validity assessment service and present it to the user. The forecast information is, for example, information showing various forecast results assuming that a carbon project is implemented in a candidate site. The forecast information includes, for example, in the case of forest credits, a forecast value of carbon stocks based on an estimate of aboveground biomass volume, in the case of AWD, a forecast value of greenhouse gas (methane gas) emission reductions based on an estimate of non-flooded periods, and a revenue forecast based on an estimate of the amount of carbon credits. This information may be generated for, for example, multiple scenarios (e.g., different tree species, different management methods for the candidate site, different market price fluctuation trends that serve as the basis for the revenue forecast, etc.).

[0025] The results of the eligibility assessment, methodology information, and forecast information are all useful information for assessing the validity of the plan.

[0026] 1 includes, for example, a terminal device 10, a server 20, a satellite 30, and an AI system 40. The terminal device 10, the server 20, and the AI ​​system 40 are communicatively connected, for example, via a network 80. The satellite 30 transmits, for example, various types of satellite data to a ground station (not shown). The ground station is communicatively connected to the network 80, and, for example, receives a transmission request from the server 20 and transmits the satellite data to the server 20 via the network 80.

[0027] 1 shows an example in which the system 1 includes one terminal device 10, but for example, the system 1 may include two or more terminal devices 10. In FIG. 1, an example in which the system 1 includes one server 20 is shown, but for example, a collection of multiple devices may be one server 20. The way in which the multiple functions required to realize the server 20 are allocated to one or multiple pieces of hardware can be determined appropriately depending on the processing capacity of each piece of hardware and / or the specifications required for the server 20, etc.

[0028] While FIG. 1 shows an example in which system 1 includes one AI system 40, system 1 may include two or more AI systems 40. Also, while FIG. 1 shows an example in which AI system 40 is independent from server 20, server 20 may include the functions of AI system 40. In other words, server 20 may store a generative AI model (details of which will be described later) included in AI system 40.

[0029] The terminal device 10 is, for example, an information processing device operated by a user who uses the validity assessment service. The user is, for example, a person who plans to implement a carbon credit project in a candidate site, and can be anyone with various attributes and backgrounds, regardless of the presence or absence and level of knowledge about carbon credits.

[0030] The terminal device 10 is realized by, for example, a mobile terminal such as a smartphone or a tablet. In this embodiment, the terminal device 10 is assumed to be a smartphone. The terminal device 10 may also be realized by, for example, a desktop personal computer (PC), a laptop PC, or the like.

[0031] The terminal device 10 includes a communication IF (Interface) 12, an input device 13, an output device 14, a memory 15, a storage 16, and a processor 19. The input device 13 is a device for receiving input operations from a user (for example, a touch panel, a touch pad, a pointing device such as a mouse, a keyboard, etc.). The output device 14 is a device for presenting information to a user (a display, a speaker, etc.). In this embodiment, the terminal device 10 is assumed to include a touch panel in which the input device 13 and the output device 14 are integrated.

[0032] The server 20 is, for example, an information processing device for managing and operating the validity assessment service, and is an information processing device realized by a computer connected to a network 80. As shown in Fig. 1, the server 20 includes a communication IF 22, an input / output IF 23, a memory 25, a storage 26, and a processor 29. The input / output IF 23 functions as an input device for receiving input operations from a person in charge of the validity assessment service, and as an interface for an output device for outputting information to the person in charge.

[0033] The artificial satellite 30, for example, acquires satellite images (hereinafter abbreviated as satellite images) of a candidate site as satellite data and transmits them to a ground station. The satellite images are used, for example, when the system 1 performs eligibility assessment, generation and presentation of methodology information, etc. FIG. 1 shows an example in which the system 1 includes one artificial satellite 30, but the system 1 may include, for example, two or more artificial satellites 30. When two or more artificial satellites 30 are included, the types of the artificial satellites 30 may be the same or different from each other. Furthermore, for example, if the content of the validity assessment service does not include eligibility assessment, generation and presentation of methodology information, etc., the system 1 may not include an artificial satellite 30.

[0034] In this embodiment, the satellite 30 acquires optical satellite images (e.g., visible light images, near-infrared images, etc.) as satellite images and transmits them to the server 20 via a ground station. Note that the satellite images transmitted by the satellite 30 to the server 20 via a ground station may be, for example, SAR (Synthetic Aperture Radar) images, spectral satellite images, or topographical satellite images (DEM: Digital Elevation Model), etc.

[0035] The AI ​​system 40 is, for example, a system equipped with a generative AI model and for generating and outputting validity information. The validity information is, for example, analytical information regarding the validity of planning to implement a carbon credit project in a candidate site, and includes at least one of information indicating the results of an evaluation of the candidate site's eligibility (hereinafter referred to as eligibility information), methodological information, and forecast information. However, the validity information may include information other than the eligibility information, methodological information, and forecast information, or may consist solely of such other information.

[0036] Specifically, justification information refers to a structured set of data that includes at least the following three information elements that correspond to logical evaluation steps in the project lifecycle:

[0037] (1) Information on normative and regulatory compliance This information is used to determine whether the prerequisites for project implementation are met. Specifically, it is equivalent to the result of an evaluation of whether a project is "feasible" according to the rules, based on the land history of the candidate site, legal regulations, international standards, and eligibility requirements set out in specific methodologies. It can also be called a compliance evaluation.

[0038] (2) Information about the project implementation plan This information presents a strategy for how a project should be implemented. After passing an evaluation of normative and regulatory compliance, it is equivalent to proposing a specific action plan, such as selecting the applicable methodology that is most suitable for the characteristics of the candidate site (soil, vegetation, climate, etc.), the tree species to plant, and the conservation methods to be adopted.

[0039] (3) Information regarding quantitative forecasts of project results This is information that quantitatively indicates "what results can be expected" if a project is implemented. It corresponds to the predicted amount of carbon credits to be generated in the future based on the selected implementation plan, and the associated profitability. It may also include simulation results that take into account multiple scenarios (e.g., fluctuations in market prices, climate change risks, etc.).

[0040] Furthermore, there are no particular limitations on the output format and output manner of the validity information. For example, the validity information may be the eligibility information, methodology information, and prediction information each output in text format, or the eligibility information may be output as polygon data of the eligible area. For example, the validity information may be presented in a report or in a chart such as a dashboard.

[0041] The AI ​​system 40 is physically built on, for example, one or more server computers. The AI ​​system 40 is connected to the server 20 via a network 80 and provides each function through an API (Application Programming Interface). The AI ​​system 40 may be built on, for example, a cloud-based infrastructure.

[0042] The generative AI model may be, for example, an LLM, or a multimodal generative AI model that can process text data, image data, audio data, etc. in an integrated manner. In this embodiment, the AI ​​system 40 is assumed to include an LLM (not shown) as the generative AI model. The number of LLMs included in the AI ​​system 40 may be one or more.

[0043] LLM is a single-modal natural language model constructed by learning from large-scale text data, and is used in many NLG (Natural Language Generation) tasks such as generating answers to specific questions, automatically generating sentences, summarizing text data, etc. Examples of LLMs include the following: OpenAI: GPT-4 Google: Gemini 1.5 Flash ·Anthropic: Claude 3.5 Sonnet Specifically, for example, the AI ​​system 40 inputs geographic information and related information of the candidate site together with a prompt into the LLM, and causes the LLM to output validity information. The prompt includes, for example, an instruction statement instructing the output of validity information, and a predetermined template is stored in the server 20 or the AI ​​system 40. Note that the AI ​​system 40 may generate a prompt whose content corresponds to, for example, the geographic information and related information.

[0044] The geographic information includes, for example, address information indicating the address of the candidate site, map data of the candidate site such as polygon data, and location information indicating the latitude and longitude of the candidate site. The address information and location information may be, for example, text data or audio data. The related information is, for example, information about other carbon credit projects related to the candidate site, and is stored in the related information database 2021 described below. Note that the related information may also be stored, for example, in a database of an external system.

[0045] The related information includes, for example, at least one of regulatory information, certification standards, past cases, and information on similar areas. Regulatory information includes, for example, the laws of the country and region where the candidate site is located, or international guidelines that must be observed when implementing a carbon credit project. Certification standards include, for example, specific conditions and rules for officially certifying carbon credits created when a carbon credit project is implemented in the candidate site. Past cases include, for example, plans, implementation reports, outcome data, and survey reports of similar projects implemented in the past. Information on similar areas includes, for example, various information on other lands and places with environmental conditions (e.g., climate, topography, vegetation, etc.) similar to those of the candidate site. Note that the related information may include information other than regulatory information, certification standards, past cases, and information on similar areas, or may consist solely of such other information.

[0046] In this embodiment, the AI ​​system 40 functions as an AI agent that acquires address information, autonomously plans and executes a series of tasks, and outputs validity information. That is, for example, upon receiving address information from the AI ​​system 40, the LLM accesses the related information database 2021 to read and acquire related information. The LLM, for example, reads the acquired related information, analyzes it together with the address information previously received, and determines and evaluates the validity of the plan based on the analysis results. The LLM, for example, converts the evaluation results into an appropriate output format depending on the content of the evaluation results and the type of target project, and outputs the result as validity information.

[0047] The related information database 2021 is an example of a knowledge base according to one embodiment of the present disclosure. The knowledge base is a collection of specialized information assets implemented in various formats for query by the LLM, specializing in evaluating the appropriateness of planning a carbon credit project in a candidate site. Specifically, the knowledge base contains specialized information necessary for the evaluation, such as regulations, certification standards, methodologies, past cases, and geospatial information. The knowledge base also integrates various data formats, such as structured data, unstructured data, and spatiotemporal data, and includes any implementation that can be queried by the LLM, such as a vector database, a relational database, a graph database, a set of APIs, and a training dataset.

[0048] The related information database 2021 in this embodiment is constructed as follows (vector database) to maximize the performance of RAG (Retrieval-Augmented Generation). First, various collected documents (e.g., PDF documents of Verra's methodology VM0047, texts of forest-related laws and regulations of various countries, etc.) are divided into chunks of a predetermined number of tokens (e.g., 512 tokens) while maintaining semantic cohesion. Next, each chunk is converted (embedded) into a high-dimensional vector (e.g., 1024 dimensions) using a pre-trained language model (e.g., Sentence-BERT, text-embedding-3-large, etc.). This vector data is stored in a vector database (e.g., Pinecone, ChromaDB) that can perform high-speed similarity searches between vectors. At this time, metadata such as the source (document name, page number) and information type (regulatory information, methodology, etc.) are assigned to each vector.

[0049] The LLM in this embodiment is not a general-purpose model that is used as is, but is tuned by combining the following methods to specialize it for the task of carbon credit evaluation.

[0050] (1) Instruction Fine Tuning We prepare thousands of high-quality datasets created by carbon credit experts, each consisting of a triplet of "Query," "Context," and "Answer." For example, the dataset might look like this: "Query: If a candidate site has not been forested for the past 15 years, does it meet the eligibility criteria of VM0047?", "Context: (Relevant clause of VM0047)," and "Answer: Yes, it does. VM0047 requires that the site must not have been forested for 10 years prior to the start of the project, and 15 years will meet this requirement." Fine-tuning the LLM using this dataset will improve its ability to generate accurate inferences and answers based on the provided context.

[0051] (2) RLHF (Reinforcement Learning with Human Feedback) After fine-tuning, the model generates multiple draft evaluation reports, which are then reviewed by experts, who then rank them based on their accuracy, completeness, clarity, etc. This ranking information is used to train the reward model, which, through reinforcement learning, further adjusts the behavior of the LLM to produce the quality output preferred by the experts.

[0052] When a user inputs geographic information, the server 20 instructs the LLM, which operates as an AI agent, to "create a carbon credit project validity assessment report related to this geographic information" as its final goal. The LLM repeats the following cycle of thinking and acting based on a framework such as ReAct (Reasoning and Acting). · Thought: What should be done first? Based on the geographical characteristics of the candidate site, it is necessary to determine the applicable methodology (e.g. forestry or agriculture). Action: Perform a vector search on the related information database using keywords such as "forest conservation methodology" and "paddy field methane reduction methodology." Observation: The search results included chunks of methodological papers on REDD+, ARR, AWD, etc. Thought: Next, we need to check the eligibility conditions of these methodologies one by one. First, we need to look at the land history conditions for REDD+. Action: Search the database with a query such as "REDD+ land history conditions past 10 years." At the same time, access an external API (satellite data provider) and request land cover data for the candidate site for the past 10 years. Observation: Repeat the cycle until all evaluation items are completed.

[0053] In this way, the LLM autonomously creates plans and collects and analyzes information while making full use of tools (database searches, API execution), ultimately generating a comprehensive and reliable evaluation report (see Figure 6).

[0054] In other words, in this embodiment, the AI ​​system 40 executes a search query based on geographic information on a pre-vectorized related information database 2021, which is composed of multiple information categories including regulatory information, certification standards, and methodology papers related to carbon credit projects, to acquire information highly relevant to the geographic information (search processing in RAG). The AI ​​system 40 also inputs the acquired related information as context into the LLM together with the geographic information, and causes the LLM to output structured validity assessment information including at least "eligibility assessment," "applicable methodologies," and "carbon credit generation amount forecast" for the candidate site (generation processing in RAG).

[0055] In order for the AI ​​system 40 to function as an AI agent, the LLM is tuned in advance, for example, using a large amount of past cases as learning data. There are no particular limitations on the tuning method, and it may be, for example, fine tuning or prompt tuning. Furthermore, for example, the LLM may be tuned through MCP (Model Context Protocol) collaboration. In MCP collaboration, a Model unit that thinks and plans, a Controller unit that executes the plan, and a Parser unit that analyzes information collaborate to enable autonomous task execution.

[0056] It is not essential that the AI ​​system 40 function as an AI agent, and the AI ​​system 40 may function, for example, as a response generation tool that automatically generates a response (validity information) in response to input of geographic information and related information.

[0057] Each information processing device, such as the terminal device 10, the server 20, and the AI ​​system 40, is configured by a computer 90 (see FIG. 7) equipped with an arithmetic unit and a storage device. The basic hardware configuration of the computer 90 and the basic functional configuration of the computer 90 realized by the basic hardware configuration will be described later. Note that for each of the terminal device 10, the server 20, and the AI ​​system 40, descriptions that overlap with the basic hardware configuration and basic functional configuration of the computer 90 will be omitted.

[0058] <2-2 Terminal Device Configuration> Fig. 2 is a block diagram showing an example configuration of the terminal device 10 shown in Fig. 1. As shown in Fig. 2, the terminal device 10 includes a communication unit 120, an input device 13, an output device 14, an audio processing unit 170, a microphone 171, a speaker 172, a camera 160, a position information sensor 150, an acceleration sensor 155, a storage unit 180, and a control unit 190. The blocks included in the terminal device 10 are electrically connected by, for example, a bus or the like.

[0059] The communication unit 120 performs processing such as modulation and demodulation for communication between the terminal device 10 and an external device (for example, the server 20). The communication unit 120 performs transmission processing on the signal generated by the control unit 190 and transmits it to the external device. The communication unit 120 performs reception processing on the signal received from the external device and outputs it to the control unit 190.

[0060] The input device 13 is a device for a user to input instructions or information. The input device 13 is realized, for example, by a touch-sensitive device 131 that inputs instructions by touching an operation surface. When the terminal device 10 is a PC or the like, the input device 13 may be realized by a reader, keyboard, mouse, or the like. The input device 13 converts instructions input by the user into electrical signals and outputs them to the control unit 190. The input device 13 may include, for example, a receiving port that receives electrical signals input from an external input device.

[0061] The output device 14 is a device for presenting information to a user. The output device 14 is realized, for example, by a display 141. The display 141 displays various types of information according to the control of the control unit 190. The display 141 is realized, for example, by an LCD (Liquid Crystal Display) or an organic EL (Electro-Luminescence) display.

[0062] The audio processing unit 170 performs, for example, digital-to-analog conversion processing of an audio signal. The audio processing unit 170 converts a signal provided from the microphone 171 into a digital signal and provides the converted signal to the control unit 190. The audio processing unit 170 also provides the audio signal to the speaker 172. The audio processing unit 170 is realized, for example, by a processor for audio processing. The microphone 171 receives audio input and provides an audio signal corresponding to the audio input to the audio processing unit 170. The speaker 172 converts the audio signal provided from the audio processing unit 170 into audio and outputs the audio to the outside of the terminal device 10.

[0063] Camera 160 is an imaging device that captures images using visible light. In other words, camera 160 is a device that receives visible light using a light-receiving element and outputs image data as an image capture signal. Camera 160 captures an image of a subject in a certain direction and within a certain image capture range relative to terminal device 10, and outputs image data as the image capture result. If camera 160 has a function that allows the image capture range, or more precisely, the angle of view, to be adjustable, camera 160 also outputs information regarding this angle of view. This function is known as a zoom function.

[0064] The position information sensor 150 is a sensor that detects the position of the terminal device 10 and is generally a GNSS device, such as a GPS module. The GPS module is a receiving device used in a satellite positioning system. In a satellite positioning system, signals are received from at least three or four satellites, and the current position of the terminal device 10 equipped with the GPS module is detected as coordinate values ​​based on the received signals. The position information sensor 150 may detect the current position of the terminal device 10 from the position of a wireless base station to which the terminal device 10 connects via the communication unit 120.

[0065] The acceleration sensor 155 is a sensor that detects acceleration applied to the terminal device 10. Preferably, the acceleration sensor 155 has a function of detecting tilt around each axis (X-axis, Y-axis, Z-axis) of a three-dimensional coordinate system with the position of the terminal device 10 as the origin. The acceleration sensor 155 having such a function can detect the attitude of the terminal device 10, that is, the direction with respect to the X-axis, Y-axis, and Z-axis, by detecting the gravitational acceleration of the gravitational force with respect to the earth.

[0066] 1, and stores data and programs used by the terminal device 10. The programs include application programs such as a web browser application.

[0067] The control unit 190 is realized, for example, by the processor 19 reading a program stored in the storage unit 180 and executing instructions included in the program. The control unit 190 controls the operation of the terminal device 10. The control unit 190 performs the functions of an operation reception unit 191, a transmission / reception unit 192, and a presentation control unit 193 by operating in accordance with the read program.

[0068] Operation acceptance unit 191 performs processing for accepting instructions or information input from input device 13. Specifically, operation acceptance unit 191 accepts instructions or information input from touch-sensitive device 131. In this embodiment, it is assumed that a user inputs text data of geographic information from touch-sensitive device 131, and operation acceptance unit 191 accepts the input.

[0069] The transmitting / receiving unit 192 performs processing for the terminal device 10 to transmit and receive data to and from an external device in accordance with a communication protocol. Specifically, the transmitting / receiving unit 192 transmits instructions or information input by a user to the server 20. The transmitting / receiving unit 192 receives information transmitted from the server 20. The presentation control unit 193 controls the output device 14 to present various types of information to the user.

[0070] <2-3 Server configuration> Fig. 3 is a block diagram showing an example of the configuration of the server 20 shown in Fig. 1. As shown in Fig. 3, the server 20 performs the functions of a communication unit 201, a storage unit 202, and a control unit 203.

[0071] The communication unit 201 performs processing for the server 20 to communicate with an external device (for example, the terminal device 10). The storage unit 202 is realized by the memory 25 and the storage 26, and stores data and programs used by the server 20. The programs include application programs such as a web browser application.

[0072] The storage unit 202 stores, for example, a related information database 2021 and an application 2022. The related information database 2021 is, for example, a database that stores related information. Note that the related information database 2021 may be stored in a storage area separate from the storage unit 202, such as an external system.

[0073] The application 2022 is an application for managing the use of the validity assessment service by the user. The application 2022 runs, for example, in the background of other applications installed on the server 20, and monitors the processes executed by the user. The user can access the application 2022 by using, for example, a web browser application installed on the terminal device 10.

[0074] The server 20 may, for example, perform a predetermined analysis process by grasping and managing the usage status of the application 2022. Furthermore, for example, the application 2022 may be installed in the terminal device 10 and stored in the storage unit 180.

[0075] The control unit 203 is realized by the processor 29 reading a program stored in the storage unit 202 and executing instructions included in the program. The control unit 203 controls the operation of the server 20. The control unit 203 operates in accordance with the read program to fulfill the functions of a reception control module 2031, a transmission control module 2032, a presentation control module 2033, and a validity evaluation module 2034.

[0076] The reception control module 2031 controls the process in which the server 20 receives signals from external devices in accordance with a communication protocol. The transmission control module 2032 controls the process in which the server 20 transmits signals to external devices in accordance with a communication protocol. The presentation control module 2033 controls the process of presenting various information to the user. The validity evaluation module 2034 executes a series of processes related to the generation and output of validity information based on, for example, geographic information acquired from the terminal device 10.

[0077] <2-4 Data Structure> 4 is a diagram showing the data structure of tables stored in the server 20. Note that FIG. 4 is merely an example and does not exclude data that is not listed. Furthermore, even data listed in the same table may be stored in separate storage areas in the storage unit 202.

[0078] 4 is a diagram showing the data structure of the related information database 2021. The various related information stored in this database is referenced, for example, when the AI ​​system 40 generates validity information as an AI agent. The related information database 2021 has columns for information type and content, with an information ID as a key, for example.

[0079] The item "information ID" is, for example, a column that stores an identifier for uniquely identifying various related information. The item "information type" is, for example, a column that stores information indicating a category of related information. In the example of Figure 4, the types of "regulatory information," "certification standards," "past cases," "similar area information," "methodology," and "market data" are stored in the item "information type." The item "content" is, for example, a column that stores specific content of various related information. In the example of Figure 4, the specific content of regulatory information, certification standards, past cases, similar area information, methodology, and market data is stored in the item "content." As an example, the item "content" corresponding to the type of "market data" stores, for example, various information regarding the market price and trading trends of carbon credits.

[0080] <2-5 Operation> An example of the operation of the server 20 according to this embodiment will be described below.

[0081] In step S11 shown in FIG. 5, the server 20 accepts input of geographic information by a user. Specifically, for example, the user inputs the geographic information using the input device 13 (touch-sensitive device 131) of the terminal device 10. The input device 13 may be, for example, a chat interface. The user may also input the geographic information by voice using the microphone 171, or may upload image data (e.g., map data, etc.) as the geographic information. In the case of voice input or image data upload, the generative AI model provided in the AI ​​system 40 may be, for example, a multimodal generative AI model. The operation acceptance unit 191, for example, accepts input of geographic information from a user. The transmission / reception unit 192, for example, transmits the input geographic information to the server 20. The reception control module 2031, for example, receives the geographic information transmitted from the terminal device 10.

[0082] In step S12, the server 20 causes the LLM to acquire related information from the related information database 2021 based on the geographic information. Specifically, for example, the validity evaluation module 2034 transfers the received geographic information to the AI ​​system 40. The LLM of the AI ​​system 40 generates a search query from the received geographic information and accesses the related information database 2021 itself to perform a vector search. The LLM reads and acquires, for example, at least one of regulatory information, certification standards, past cases, similar area data, etc. related to the geographic information from the related information database 2021.

[0083] This LLM search process is an example of RAG, a technique in which a generative AI model searches for information from an external knowledge base to base its answer on before generating it.

[0084] In step S13, the server 20 causes the LLM to analyze the geographic information and related information, thereby outputting validity information. Specifically, for example, the LLM analyzes the geographic information received from the server 20 in step S12 together with the related information it also acquired in step S12 to evaluate the validity of planning a carbon credit project in the candidate site. Based on the evaluation results, the LLM generates validity information including, for example, an evaluation result of the candidate site's eligibility (eligibility information), a proposal for an applicable methodology (methodology information), and a revenue forecast based on multiple scenarios (forecast information), and outputs the validity information. Here, for example, if it is determined that it is desirable to include the eligibility information and methodology information in the content of the validity information, the LLM acquires and analyzes a predetermined number of satellite images from the satellite 30 via a ground station. For example, the AI ​​system 40 acquires the validity information output from the LLM and transmits it to the server 20.

[0085] In step S14, the server 20 presents the validity information to the user. Specifically, for example, the presentation control module 2033 transmits display information for displaying the validity information received from the AI ​​system 40 to the terminal device 10. The presentation control unit 193, for example, accepts a display request from the user and displays the validity information on the display 141 based on the display information received from the server 20. Note that if the output device 14 is a printer, the presentation control module 2033 may, for example, control the presentation control unit 193 to cause the output device 14 to print out a paper medium on which the validity information is printed.

[0086] <2-6 Summary> As described above, in this embodiment, when the user inputs geographic information of a candidate location, the server 20 transmits it to the AI ​​system 40. The AI ​​system 40 functions as an AI agent and autonomously acquires related information from the related information database 2021 based on the geographic information (RAG). The AI ​​system 40 (specifically, the LLM) generates and outputs validity information based on the geographic information and related information. The presentation control module 2033 controls the presentation control unit 193 to display the validity information on the display 141.

[0087] As described above, according to this embodiment, the AI ​​system 40 autonomously executes complex research and analysis processes similar to those performed by experts as an AI agent, while the RAG function ensures the reliability of the validity information output from the LLM. In other words, the AI ​​system 40 as an AI agent automates a series of tasks that would normally be performed manually by humans, such as checking regulatory information, searching for past cases, and comparing methodologies. Therefore, the user only needs to input geographic information, simplifying the process leading up to the output of validity information. Furthermore, since the LLM does not generate answers based on its own knowledge but instead relies on information from a highly reliable database via the RAG function, the accuracy of the output validity information is guaranteed.

[0088] Therefore, according to this embodiment, even a user with little specialized knowledge about carbon credits can easily and accurately grasp the validity of a project plan at an early stage. This makes it possible to evaluate the validity of a plan before the creation of a business plan, which was difficult with conventional techniques, and can provide strong support for user decision-making.

[0089] <2-7 Screen example> An example screen according to one embodiment of the present disclosure will be described below with reference to Fig. 6. Fig. 6 is a schematic diagram showing an example screen of display 141 presented to the user.

[0090] 6, an evaluation report 1411 is displayed on the display 141. This evaluation report 1411 is an example of validity information, and is displayed on the display 141 under the control of the presentation control module 2033 via the presentation control unit 193. The evaluation report 1411 includes, for example, four areas: an area 1412 for "qualification information," an area 1413 for "methodology information," an area 1414 for "prediction information," and an area 1415 for "validity evaluation."

[0091] Area 1412 is an area that displays, for example, eligibility information. In the example of FIG. 6, the area of ​​the candidate site, the total area of ​​the condition-satisfying areas, and the final eligibility evaluation results are displayed in area 1412. The condition-satisfying areas are, for example, areas of the candidate site that satisfy the eligibility conditions. The information displayed in area 1412 corresponds, for example, to the results of analysis performed by AI system 40 based on geographic information and related information.

[0092] Area 1413 is an area that displays, for example, methodology information. In the example of FIG. 6, applicable methodologies are listed for each condition-satisfying area (areas A, B, and C in the figure). Specific suggestions are also shown, including the tree species to be planted and the conservation methods to be adopted. The information displayed in area 1413 corresponds to, for example, the results of the AI ​​system 40 proposing an optimal project implementation plan based on the analysis results and various information in the related information database 2021.

[0093] Area 1414 is an area that displays, for example, forecast information. In the example of Fig. 6, the forecast values ​​for carbon accumulation and methane gas reduction, as well as revenue forecasts for the entire condition-satisfying area, are displayed in area 1414 for each of multiple scenarios. The information displayed in area 1414 corresponds to, for example, the results of a simulation of future results, etc., performed by the AI ​​system 40 using market data, etc., stored in the related information database 2021.

[0094] Area 1415 is an area that displays a final assessment of validity, for example, by integrating the various pieces of information displayed in areas 1412 to 1414. In the example of Fig. 6, area 1415 displays a clear assessment to support the user's decision-making, such as "The planning of project implementation at the candidate site is valid." This final assessment of validity corresponds to, for example, the result of AI system 40 integrating all analysis results and outputting a final judgment.

[0095] 6 is merely an example, and various variations are expected in the display content and display manner of the evaluation report 1411. For example, information may be presented in various formats according to the convenience of the user, such as a graph display of each predicted value, visualization of the area on a map, or addition of detailed explanatory text.

[0096] 3. Second Embodiment In the first embodiment, an example was described in which the user is not involved in the process from inputting geographic information to generating and outputting validity information by the LLM. However, this is not limiting. For example, the AI ​​system 40 as an AI agent may be configured to perform a deeper analysis that more accurately reflects the user's intentions through an interactive interface with the user. This interactive interface may be, for example, a chat format in which the user interacts with the user via text or voice, or a format in which an expert-like avatar responds using voice and gestures. This allows the user to ask questions and perform related operations more naturally and intuitively.

[0097] First, the LLM of the AI ​​system 40 may be tuned to answer questions from users about, for example, a carbon credit project. Such a function of the AI ​​system 40 is realized, for example, as an interactive Q&A function that answers questions posed by the user at any time.

[0098] For example, if a user asks a general question such as "What types of carbon credits are available in Japan?" or "What kind of system is REDD+?" before entering geographic information, LLM will generate an answer by referencing the relevant information database 2021 and output it as text or audio data. This allows users to obtain the basic knowledge necessary for initial project consideration.

[0099] For example, if a user asks a more specific and detailed question while entering geographic information, such as "Which authentication standard applies to this address?", or after receiving validity information (in report format), such as "Please explain in detail the additionality requirements of the ARR methodology recommended in this report," the LLM will generate a detailed answer based on the context of the geographic information and validity information, as well as related information. This allows users to dig deeper into the validity assessment and actively deepen their understanding.

[0100] To realize such an interactive Q&A function, the LLM is tuned using, for example, a dataset consisting of pairs of various anticipated questions and their corresponding answers. This dataset may be, for example, dialogue scenario data consisting of exchanges between anticipated questions and answers. Furthermore, this dataset does not have to be composed of question-answer pairs; for example, it may simply contain information related to a predetermined theme in a structured manner. Furthermore, the information constituting this dataset may be any information obtained from a reliable site, etc. Possible tuning methods include, for example, fine tuning and RLHF (Reinforcement Learning from Human Feedback).

[0101] When the AI ​​system 40 performs an interactive Q&A function, the operation of the server 20 is, for example, as follows.

[0102] That is, the user inputs a question about the carbon credit project, for example, via the input device 13 or microphone 171 of the terminal device 10. For example, the operation reception unit 191 receives the question, and the transmission / reception unit 192 transmits the question to the server 20.

[0103] The validity evaluation module 2034, for example, transfers a question received from the terminal device 10 to the AI ​​system 40. The LLM interprets the intent of the received question and generates and outputs an answer to the question while referring to the related information database 2021 as necessary. The AI ​​system 40, for example, acquires the answer output from the LLM and returns it to the server 20.

[0104] The presentation control module 2033, for example, transmits presentation information for presenting the answer received from the AI ​​system 40 to the terminal device 10. The presentation control unit 193, for example, causes the answer to be displayed as text on the display 141 or output as audio from the speaker 172 based on the presentation information.

[0105] Each time the server 20 and the AI ​​system 40 receives a question from the user in response to the presented answer, or each time they receive another question from the user, the server 20 and the AI ​​system 40 repeat, for example, the execution of this series of processes.

[0106] Next, the LLM may be tuned to autonomously ask questions and / or make suggestions depending on the content of the input geographic information. Such a function of the AI ​​system 40 is realized, for example, as part of a dialogue management technique in a dialogue system.

[0107] For example, if the input geographic information alone is not enough to narrow down the applicable methodology to one, the LLM will autonomously generate a clarification dialogue such as, "Both REDD+ and ARR are possible in this region. Which scenario would you prioritize in your analysis?" Also, if the specified area included in the input geographic information is too narrow for the survey area, the LLM will autonomously create a suggestion such as, "Should we expand the survey area to a radius of XX km?" This allows the system to elicit information from the user that is necessary for inference, enabling more accurate analysis.

[0108] To realize such autonomous questioning and suggestion functions, the LLM is tuned using datasets consisting of pairs of "situations" and "next actions." Examples of such datasets include a clarification request dataset, which returns appropriate questions when the user's input is ambiguous or lacks information, and a proposal scenario dataset, which suggests useful options to the user when the analysis process reveals them. A clarification request dataset consists of a pair of a "situation" such as "The user entered polygon data over a large area and simply instructed the system to "analyze this location." A pair of a "next action" such as "The specified area contains both agricultural land and forests. Which land use would you like the project to target?" A proposal scenario dataset consists of a pair of a "situation" such as "The analysis results show that the candidate site meets the ARR eligibility criteria to a high degree." A pair of a "next action" such as "This candidate site has been determined to be highly suitable for an ARR project. Would you like to view a simulation of carbon absorption forecasts from similar past cases for reference?" Possible tuning methods include fine tuning and prompt tuning.

[0109] When the AI ​​system 40 performs the autonomous question / suggestion function, the operation of the server 20 is, for example, as follows: This operation is performed, for example, at an intermediate stage after the server 20 receives the geographic information and before the LLM outputs the validity information.

[0110] That is, the LLM, for example, performs an initial analysis of the geographic information transferred from the server 20. If, during the initial analysis, it determines that the content of the geographic information is ambiguous or that the direction of the analysis needs to be narrowed, the LLM, for example, autonomously generates and outputs at least one of a question and a suggestion to the user. The AI ​​system 40, for example, acquires at least one of the question and the suggestion output from the LLM and returns it to the server 20.

[0111] The presentation control module 2033 transmits, for example, presentation information for presenting at least one of a question and a proposal received from the AI ​​system 40 to the terminal device 10. The presentation control unit 193, for example, displays at least one of a question and a proposal as text on the display 141 or outputs it as audio from the speaker 172 based on the presentation information.

[0112] When the LLM receives a user question or answer to at least one of the presented question and suggestion, the LLM may generate and output at least one of the question, answer, or suggestion, for example, depending on the content of the received user question or answer.

[0113] 4. Third Embodiment In the first embodiment, an example has been described in which a series of processes is completed by presenting the output validity information to the user, but this is not limiting. The server 20 may be configured to receive feedback (e.g., evaluation, improvement suggestions, correction instructions, etc.) on the presented validity information and utilize the feedback to continuously improve the performance of the entire system 1. The feedback is, for example, general information provided for the purpose of evaluating, correcting, and improving the quality of the validity information output by the LLM. The feedback may be provided by a person (e.g., a user or a person in charge) or by an AI model, for example.

[0114] In one application, the server 20 may use the received feedback to tune the LLM, for example, by applying a technique known as RLHF or RLAIF (Reinforcement Learning from AI Feedback).

[0115] Specifically, for example, in the case of human feedback, when a user evaluates the validity information as "useful" or "inaccurate" or suggests a correction, the validity evaluation module 2034 accepts the feedback. Also, for example, in the case of evaluation by an AI model, the AI ​​model analyzes the validity information and outputs feedback, and the validity evaluation module 2034 accepts the feedback output from the AI ​​model.

[0116] Here, the AI ​​model that outputs the feedback is, for example, a separate evaluation model different from the LLM of the AI ​​system 40. The evaluation model may be, for example, a generative AI model or a machine learning model such as a neural network. Also, for example, the evaluation model may be provided in the AI ​​system 40, the server 20, or an external system.

[0117] The validity evaluation module 2034, for example, transfers the received feedback as a reward signal to the AI ​​system 40. The AI ​​system 40, for example, provides the reward signal transferred from the server 20 to the LLM. By updating (fine-tuning) the model parameters based on this reward signal, the LLM improves itself so as to generate highly accurate information that is more in line with the user's intentions.

[0118] Alternatively, the server 20 may have the LLM modify the validity information based on the received feedback, e.g., a form of in-context learning without tuning the LLM.

[0119] Specifically, for example, a user who has received the presented validity information inputs feedback on the content of the information from the input device 13 or microphone 171 of the terminal device 10. For example, a specific correction instruction such as "Please recalculate this revenue forecast based on a more conservative market price" is input. For example, the operation reception unit 191 receives this feedback, and the transmission / reception unit 192 transmits the feedback to the server 20. For example, the validity evaluation module 2034 transfers the feedback received from the terminal device 10 to the AI ​​system 40 together with the context of the current validity information.

[0120] For example, the evaluation model automatically reviews the validity information received from the AI ​​system 40 from the perspectives of whether there are logical inconsistencies and whether it is consistent with internal data. If the evaluation model detects a defect, such as an insufficient explanation of additionality, it will provide feedback to the LLM with instructions on how to address the defect.

[0121] The LLM, for example, interprets the content of the received feedback and modifies the validity information according to that interpretation. The LLM, for example, outputs the modified validity information. The AI ​​system 40, for example, transmits the modified validity information output from the LLM to the server 20.

[0122] The presentation control module 2033, for example, transmits display information for displaying the corrected validity information received from the AI ​​system 40 to the terminal device 10. The presentation control unit 193, for example, causes the display 141 to display the corrected validity information based on the display information.

[0123] 5. Fourth Embodiment In the first embodiment, an example has been described in which a series of processes is completed by presenting the output validity information to the user, but this is not limiting. The server 20 may be configured to create a project plan according to the content of the validity information, for example. The project plan is, for example, a plan for implementing a carbon credit project in a candidate site.

[0124] Specifically, for example, two possibilities are possible: a user's instruction to create a plan is used as a trigger for creating the plan, or the plan is automatically created by the LLM of the AI ​​system 40. When a user's instruction is used as a trigger, for example, a user who has received the validity information decides to execute a carbon credit project plan and issues an instruction to create a plan from the terminal device 10. This instruction to create a plan is sent from the terminal device 10 to the server 20. For example, the validity evaluation module 2034 may receive the instruction to create a project plan from the terminal device 10 and create the project plan itself based on the validity information, or it may transfer the instruction to the AI ​​system 40 and have the LLM create the project plan.

[0125] Whether the validity assessment module 2034 or the LLM creates the project plan, it reads a template, such as a PDD (Project Design Document) pre-stored in the storage unit 202, and creates a structured project plan by mapping validity information to the template. When the LLM creates the project plan, the LLM can, for example, create a project plan that is more natural and persuasive based on the validity information. Furthermore, for example, the LLM can not only simply map the validity information, but also autonomously add explanatory text, text emphasizing the significance of the carbon credit project, etc. to the plan.

[0126] Once the project plan is created by either method, the validity evaluation module 2034 transmits the project plan to, for example, the terminal device 10. Finally, the presentation control unit 193 displays the project plan on the display 141 so that the user can view or download it. This makes it possible to automate most of the process of creating project plans such as PDDs, which require specialized knowledge, and dramatically reduce the burden on the user.

[0127] The validity assessment module 2034 or the LLM may, for example, perform at least one of recommending and matching project developers who have a proven track record in the country of the applicable methodology or project plan. Information about project developers may be stored, for example, in the related information database 2021 or in an external database. The validity assessment module 2034 or the LLM may, for example, refer to the database based on the output validity information to select a relevant project developer. This can accelerate decision-making on planning and implementing carbon credit projects.

[0128] 6. Fifth Embodiment In this embodiment, the information acquisition process is realized by internalizing knowledge through fine-tuning of the LLM. Unlike the first embodiment, which searches for external information at runtime, this embodiment is characterized by embedding the specialized knowledge required for evaluation into the internal parameters of the LLM in advance.

[0129] The knowledge base in this embodiment is used as a large-scale data set for training the model. Specifically, for example, tens of thousands to millions of pairs of "geographic information" and corresponding "correct answer validity information" created by experts are prepared. Then, this data set is used to fine-tune the LLM. As a result, the LLM learns the complex correlation between the geographic information and the validity information within its parameters and "internalizes" the knowledge.

[0130] When a user inputs geographic information, the LLM responds to the input by recalling relevant knowledge from its internal neural network without accessing an external database. This recall process corresponds to the "information acquisition step," and it directly generates validity information based on the internalized knowledge.

[0131] 7 Sixth Embodiment In this embodiment, the information acquisition process is realized through the cooperation of multiple AI agents with different specialties. In this approach, rather than a single AI agent taking on all tasks, the agents work as a team with assigned roles.

[0132] In this embodiment, the system 1 includes a "supervisor agent (not shown)" that manages the overall tasks and integrates the analysis results, and multiple "specialist agents (not shown)" that perform specific specialized tasks. When a user inputs geographic information, the supervising agent first analyzes the content and breaks it down into subtasks required to generate validity information (e.g., researching applicable regulations, geospatial analysis of the target area, analysis of related market trends, etc.). The supervising agent then assigns each of the broken down tasks to the most appropriate specialist agent.

[0133] Specialized agents include, for example, regulatory agents, geospatial agents, and market agents. Regulatory agents perform keyword searches on structured databases (such as SQL) that store legal regulations to identify applicable laws or ordinances. Geospatial agents access APIs published by external satellite data providers to obtain and analyze historical land cover data and vegetation indices for the target area. Market agents use web scraping technology to collect the latest trading prices and market reports for related carbon credits.

[0134] The entire series of autonomous information gathering activities carried out in parallel by these specialized agents corresponds to the "information acquisition step." Finally, the supervisory agent receives reports from each specialized agent, integrates and summarizes them, and generates comprehensive analytical information.

[0135] 8 Seventh Embodiment In this embodiment, the information acquisition process is realized by searching a knowledge graph, which structures specialized knowledge using entities and relationships. This method is specialized in accurately tracing the logical connections of information.

[0136] The knowledge base in this embodiment is constructed as a huge knowledge network in which pieces of information are interrelated. For example, an entity called "Methodology VM0047" stores knowledge in a form connected to an entity called "ARR (Afforestation)" via the relationship "is applicable to." Furthermore, the entity called "ARR" stores knowledge in a form connected to an entity called "Land history of being non-forested land for 10 years or more" via the relationship "has as a requirement."

[0137] When a user inputs geographic information, the server 20 identifies elements contained in the information, such as place names and land types, as entities. Next, using the entities as a starting point, the server 20 executes a structured query (e.g., SPARQL, Cypher) on the knowledge graph.

[0138] This query allows the server 20 to logically trace the defined relationships between entities and accurately extract facts and relationships relevant to the question. This entire logical search process on the graph corresponds to the "information acquisition step." By providing the extracted, highly accurate facts as context to the LLM, it is possible to minimize factual errors (hallucination) and generate highly reliable analytical information.

[0139] [9 Basic Computer Hardware Configuration] 7 is a block diagram showing the basic hardware configuration of a computer 90. The computer 90 includes at least a processor 901, a main memory device 902, an auxiliary memory device 903, and a communication IF 991 (interface), which are electrically connected to one another by a communication bus 921.

[0140] The processor 901 is hardware for executing an instruction set written in a program, and is composed of an arithmetic unit, registers, peripheral circuits, and the like.

[0141] The main memory device 902 is used to temporarily store programs, data to be processed by the programs, etc. For example, it is a volatile memory such as a DRAM (Dynamic Random Access Memory).

[0142] The auxiliary storage device 903 is a storage device for saving data and programs, such as a flash memory, a hard disk drive (HDD), a magneto-optical disk, a CD-ROM, a DVD-ROM, or a semiconductor memory.

[0143] The communication IF 991 is an interface for inputting and outputting signals for communicating with other computers via a network using wired or wireless communication standards.

[0144] The network is composed of the Internet, a LAN, various mobile communication systems constructed by wireless base stations, etc. For example, the network includes 3G, 4G, and 5G mobile communication systems, LTE (Long Term Evolution), and wireless networks (e.g., Wi-Fi (registered trademark)) that can connect to the Internet via a predetermined access point. In the case of a wireless connection, communication protocols include, for example, Z-Wave (registered trademark), ZigBee (registered trademark), and Bluetooth (registered trademark). In the case of a wired connection, the network also includes a direct connection using a USB (Universal Serial Bus) cable, etc.

[0145] It should be noted that the computer 90 can be virtually realized by distributing all or part of each hardware configuration across multiple computers 90 and interconnecting them via a network. In this way, the computer 90 is a concept that includes not only a computer 90 housed in a single housing or case, but also a virtualized computer system.

[0146] [10 Basic Functional Configuration of a Computer] The following describes the functional configuration of a computer realized by the basic hardware configuration (FIG. 7) of the computer 90. The computer includes at least the functional units of a control unit, a storage unit, and a communication unit.

[0147] The functional units of the computer 90 can also be realized by distributing all or part of the functional units among multiple computers 90 interconnected via a network. The computer 90 is a concept that includes not only a single computer 90 but also a virtualized computer system.

[0148] The control unit is realized by the processor 901 reading out various programs stored in the auxiliary storage device 903, expanding them in the main storage device 902, and executing processing in accordance with the programs. The control unit can realize functional units that perform various types of information processing depending on the type of program. In this way, the computer is realized as an information processing device that performs information processing.

[0149] The storage unit is realized by a main storage device 902 and an auxiliary storage device 903. The storage unit stores data, various programs, and various databases. Furthermore, the processor 901 can allocate a storage area corresponding to the storage unit in the main storage device 902 or the auxiliary storage device 903 in accordance with the programs. Furthermore, the control unit can cause the processor 901 to execute processes for adding, updating, and deleting data stored in the storage unit in accordance with the various programs.

[0150] A database refers to a relational database, which manages data sets called masters and tables in a tabular format structurally defined by rows and columns, by relating them to each other. In a database, a table is called a table, a master, a column in a table is called a column, and a row in a table is called a record. In a relational database, relationships between tables and masters can be set and associated.

[0151] Typically, each table and each master has a column set as a primary key to uniquely identify a record, but setting a primary key to a column is not essential. The control unit can cause the processor 901 to add, delete, or update records in specific tables and masters stored in the storage unit according to various programs.

[0152] Furthermore, by storing data, various programs, and various databases in the storage unit, it can be considered that the information processing device and information processing system according to the present disclosure have been manufactured.

[0153] Note that the databases and masters in this disclosure may include any data structure in which information is structurally defined (such as a list, dictionary, associative array, or object). The data structure also includes data that can be considered as a data structure by combining data with functions, classes, methods, etc. written in any programming language.

[0154] The communication unit is realized by the communication IF 991. The communication unit realizes a function of communicating with other computers 90 via a network. The communication unit can receive information transmitted from other computers 90 and input the information to the control unit. The control unit can cause the processor 901 to execute information processing on the received information in accordance with various programs. In addition, the communication unit can transmit information output from the control unit to other computers 90.

[0155] Furthermore, the above-described configurations, functions, processing units, processing means, etc. may be partially or entirely implemented in hardware, for example, by designing them as integrated circuits. The present invention can also be realized by software program code that implements the functions of the embodiments. In this case, a storage medium on which the program code is recorded is provided to a computer, and a processor included in the computer reads the program code stored in the storage medium. In this case, the program code itself read from the storage medium implements the functions of the above-described embodiments, and the program code itself and the storage medium on which it is stored constitute the present invention. Examples of storage media for providing such program code include flexible disks, CD-ROMs, DVD-ROMs, hard disks, SSDs, optical disks, magneto-optical disks, CD-Rs, magnetic tapes, non-volatile memory cards, and ROMs.

[0156] Furthermore, the program code that realizes the functions described in this embodiment can be implemented in a wide range of program or script languages, such as assembler, C / C++, perl, Shell, PHP, Java (registered trademark), JavaScript, and TypeScript.

[0157] Furthermore, the program code of the software that realizes the functions of the embodiments may be distributed via a network and stored in a storage means such as a computer's hard disk or memory, or in a storage medium such as a CD-RW or CD-R, and the processor of the computer may read and execute the program code stored in the storage means or the storage medium.

[0158] The functions performed by the components described herein may be implemented in circuitry or processing circuitry, including general-purpose processors, application-specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), CPUs (Central Processing Units), conventional circuits, and / or combinations thereof, programmed to perform the described functions. A processor includes transistors and other circuits and is considered to be circuitry or processing circuitry. A processor may also be a programmed processor that executes programs stored in memory.

[0159] In this specification, a circuitry, unit, or means is hardware that is programmed to realize or performs the described functions, which may be any hardware disclosed herein or any hardware known to be programmed to realize or perform the described functions.

[0160] If the hardware is a processor considered to be a type of circuitry, the circuitry, means, or unit is a combination of the hardware and the software used to configure the hardware and / or processor.

[0161] Although several embodiments of the present disclosure have been described above, these embodiments can be embodied in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and modifications are intended to be included in the scope of the inventions and their equivalents as defined in the claims, as well as in the scope and spirit of the inventions.

[0162] [11 Appendix] The matters described in the above embodiments will be supplemented below.

[0163] <Appendix 1> A program to be executed by a computer having a processor and a memory, the program causing the processor to execute the following steps: accepting input of geographic information of a candidate site for a carbon credit project from a user; obtaining information about other carbon credit projects from a knowledge base pre-constructed for evaluating the appropriateness of planning to implement a carbon credit project in the candidate site based on the accepted geographic information; inputting the geographic information and information about other carbon credit projects into a generative AI model and causing the generative AI model to output analytical information regarding the appropriateness of planning to implement a carbon credit project in the candidate site; and presenting the analytical information to the user.

[0164] <Appendix 2> The program described in Appendix 1 further causes the processor to execute the steps of accepting a question from a user regarding a carbon credit project and the generative AI model outputting an answer to the question.

[0165] <Appendix 3> A program described in (Appendix 1) or (Appendix 2), further causing the processor to execute a step in which the generative AI model outputs at least one of a question and a suggestion depending on the content of the input geographic information.

[0166] <Appendix 4> A program described in any of (Appendix 1) to (Appendix 3), further causing the processor to execute the steps of receiving feedback on the analytical information and inputting the received feedback into the generative AI model to tune the generative AI model.

[0167] <Appendix 5> A program described in any one of (Appendix 1) to (Appendix 4), which causes the processor to further execute the steps of receiving feedback on the analytical information, and inputting the received feedback into a generative AI model, causing the generative AI model to modify the analytical information according to the content of the feedback, and outputting the modified analytical information.

[0168] <Appendix 6> A program described in any one of (Appendix 1) to (Appendix 5), further causing the processor to execute a step of creating a plan for implementing a carbon credit project in a candidate location depending on the content of the analysis information.

[0169] <Appendix 7> An information processing device comprising a control unit and a storage unit, wherein the control unit executes all steps in the program according to any one of (Supplementary Note 1) to (Supplementary Note 6).

[0170] <Appendix 8> A method executed by a computer having a processor and a memory, wherein the processor executes all steps in the program described in any one of (Appendix 1) to (Appendix 6).

[0171] <Appendix 9> A system comprising one or more processors that execute all steps in the program described in any one of (Appendix 1) to (Appendix 6). [Explanation of symbols]

[0172] 1. System 10...Terminal device 120…Communications Department 13...Input device 14...Output device 15...Memory 16…Storage 19...Processor 20...Server 22...Communication IF 23...Input / output IF 25…Memory 26…Storage 29...Processor 30…Artificial satellite 40...AI system

Claims

1. A program to be executed by a computer having a processor and a memory, The program causes the processor to: receiving input of geographic information of a candidate site for a carbon credit project from a user; A step of obtaining information on other carbon credit projects from a knowledge base that has been previously constructed for evaluating the appropriateness of planning to implement the carbon credit project in the candidate location based on the received geographic information; inputting the geographical information and information about the other carbon credit projects into a generative AI model, and outputting analytical information from the generative AI model regarding the appropriateness of planning to implement the carbon credit project in the candidate location; presenting the analytical information to a user; A program that executes the following.

2. accepting inquiries from the user regarding the carbon credit project; the generative AI model outputting an answer to the question; The program according to claim 1 , further causing the processor to execute:

3. The program according to claim 1 , further causing the processor to execute a step in which the generative AI model outputs at least one of a question and a suggestion depending on the content of the input geographic information.

4. receiving feedback on the analysis information; inputting the received feedback into the generative AI model to tune the generative AI model; The program according to claim 1 , further causing the processor to execute:

5. receiving feedback on the analysis information; inputting the received feedback into the generative AI model, causing the generative AI model to modify the analytical information according to the content of the feedback and output the modified analytical information; The program according to claim 1 , further causing the processor to execute:

6. The program according to claim 1 , further causing the processor to execute a step of creating a plan for implementing the carbon credit project in the candidate site according to the content of the analysis information.

7. 7. An information processing apparatus comprising a control unit and a storage unit, wherein the control unit executes all steps of the program according to claim 1.

8. A method executed by a computer having a processor and a memory, wherein the processor executes all the steps of the program of any one of claims 1 to 6.

9. A system comprising one or more processors that execute all steps in the program according to any one of claims 1 to 6.

Images