Green service identification method, device, equipment, medium and program product

Abstract

This application provides a method, apparatus, device, medium, and program product for identifying green businesses, relating to the fields of artificial intelligence and fintech, specifically involving the application of large-scale models in green finance scenarios. The method includes: acquiring multimodal data of the business to be identified and preprocessing the multimodal data to obtain core data; based on retrieval-enhanced generation technology, semantically matching the core data with a dynamic knowledge base using a domain-wide large-scale model to obtain initial category results; wherein, the dynamic knowledge base is dynamically updated based on a set of green industry standard specifications; the domain-wide large-scale model is generated by fine-tuning and training the large-scale model based on historical green business data; and, if the initial category result is a potential green category, performing hierarchical verification of the core data based on an expert rule base, the dynamic knowledge base, and the domain-wide large-scale model to generate identification results; wherein, the expert rule base is constructed based on constraint risk control rules and green identification rules.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and fintech, specifically to the application of large models in green finance scenarios, and more specifically to a green business identification method, device, equipment, medium, and program product. Background Technology

[0002] With the rapid development of green finance, accurate identification of green businesses can guide financial resources from high-pollution and high-energy-consuming sectors to clean energy, energy conservation and environmental protection, and green manufacturing, optimizing resource allocation and contributing to the achievement of dual-carbon goals. This is a key lever for promoting the green and low-carbon transformation of the economy and society. Furthermore, green business identification enables precise determination of the green attributes and carbon footprint of projects, enterprises, and assets, solving information asymmetry and "greenwashing" problems, and supporting the implementation of green finance standards and intelligent risk control.

[0003] In existing green identification technologies, traditional deep learning models and machine learning models are typically used for identification. However, traditional models suffer from high maintenance costs and delayed updates. They require retraining when green standards change, resulting in long iteration cycles. Furthermore, the black-box nature of traditional models leads to poor interpretability, making it difficult to meet financial regulatory compliance requirements. In addition, they have weak processing capabilities for unstructured text data, and the process is highly dependent on manual comparison, resulting in low efficiency and difficulty in ensuring consistency in standard implementation. Summary of the Invention

[0004] In view of the above problems, embodiments of this application provide a green business identification method, apparatus, device, medium, and program product.

[0005] According to a first aspect of this application, a method for identifying green businesses is provided, comprising: acquiring multimodal data of the business to be identified, and preprocessing the multimodal data to obtain core data; based on retrieval enhancement generation technology, semantically matching the core data with a dynamic knowledge base through a domain large model to obtain an initial category result; wherein the dynamic knowledge base is dynamically updated based on a set of green industry standard specifications; the domain large model is generated by fine-tuning and training the large model based on historical green business data; and, if the initial category result is a potential green class, performing hierarchical verification on the core data based on an expert rule base, the dynamic knowledge base, and the domain large model to generate an identification result; wherein the expert rule base is constructed based on constraint risk control rules and green identification rules.

[0006] According to an embodiment of this application, the step of performing hierarchical verification on the core data based on the expert rule base, the dynamic knowledge base, and the domain big model to generate identification results includes: performing scenario compliance verification on the core data based on the business identification scenario of the business to be identified, using the domain big model to generate green classification results and compliance explanation information; and performing cross-validation on the green classification results based on the expert rule base, using the domain big model to generate the identification results and verification explanation information.

[0007] According to embodiments of this application, the business identification scenario includes at least one of a subject identification scenario, a project identification scenario, and a credential identification scenario. The business identification scenario based on the business to be identified, through the domain-wide model, performs scenario compliance verification on the core data, including: for the subject identification scenario, verifying the energy efficiency of the main products of the core data based on the domain-wide model and the dynamic knowledge base; for the project identification scenario, extracting project data from the core data and verifying the project data against environmental standards based on the domain-wide model and the dynamic knowledge base; and / or for the credential identification scenario, obtaining the subject classification result and project classification result corresponding to the credential to be identified, and verifying the transaction direction of the core data through the domain-wide model based on the subject classification result and the project classification result.

[0008] According to an embodiment of this application, the step of verifying the energy efficiency of the main products based on the domain-wide model and the dynamic knowledge base includes: extracting the main business data from the core data, analyzing the green industry classification and confidence level corresponding to the main business data, and extracting main business information based on the main business data; obtaining intellectual property information corresponding to the main business information when the confidence level meets a first-level threshold, and determining the consistency between the intellectual property information, the main business information, and the green industry classification; and extracting product energy efficiency standard data based on the main business information and the dynamic knowledge base when the confidence level meets a second-level threshold, and comparing the energy efficiency testing data corresponding to the main business information with the product energy efficiency standard data item by item through the domain-wide model.

[0009] According to an embodiment of this application, the project data includes report data, address data, and environmental assessment data. The step of verifying the project data against environmental standards based on the domain-wide model and the dynamic knowledge base includes: parsing the construction indicator information corresponding to the report data, and comparing the construction indicator information with the green industry information in the dynamic knowledge base using the domain-wide model; performing spatial overlay analysis on the address data and the ecological protection red line database using the domain-wide model; and extracting environmental information from the environmental assessment data, and comparing the environmental information with the environmental standard information in the dynamic knowledge base using the domain-wide model.

[0010] According to an embodiment of this application, the step of verifying the transaction direction of the core data using the domain-wide model based on the subject classification result and the project classification result includes: determining the greenness of the project classification result; if the project classification result is not a green project, extracting the transaction voucher information of the core data and verifying the industry direction of the transaction voucher information using the domain-wide model; if the industry direction verification fails, determining the greenness of the subject classification result; and if the subject classification result is not a green subject, extracting the transaction business information of the core data, verifying the business direction of the transaction business information using the domain-wide model, and verifying the consistency between the fund payment object and the transaction object of the transaction voucher information.

[0011] According to an embodiment of this application, the step of semantically matching the core data with a dynamic knowledge base using a domain-wide model based on retrieval enhancement generation technology to obtain an initial category result includes: retrieving green standard rule data associated with the core data from the dynamic knowledge base based on the retrieval enhancement generation technology; and performing semantic fuzzy matching between the core data and the green standard rule data using the domain-wide model, and determining the initial category result based on the semantic fuzzy matching result and the dynamic knowledge base.

[0012] According to an embodiment of this application, the preprocessing of the multimodal data to obtain core data includes: extracting structured data from the multimodal data based on keywords; parsing the unstructured data of the multimodal data using optical character recognition technology to obtain parsed data; extracting key data from the structured data and the parsed data based on the business recognition scenario; and standardizing the key data to obtain the core data.

[0013] According to a second aspect of this application, a green business identification device is provided, comprising: a data processing module for acquiring multimodal data of the business to be identified and preprocessing the multimodal data to obtain core data; a semantic matching module for semantically matching the core data with a dynamic knowledge base based on retrieval enhancement generation technology and a domain large model to obtain an initial category result; wherein the dynamic knowledge base is dynamically updated based on a set of green industry standards and specifications; the domain large model is generated by fine-tuning and training the large model based on historical green business data; and a hierarchical verification module for performing hierarchical verification on the core data based on an expert rule base, the dynamic knowledge base, and the domain large model when the initial category result is a potential green category, to generate an identification result; wherein the expert rule base is constructed based on constraint risk control rules and green identification rules.

[0014] According to a third aspect of this application, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0015] According to a fourth aspect of this application, a computer-readable storage medium is also provided, on which a computer program or instructions are stored, wherein the computer program or instructions, when executed by a processor, implement the steps of the above-described method.

[0016] According to a fifth aspect of this application, a computer program product is also provided, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0017] In the embodiments of this application, multimodal business data is preprocessed automatically to efficiently obtain core data; based on retrieval-enhanced generation technology, a dynamic knowledge base is used to achieve real-time adaptation to green standards, fine-tuning training and using a domain-wide large model, significantly reducing maintenance costs and shortening the iteration cycle; combined with hierarchical verification of the domain-wide large model and expert rule base, the interpretability of recognition is improved, meeting the compliance requirements of financial regulatory authorities; combined with a dual-track judgment mechanism of semantic matching of the large model and precise verification of expert rules, it not only ensures the ability to understand non-standard texts, but also ensures the rigor of financial business, while eliminating the high dependence on manual comparison, improving recognition efficiency, accurately determining the green attributes of business, and optimizing the green allocation of financial resources. Attached Figure Description

[0018] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0019] Figure 1The illustrations depict application scenarios of the green service identification method, apparatus, device, medium, and program products according to embodiments of this application.

[0020] Figure 2 A flowchart illustrating a green service identification method according to an embodiment of this application is shown schematically.

[0021] Figure 3 This illustration shows a flowchart of the scenario compliance verification process for the green business identification method according to an embodiment of this application.

[0022] Figure 4 A schematic diagram illustrates a green customer identification flowchart of a green business identification method according to an embodiment of this application;

[0023] Figure 5 A flowchart illustrating the green project identification process of the green business identification method according to an embodiment of this application is shown.

[0024] Figure 6 A schematic diagram illustrating a green loan document flowchart of a green business identification method according to an embodiment of this application is shown.

[0025] Figure 7 A schematic diagram illustrating the structure of a green service identification device according to an embodiment of this application is shown.

[0026] Figure 8 A block diagram of an electronic device suitable for implementing a green business identification method according to an embodiment of this application is shown schematically. Detailed Implementation

[0027] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0028] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0029] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0030] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0031] In one or more embodiments described herein, the term "large model" can refer to a deep learning model with a large number of model parameters, which can include hundreds of millions, tens of billions, hundreds of billions, trillions, or even tens of trillions of model parameters. Large models can also be called foundational models or basic models. They are pre-trained using large-scale unlabeled corpora to produce pre-trained models with hundreds of millions of parameters. Such models can adapt to a wide range of downstream tasks and have good generalization ability, such as large language models and multimodal pre-trained models. It should be understood that in practical applications, large models only require a small number of samples to fine-tune the pre-trained model before being applied to different tasks. Large models can be widely used in natural language processing, computer vision, and other fields. Specifically, they can be applied to computer vision tasks such as visual question answering, image captioning, and image generation, as well as natural language processing tasks such as text-based sentiment classification, text summarization, and machine translation. Major application scenarios for large models can include digital assistants, intelligent robots, search, online education, office software, e-commerce, and intelligent design.

[0032] Identifying green financial businesses is a key lever for promoting the green and low-carbon transformation of the economy and society. Its importance and necessity are reflected in the following aspects: First, accurately identifying green projects and green assets can guide funds from high-pollution and high-energy-consuming sectors to clean energy, energy conservation and environmental protection, and green manufacturing, optimizing resource allocation and helping to achieve dual-carbon goals. Second, it effectively prevents the risks of greenwashing and ensures that funds are truly used for green and low-carbon activities, maintaining market order and financial stability. Third, it provides a basis for financial institutions to carry out green lending, green bonds, and green investment, improving their risk management capabilities and sustainable development levels. Fourth, it meets regulatory and international rules requirements, promotes the alignment of the financial system with global green governance, and enhances financial competitiveness.

[0033] With the rapid expansion of green finance, financial institutions face increasing demands for the identification, classification, and compliance review of green projects, necessitating efficient, accurate, and scalable green business identification capabilities. However, existing green identification technologies still have significant shortcomings in practical applications, failing to fully support business development and regulatory requirements. These shortcomings are mainly reflected in the following aspects:

[0034] Traditional machine learning models are costly to maintain and slow to update: Traditional classification algorithms require extensive training on specific data. When green finance standards change, the models often need to be retrained, resulting in long iteration cycles and difficulty in adapting to frequent updates to the latest requirements.

[0035] The black box nature of models leads to poor interpretability: Although traditional deep learning or machine learning models can provide classification results, they cannot provide specific judgment criteria and logic, making it difficult for business personnel to understand and verify them, which does not meet the compliance requirements of financial regulators.

[0036] Weak ability to process unstructured data: Green certification relies on a large amount of unstructured text (such as environmental impact assessment reports, feasibility study reports, business scope descriptions, etc.). Traditional methods mainly rely on manual reading or simple keyword matching, which is inefficient and prone to omissions.

[0037] High reliance on manual labor: In the existing process, hundreds of industry standards and thousands of product rules need to be compared manually one by one, which is time-consuming, labor-intensive, and difficult to guarantee the consistency of standard implementation.

[0038] This application provides a method for identifying green businesses, comprising: acquiring multimodal data of the business to be identified and preprocessing the multimodal data to obtain core data; based on retrieval enhancement generation technology, semantically matching the core data with a dynamic knowledge base through a domain large model to obtain an initial category result; wherein, the dynamic knowledge base is dynamically updated based on a set of green industry standards and specifications; the domain large model is generated by fine-tuning and training the large model based on historical green business data; and, if the initial category result is a potential green class, performing hierarchical verification on the core data based on an expert rule base, a dynamic knowledge base, and a domain large model to generate an identification result; wherein, the expert rule base is constructed based on constraint risk control rules and green identification rules. In the embodiments of this application, multimodal business data is preprocessed automatically to efficiently obtain core data; based on retrieval-enhanced generation technology, a dynamic knowledge base is used to achieve real-time adaptation to green standards, fine-tuning training and using a domain-wide large model, significantly reducing maintenance costs and shortening the iteration cycle; combined with hierarchical verification of the domain-wide large model and expert rule base, the interpretability of recognition is improved, meeting the compliance requirements of financial regulatory authorities; combined with a dual-track judgment mechanism of semantic matching of the large model and precise verification of expert rules, it not only ensures the ability to understand non-standard texts, but also ensures the rigor of financial business, while eliminating the high dependence on manual comparison, improving recognition efficiency, accurately determining the green attributes of business, and optimizing the green allocation of financial resources.

[0039] It should be noted that the green business identification method and device of this application can be used in the fields of artificial intelligence and fintech, specifically involving the application of large models in green finance scenarios, and can also be used in any field other than artificial intelligence and fintech. The application fields of the green business identification method and device of this application are not limited.

[0040] In the technical solution of this application, the customer / enterprise user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0041] In scenarios where personal information is used for automated decision-making, the methods, devices, and systems provided in this application all provide users with corresponding operation entry points for users to choose to agree to or reject the automated decision results; if the user chooses to reject, the process enters the expert decision-making process.

[0042] Figure 1The illustration shows an application scenario diagram of the green service identification method, apparatus, device, medium, and program product according to embodiments of this application.

[0043] like Figure 1 As shown, application scenario 100 according to an embodiment of this application may include a first terminal device 101, a second terminal device 102, a database 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the database 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables. For example, a user can use the first terminal device 101 and the second terminal device 102 to interact with the server 105 and / or the database 103 through the network 104 to receive or send information, etc.

[0044] The first terminal device 101 and the second terminal device 102 can be electronic devices such as smartphones, wearable devices, personal computers, intelligent voice interaction devices, smart home appliances, intelligent vehicles, in-vehicle terminals, aircraft, unmanned vending terminals, and extended reality devices. Extended reality devices can include virtual reality devices, augmented reality devices, and mixed reality devices. A client application for the target application can be installed and run on the terminal device. This target application can include, but is not limited to, financial transaction applications, payment applications, shopping applications, web browser applications, search applications, instant messaging tools, email clients, and social media platform software (these are just examples). Furthermore, this application embodiment does not limit the form of the target application, including but not limited to applications, mini-programs, etc., installed on the terminal device, and can also be in web page form.

[0045] Server 105 can be a server providing various services, such as a backend management server supporting websites browsed by users using the first terminal device 101 and the second terminal device 102 (this is just an example). The backend management server can analyze and process received user requests and other data, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services such as cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and big data. The server can be a backend server for the aforementioned target application, used to provide backend services to the clients of the target application.

[0046] Database 103 is a professional storage system for storing and managing data. It can store various types of data related to the target application, such as user account information, business transaction records, and content resource data. It supports structured, semi-structured, or unstructured data storage and has management capabilities such as adding, deleting, modifying, querying, backing up, and restoring data. In this application scenario, database 103 can be connected to server 105 via a communication link. Server 104 can retrieve the required data from database 103 for processing based on requests from the first terminal device 101 and the second terminal device 102. It can also synchronously store new data generated by the operations of the first terminal device 101 and the second terminal device 102 into database 103, thereby achieving data persistence and efficient retrieval.

[0047] It should be noted that the green service identification method provided in this application embodiment can generally be executed by server 105 and / or terminal devices 101-102. Accordingly, the green service identification device provided in this application embodiment can generally be set in server 105 and / or terminal devices 101-102.

[0048] It should be understood that Figure 1 The number of terminal devices, networks, databases, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, databases, and servers can be included.

[0049] Figure 2 A flowchart illustrating a green service identification method according to an embodiment of this application is shown. Figure 2 As shown, the green service identification method 200 according to the embodiments of this application may include steps S210 to S230.

[0050] In step S210, multimodal data of the service to be identified is acquired and preprocessed to obtain core data.

[0051] Multimodal data acquisition channels cover system interface calls, user uploads, and integration with third-party public databases to comprehensively obtain basic information and related materials of the business to be identified. Basic information must include at least one of the following: customer name, unified social credit code, project name, and contract number. Related materials must cover key documents such as business registration information, corporate financial statements, project application forms, feasibility study reports, environmental impact assessment reports, trade contracts, and invoices.

[0052] In the preprocessing of multimodal data, for structured data (such as business registration information and transaction records), core key fields such as business scope and capital flow are directly extracted through preset field mapping rules. For unstructured data, Optical Character Recognition (OCR) technology is used to parse the text content, supporting character recognition in complex document formats (such as scanned documents) to extract key information such as project construction content, product parameters, and trading partners. Then, data cleaning removes duplicate and invalid data, and standardized format conversion (such as date unification and unit standardization) is used to obtain core data. A data quality verification mechanism is established to ensure that the core data is complete and free of logical contradictions.

[0053] In step S220, based on the retrieval enhancement generation technology, the core data is semantically matched with the dynamic knowledge base through the domain big model to obtain the initial category results; wherein, the dynamic knowledge base is dynamically updated based on the green industry standard specification set; the domain big model is generated by fine-tuning and training the big model based on historical green business data.

[0054] The Retrieval-Augmented Generation (RAG) technology process is initiated. First, a combination of keyword retrieval and semantic indexing is used to accurately extract green technology standards, judgment rules, and industry regulations related to the core data from a dynamic knowledge base. Then, a domain-wide model is invoked to perform deep semantic fuzzy matching between the core data (including the company's business scope, project construction content, and investment information) and the search results. This overcomes the limitations of traditional literal matching and accurately captures the deep semantic relationship between the data and green industry classifications. Based on the matching degree and the classification standards in the Green and Low-Carbon Transformation Industry Guidance Catalogue, the initial green category corresponding to the business is determined. For example, a confidence level of 50% (only the business scope involves green industries), 70% (the main business is highly related to green industries), or 90% (complies with detailed national standards) is assigned. A confidence level higher than 50% is identified as a potential green category for subsequent judgment. If the company's business scope and project construction content in the core data do not involve any green industry classification, the system directly determines it as a non-green business and simultaneously generates a reasoning that the business scope / project construction content does not involve green industries, which is then fed back to the user.

[0055] Initially based on green technology standards, identification and judgment rules, and industry-specific technical regulations, and integrating core documents such as the Green and Low-Carbon Transformation Industry Guidance Catalogue, a set of green industry standards and specifications was formed, constructing a basic knowledge base. Through manual uploading or automatic data retrieval, the latest standard documents issued by regulatory authorities and the practical experience of business personnel were synchronized to enrich the content of the knowledge base. Regular data verification was conducted within the knowledge base, invalid rules were removed, new standards were added, and the rule association logic was optimized based on business identification feedback, forming a dynamic knowledge base that achieved dynamic iteration and precise adaptation.

[0056] A large-scale language model (large model) was selected as the basic architecture. Historical green business classification data, judgment criteria, and compliance standards were collected to construct a high-quality training dataset. After cleaning and labeling the dataset, a fine-tuning training method was adopted, incorporating green finance expertise and semantic association logic to optimize the large model's ability to understand non-standard text. Continuous iterative training was conducted using dynamic knowledge base content. The classification accuracy of the large model was verified through a validation set, and model parameters were adjusted based on feedback from business scenarios, ultimately forming a domain-specific large model for green business identification.

[0057] In step S230, if the initial category result is a potential green category, the core data is hierarchically verified based on the expert rule base, dynamic knowledge base and domain big model to generate the identification result; wherein, the expert rule base is constructed based on the constraint risk control rules and green identification rules.

[0058] When the initial category is a potential green category, a multi-dimensional, hierarchical verification process is initiated. First, scenario-based compliance verification is conducted. For entity (e.g., customer) identification scenarios, the energy efficiency of the main products is verified to meet detailed national standards. For project identification scenarios, the project address is verified to be within the ecological protection red line and to comply with national standards. For voucher (e.g., IOU) identification scenarios, the compliance of fund allocation and related relationships is verified. If the verification fails, the business is directly determined to be non-green. To ensure the uniqueness of the classification, information such as the proportion of the enterprise's main business revenue and the core attributes of the products is combined to narrow down the green category to a unique category and update the confidence level. Next, a domain-wide model is invoked, and cross-validation of the classification results is conducted through an expert rule base to verify compliance with risk control rules and green identification rules. Finally, the verification data from the entire process is integrated. If cross-validation passes, the business is determined to be green, and the chain of evidence for compliance with the catalog classification, compliance verification, and expert rule verification is listed. If it fails, the specific reasons for non-green status are clearly stated, forming standardized identification results to ensure accuracy and compliance.

[0059] Once green businesses are identified, the identification results can be reported to the regulatory statistics system to achieve unified supervision of green businesses and meet regulatory requirements. The green business identification results support carbon accounting throughout the entire investment and financing process, accurately measuring project carbon reduction and environmental benefits, and achieving quantitative assessments of environmental benefits such as energy conservation, carbon reduction, and pollution prevention. Simultaneously, green businesses will be subject to internal ledger management, performance evaluation, and full-process risk control within the green finance sector, establishing a robust risk control defense line for green businesses.

[0060] Based on risk control rules constrained by high-energy-consuming and high-polluting industries, ecological protection red line verification, and the experience of green industry identification experts, a basic expert rule base is constructed by integrating green finance system databases and revised payment statistics. A rule review mechanism is established to regularly add compliance verification rules and revise invalid clauses based on regulatory policy adjustments, business scenario expansion, and feedback from identification practices. The effectiveness of the rules is ensured through manual review or system testing, achieving dynamic updates to the expert rule base.

[0061] In the embodiments of this application, multimodal business data is preprocessed automatically to efficiently obtain core data; based on retrieval-enhanced generation technology, a dynamic knowledge base is used to achieve real-time adaptation to green standards, fine-tuning training and using a domain-wide large model, significantly reducing maintenance costs and shortening the iteration cycle; combined with hierarchical verification of the domain-wide large model and expert rule base, the interpretability of recognition is improved, meeting the compliance requirements of financial regulatory authorities; combined with a dual-track judgment mechanism of semantic matching of the large model and precise verification of expert rules, it not only ensures the ability to understand non-standard texts, but also ensures the rigor of financial business, while eliminating the high dependence on manual comparison, improving recognition efficiency, accurately determining the green attributes of business, and optimizing the green allocation of financial resources.

[0062] According to embodiments of this application, preprocessing multimodal data to obtain core data includes: extracting structured data from multimodal data based on keywords; parsing unstructured data from multimodal data using optical character recognition technology to obtain parsed data; extracting key data from structured data and parsed data based on business recognition scenarios; and standardizing key data to obtain core data.

[0063] For structured data, a set of key fields corresponding to business scenarios is preset (such as the main business and product names in customer identification scenarios, and the construction content and site address in project identification scenarios). Through data interfaces, the system connects to business information databases, transaction flow systems, etc., and accurately extracts target fields from structured data by keywords.

[0064] For unstructured data, it supports parsing various formats such as scanned documents and images. First, image preprocessing (such as noise reduction, skew correction, and enhancement) is performed to improve recognition accuracy. Then, OCR technology is used to recognize the image content, extract the text content, and generate parsed data.

[0065] Based on the differentiated needs of business identification scenarios (such as entities, projects, and vouchers), key data is filtered. For customer scenarios, the focus is on extracting the proportion of main business revenue, product patents, and energy efficiency indicators; for project scenarios, the focus is on construction indicators, site latitude and longitude, and pollutant emission data; for voucher scenarios, the focus is on extracting fund allocation, counterparties, and payment methods. Finally, through standardization processing, the data format is unified (e.g., dates are "YYYY-MM-DD", percentages are rounded to two decimal places), data ambiguities are corrected, and invalid values ​​are removed to obtain standardized and unified core data.

[0066] In the embodiments of this application, for multimodal data, keyword extraction is performed on structured data, optical character recognition is performed on unstructured data, and key data is extracted and standardized in combination with business scenarios. This efficiently integrates various types of data, accurately filters the core information required for business recognition, improves the capabilities of unstructured data, and unifies data standards, laying a high-quality data foundation for subsequent accurate matching and verification.

[0067] According to embodiments of this application, based on retrieval enhancement generation technology, core data is semantically matched with a dynamic knowledge base through a domain-wide model to obtain initial category results. This includes: retrieving green standard rule data associated with the core data from the dynamic knowledge base based on retrieval enhancement generation technology; and performing semantic fuzzy matching between the core data and the green standard rule data through a domain-wide model, and determining the initial category results based on the semantic fuzzy matching results and the dynamic knowledge base.

[0068] The process of enhancing search generation technology is initiated by first performing semantic word segmentation and keyword extraction on core data (such as new energy, low-carbon transformation, and environmental governance related to green industries) to construct a search vector. Based on this vector, a multi-dimensional search is performed in a dynamic knowledge base, prioritizing matching with category entries in the Green and Low-Carbon Transformation Industry Guidance Catalogue, and then associating them with corresponding green technology standards, identification and judgment rules, and industry-specific technical regulations to filter out highly relevant green standard and rule data.

[0069] Subsequently, a domain-wide model is invoked to perform deep semantic fuzzy matching between the core data and the retrieved green standard rule data. This overcomes the limitations of literal matching and captures the implicit connections between the company's business scope, project construction content, and the green industry through contextual semantic understanding. The model scores the matching similarity and determines the initial green category based on confidence level standards (e.g., 50%, 70%, 90%). If the core data has no effective connection with any of the green standard rule data, it is directly classified as a non-green business, and the initial category result is labeled "Non-Potential Green Category," along with relevant judgment criteria. If the core data has an effective connection with all the green standard rule data (e.g., a confidence level of 70%), it is directly classified as a potential green business, and the initial category result is labeled "Potential Green Category," along with relevant judgment criteria.

[0070] In the embodiments of this application, retrieval enhancement generation technology is used to accurately retrieve associated green standard rules based on a dynamic knowledge base. Then, a domain-wide model is combined to achieve semantic fuzzy matching between core data and rules, which improves the flexibility and adaptability of matching and can more accurately match various types of business data. The matching process combines the dynamic knowledge base to determine the initial category results, ensuring that the results conform to the latest green standards and greatly improving the accuracy and efficiency of initial identification.

[0071] According to embodiments of this application, core data is subjected to hierarchical verification based on an expert rule base, a dynamic knowledge base, and a domain-wide model to generate identification results. This includes: performing scenario compliance verification on the core data based on the business identification scenario of the business to be identified, using the domain-wide model to generate green classification results and compliance explanation information; and performing cross-validation on the green classification results based on the expert rule base and the domain-wide model to generate identification results and verification explanation information.

[0072] The system clarifies the scenario type (entity, project, voucher) of the business to be identified and loads the corresponding compliance verification rules into the domain-wide model. In the entity identification scenario, the model extracts the core data—the energy efficiency data of the main products—and compares it item by item with the national-level detailed standards in the dynamic knowledge base to generate energy efficiency compliance conclusions and explanations. In the project identification scenario, the large model analyzes the project address's latitude and longitude against the ecological protection red line database to perform spatial overlay analysis, verifying the fit between the construction content and green industry technical requirements, and outputting compliance results for site selection and construction content. In the voucher identification scenario, the large model checks whether the funding direction matches the green classification and whether the payment recipient is consistent with the contract agreement, forming a related compliance judgment.

[0073] Subsequently, the expert rule base is accessed through a large model, and the green classification results generated by scenario compliance verification are cross-compared with the rules in the expert rule base (such as constraint risk control rules and green identification rules). If all rules pass the verification, the business is determined to be green, and the verification explanation information lists the specific rules and standards that are met; if any rule fails, the business is determined to be non-green, and the non-compliant rule and core data discrepancies are explained in detail. Finally, the results are integrated to form a complete identification result containing classification results, compliance explanations, and verification explanations.

[0074] In the embodiments of this application, accurate identification is achieved through phased hierarchical verification. First, compliance verification is performed in conjunction with business scenarios, and classification results and compliance explanations are output simultaneously to ensure the interpretability of the judgment. Then, the results are cross-verified based on the expert rule base. The dual verification effectively suppresses the illusion of large models and improves the recognition accuracy and result stability. The process is completed by the domain large model in conjunction with the dynamic knowledge base, which adapts to policy updates and does not require full retraining of the large model, thus reducing operation and maintenance costs.

[0075] For example, based on the green business identification method of this application, a green business identification system is constructed, which consists of a data layer, a knowledge layer, a cognition layer, a rule layer, and an application layer.

[0076] Data Layer: Responsible for accessing multimodal data, including structured data (business information, transaction records) and unstructured data (project applications, environmental impact assessment reports, trade contracts, invoices). Optical character recognition (OCR) technology is used to parse unstructured data.

[0077] Knowledge Layer: A dynamic knowledge base for green finance is built based on a set of green industry standards and specifications. The set of green industry standards and specifications includes more than 500 green technology standards, more than 8,000 identification and judgment rules, and more than 400 industry-specific technical regulations, such as the Green and Low-Carbon Transformation Industry Guidance Catalogue.

[0078] Cognitive layer: Deploy a finely tuned domain-specific large model, utilize its reasoning capabilities, and combine it with retrieval-enhanced generation techniques to semantically match the extracted business information with the knowledge base.

[0079] Rule layer: Built-in hard constraint risk control rules (such as veto power for high-energy-consuming and high-polluting industries, verification of ecological protection red lines, etc.).

[0080] Application Layer: Receives business requests to be identified, connects with the data layer to perform multimodal data preprocessing and extract core information, and collaborates with the cognitive layer to achieve semantic matching between the core information and the dynamic knowledge base of the knowledge layer to obtain initial category results. Combined with risk control rules from the rules layer, hierarchical verification is performed, and finally, a green classification result is output, simultaneously providing the reasons for the judgment, policy basis, and specific clauses to support multi-dimensional green business identification.

[0081] By constructing a dynamic knowledge base and expert rule base incorporating green standards, and leveraging the powerful semantic understanding and reasoning capabilities of large-scale models, automated analysis of clients' business scope, project construction content, and fund usage is achieved. White-box identification of green business is realized, outputting not only green classification results but also detailed reasoning, supporting policy documents, and specific clauses, resolving interpretability issues. Rapid iteration of policy standards is enabled; by mounting the latest policy documents to the dynamic knowledge base, new regulations can take effect immediately without retraining the large-scale model (e.g., policy update training completed within 72 hours). Multi-dimensional intelligent judgment is achieved, simultaneously supporting the automated identification of green entities (e.g., clients, enterprises), green projects, and green vouchers (e.g., IOUs).

[0082] Figure 3 A flowchart illustrating the scenario compliance verification process of the green business identification method according to an embodiment of this application is shown. Figure 3As shown, according to the embodiments of this application, the business identification scenario includes at least one of the subject identification scenario, project identification scenario, and credential identification scenario. Based on the business identification scenario of the business to be identified, the core data is verified for scenario compliance through a domain big model, including steps S310 to S330.

[0083] In step S310, for the subject identification scenario, the energy efficiency of the main products is verified based on the domain big model and dynamic knowledge base.

[0084] According to embodiments of this application, based on a domain-wide model and a dynamic knowledge base, the energy efficiency of core data for main products is verified, including: extracting main business data from the core data, analyzing the green industry classification and confidence level corresponding to the main business data, and extracting main business information based on the main business data; under the condition that the confidence level meets the first level threshold, obtaining the intellectual property information corresponding to the main business information, and determining the consistency between the intellectual property information, the main business information, and the green industry classification; under the condition that the confidence level meets the second level threshold, extracting product energy efficiency standard data based on the main business information and the dynamic knowledge base, and comparing the energy efficiency testing data corresponding to the main business information with the product energy efficiency standard data item by item through the domain-wide model.

[0085] The core business data is precisely extracted from the preprocessed data, including the names of the top three products / businesses by revenue share, core product parameters, and business operation descriptions. A domain-wide model is invoked, combined with the green and low-carbon transformation industry guidance catalog and identification rules in the dynamic knowledge base, to perform semantic analysis on the core business data. This matches the data with corresponding green industry categories (such as new energy and clean energy equipment manufacturing, high-efficiency energy-saving product production, etc.), and generates an initial confidence level based on the degree of matching. Simultaneously, the model's information extraction capabilities are used to extract core core business information from the data, clarifying key elements such as core product models, production scale, and technological principles. For example, if no green industry category is matched, and the initial confidence level is below 50%, the entity is directly identified as a non-green entity whose business scope does not involve any green industries. If a green industry category is matched, and the initial confidence level is between 50% and 90%, the entity proceeds to the subsequent tiered threshold verification stage. If the initial confidence level directly reaches 90% (with a very few core indicators fully meeting the refined standards), the entity is tentatively designated as a potential green entity, requiring subsequent energy efficiency verification for confirmation.

[0086] The system determines whether the corresponding tiered threshold conditions are met based on the initial confidence level. If it falls within the first tier threshold (50%~70%), meaning the main business is highly related to the green industry but does not meet the detailed standard requirements, the system obtains the corresponding intellectual property information, including product-related patent names, patent types, technical features, and authorization status, by connecting to external data interfaces such as patent publication databases and intellectual property information platforms. The system then calls upon a domain-wide model to perform a deep consistency check between the intellectual property information, the main business information, and the matched green industry classifications, focusing on whether the patent technical features match the core technical requirements of the green industry classification and whether the main product functions are consistent with the low-carbon and environmentally friendly attributes of the green industry. For example: if the matching degree of the three reaches a preset threshold (e.g., above 85%), the original green industry classification and confidence level are maintained, and the entity is judged as a non-green entity whose business scope involves green industries (related to the main business but does not meet the detailed standards); if there is a significant inconsistency (e.g., the patented technology does not involve green core indicators), the confidence level is lowered to below 50%, or if there is no suitable green industry classification after rematching, the entity is judged as a non-green entity whose business scope does not involve any green industries; if the matching degree meets the standard but there is no further detailed standard verification basis, the judgment of "non-green entity whose business scope involves green industries" is maintained.

[0087] If the confidence level meets the second-level threshold (70%~90%), compliance needs to be confirmed through refined standards. Based on the product model and category in the main business information, the system accurately extracts the corresponding product energy efficiency standard data from the dynamic knowledge base, including refined energy efficiency level indicators, energy consumption limit standards, and environmental performance requirements. Subsequently, it calls the domain-wide model to compare the extracted main product energy efficiency test data (such as rated power, energy consumption value, emission reduction efficiency, etc.) with the extracted product energy efficiency standard data item by item, verifying whether the test data meets the minimum requirements stipulated by the standard. During the comparison process, the large model automatically identifies and converts data unit differences, determining the compatibility between the test data fluctuation range and the standard threshold. For example, if all energy efficiency indicators meet the standards, maintaining a high confidence level of 90% confirms compliance with the detailed standards for green industries, and the entity is ultimately identified as a green entity. If any indicator fails to meet the standards, the main product's energy efficiency is directly deemed non-compliant, and the confidence level is lowered to the 50%~70% range. Based on the relevance of the main business to the green industry, the entity is identified as a non-green entity whose business scope involves the green industry. If energy efficiency indicators are severely substandard, or if verification reveals that the core attributes of the main business have no substantial connection with the green industry, the confidence level is lowered to below 50%, and the entity is identified as a non-green entity whose business scope does not involve any green industry. If the initial confidence level is 70%~90% but no valid energy efficiency testing data is obtained, and the detailed standard verification cannot be completed, the entity is identified as a non-green entity whose business scope involves the green industry.

[0088] In the embodiments of this application, the energy efficiency verification of the main products is carried out according to the confidence level. First, business data is accurately extracted and industry classification is determined. Then, the consistency verification of intellectual property rights and industry classification is carried out by combining different thresholds, and the energy efficiency data is compared with the standard data item by item, so that the identification is more accurate and detailed. The domain large model and dynamic knowledge base are used to realize automated analysis and comparison, improve the verification efficiency, and ensure the accuracy and reliability of green business identification results.

[0089] In step S320, for the project identification scenario, the core project data is extracted, and the project data is verified against environmental protection standards based on the domain big model and dynamic knowledge base.

[0090] According to an embodiment of this application, the project data includes report data, address data, and environmental assessment data. Based on a domain-wide model and a dynamic knowledge base, the project data is verified against environmental standards. This includes: parsing the construction indicator information corresponding to the report data and comparing the construction indicator information with the green industry information in the dynamic knowledge base using the domain-wide model; performing spatial overlay analysis on the address data and the ecological protection red line database using the domain-wide model; and extracting environmental information from the environmental assessment data and comparing the environmental information with the environmental standard information in the dynamic knowledge base using the domain-wide model.

[0091] Green Industry Information Comparison: From project application documents, feasibility study reports, and other reports, the core construction indicator information is extracted using the text parsing capabilities of a large-scale model. This includes key aspects such as project production processes, product capacity, core equipment parameters, resource consumption quotas (e.g., energy and water consumption), and pollution control facility configuration. The large-scale model is then invoked, loading the corresponding green industry category's technical requirements from the dynamic knowledge base (e.g., pollution control technology standards and resource utilization efficiency indicators specified in the Green and Low-Carbon Transformation Industry Guidance Catalogue). The extracted construction indicator information is then compared item by item with the knowledge base content. The focus is on verifying whether the production process belongs to a green and low-carbon technology route, whether the core equipment meets environmental protection requirements, and whether resource consumption is below industry limits. For example, if the consistency between the construction indicators and green industry information reaches a preset threshold (e.g., above 90%), and no indicators fail to meet the standards, the construction content is deemed environmentally compliant and preliminarily identified as a green project. If some construction indicators meet green industry technical requirements, but one or more core processes, key equipment, or resource consumption fail to meet the standards, and the overall project is primarily non-green in function, it is determined to be a non-green project containing green content. If the construction indicators do not effectively align with green industry information, or if the core processes are outdated, the equipment seriously fails to meet environmental protection requirements, or the resource consumption far exceeds industry limits, the project will be directly marked as non-compliant with the construction indicators and judged as a non-green project that does not contain any green content.

[0092] Spatial Overlay Analysis: The project address data is analyzed using a large model to extract precise latitude and longitude coordinates. If the original address data is a text description (e.g., "XX City, XX District, XX Industrial Park"), it is converted to standardized latitude and longitude coordinates via a geocoding interface. The large model is then connected to the ecological protection red line database, and a spatial overlay analysis algorithm is used to spatially match the project address's latitude and longitude with the red line range data to determine whether the project site is within the ecological protection red line, permanent basic farmland, or other prohibited development areas. Simultaneously, national regional environmental protection standards in a dynamic knowledge base are used to verify whether the environmental carrying capacity of the project area is suitable for the project's construction scale. For example, if the site is not within the ecological protection red line or prohibited development areas, and the regional environmental carrying capacity is suitable for the project's construction scale, the address compliance verification passes. Combined with the construction indicator verification results, if the construction indicators meet the rules, it is ultimately identified as a green project; if the construction indicators partially meet the standards, it is identified as a non-green project with green content. If the site is within the red line range or exceeds the regional environmental carrying capacity, the address compliance verification fails, and regardless of whether the construction indicators meet green requirements, it is directly identified as a non-green project without any green content. If the address is compliant but the construction indicators do not meet the requirements of green industries, it will be judged as a non-green project that does not contain any green content.

[0093] Environmental Standard Information Comparison: Key environmental information is extracted from environmental assessment data such as environmental impact assessment reports using a domain-wide model. This includes the generation and emission volumes of major pollutants (such as waste gas, wastewater, and solid waste), pollutant treatment processes and efficiencies, operating parameters of environmental protection facilities, and environmental risk prevention and control measures. The large model retrieves corresponding national environmental standards (such as pollutant emission standards and environmental quality standards) from a dynamic knowledge base, and the extracted environmental information is compared with the standard information one by one. The focus is on verifying whether pollutant emission concentrations are below national standard limits, whether environmental protection measures meet emission standards, and whether environmental risk prevention and control comply with regulations. For example, if all environmental indicators meet the standards and the environmental information is compliant, combined with the compliance of construction indicators and site verification, the project is ultimately determined to be a green project. If some environmental indicators meet the standards and some construction indicators have green attributes, but the core environmental requirements are not met, the project is determined to be a non-green project with green content. If any indicator exceeds the standard or environmental protection measures are inadequate, the non-compliant item will be clearly marked, and the environmental standard verification will be deemed to have failed. Regardless of whether the construction indicators and the address are compliant, the project will be deemed a non-green project without any green content. If the construction indicators are only partially green and the environmental indicators fail to meet the standards, or if the address is compliant but neither the environmental protection nor the construction indicators have green attributes, the project will be uniformly deemed a non-green project without any green content.

[0094] In the embodiments of this application, multi-dimensional project data is broken down for project identification scenarios to carry out refined environmental protection standard verification. The data is compared and analyzed with the dynamic knowledge base and the ecological red line database from three aspects: construction indicators, address space, and environmental assessment. This achieves comprehensive verification of the green attributes of the project and improves the accuracy of green project identification. Each step uses a domain-wide model to complete intelligent analysis, replacing manual comparison and improving identification efficiency. At the same time, the verification basis is traceable, ensuring the interpretability and compliance of the judgment.

[0095] In step S330, for the credential recognition scenario, the subject classification result and project classification result corresponding to the credential to be recognized are obtained, and the transaction direction is verified by the domain big model based on the subject classification result and project classification result.

[0096] In the embodiments of this application, compliance verification is carried out in a targeted manner according to three scenarios: subject, project, and voucher. Each scenario is matched with exclusive verification logic to achieve multi-dimensional and accurate judgment of green business. The domain big model is used to link dynamic knowledge base to complete the verification of each scenario, which is in line with the recognition needs of different scenarios, improves the adaptability and accuracy of verification, and can output corresponding explanatory information to ensure the interpretability of the judgment, greatly improving the professionalism and comprehensiveness of green business recognition.

[0097] According to embodiments of this application, based on the subject classification results and project classification results, the core data is used to verify the transaction direction through a domain-wide model. This includes: determining the greenness of the project classification results; if the project classification result is not a green project, extracting the transaction voucher information of the core data and verifying the industry direction of the transaction voucher information through the domain-wide model; if the industry direction verification fails, determining the greenness of the subject classification results; and if the subject classification result is not a green subject, extracting the transaction business information of the core data, verifying the business direction of the transaction business information through the domain-wide model, and verifying the consistency between the fund payment object and the transaction object of the transaction voucher information.

[0098] The system uses a large model to read project classification results and determine their green attributes. First, it retrieves the corresponding identifiers for the project classification results (e.g., green project, non-green project with green content, non-green project without any green content). If the identifier is a green project, the transaction is directly confirmed to be strongly correlated with a green project, and the transaction is deemed green and compliant without further verification. If the project classification result is not a green project, the industry investment verification process is initiated: Transaction voucher information is extracted from core data, including key data such as the procurement content in the trade contract, the specifications and models of the transaction target, the names of goods and services corresponding to the invoice, and the description of the procurement purpose. The large model loads the green industry classification standards and identification rules from the dynamic knowledge base, and performs semantic parsing and industry investment matching on the transaction voucher information. The system focuses on verifying whether the procurement content falls within the green industry category of the Green and Low-Carbon Transformation Industry Guidance Catalogue and whether the transaction target possesses green attributes (e.g., energy-saving equipment, environmentally friendly materials). If the matching degree reaches a preset compliance threshold (e.g., above 80%), the industry investment verification is passed, and the transaction is green; if the threshold is not reached (e.g., procurement of non-green raw materials, non-environmentally friendly services), the industry investment verification fails, and the system proceeds to the main body greenness determination stage.

[0099] If the industry investment direction verification fails, the large model further reads the entity classification results (such as green customers, non-green customers whose business scope involves green industries, and non-green customers whose business scope does not involve any green industries) to determine the entity's green attributes. If the entity classification result is a green customer, the transaction investment direction is considered to be related to the main business of the green entity, and the transaction investment direction is determined to be green; if the entity classification result is not a green entity, a dual verification of business investment direction and payment consistency is initiated. On the one hand, transaction business information is extracted from the core data, including the explanation of the purpose of payment, the fund usage plan, and the service content in the business cooperation agreement, etc., and compared with the green business judgment standards in the dynamic knowledge base of the domain large model to verify whether the business investment direction belongs to the procurement of green industry-related products, project construction, or service acquisition (such as green technology R&D services, low-carbon project operation and maintenance), and to determine whether the business investment direction is green and compliant; on the other hand, the transaction counterparty name, unified social credit code, and other transaction object information in the transaction voucher information are extracted and compared with the fund payment object information in the core data (such as the name of the payee and account information in entrusted payment) to verify consistency and confirm whether the funds have actually been paid to the green business partner agreed in the transaction voucher. If the business direction verification passes and the payment recipient is consistent, the transaction direction is determined to be green; if any verification fails (such as the business direction not being green, or the payment recipient not matching the counterparty), the transaction direction is determined to be non-green, and the non-compliant item is clearly marked, providing the core basis for the final green certificate determination.

[0100] In the embodiments of this application, a layered and progressive transaction direction verification logic is adopted for the credential recognition scenario. First, the greenness of the project is verified, and then the greenness of the industry, entity, and business direction is verified. At the same time, the consistency between funds and transaction objects is also verified. This layered approach makes the determination of funds being invested in green businesses more accurate and effectively avoids the risk of greenwashing.

[0101] For example, taking the identification of green customers in a subject recognition scenario as an example, Figure 4 A schematic diagram illustrating a green customer identification flowchart of a green business identification method according to an embodiment of this application is provided. Figure 4 As shown, customer classification and whether they belong to the green industry and their green category are determined by their industry, main business revenue, and main products. The accuracy and confidence level of green customer identification can be graded at different stages based on the ease and certainty of information acquisition. 50% indicates that the business scope is only considered to be in the green industry, but there is a lack of evidence regarding the proportion of main business revenue. 70% indicates that the business scope is compliant, and the customer's main business is highly relevant to the green industry. 90% indicates that, based on the energy efficiency indicators and patent information of the main products, compliance with detailed national standards is confirmed.

[0102] First, the green classification is determined based on the company's business scope. The user initiates a customer identification request by opening the Smart Green Assistant (an intelligent tool for identifying green businesses in green finance scenarios), entering the company name, unified social credit code, or customer number based on the request. The system automatically retrieves the company's business scope based on the company name and unified social credit code by calling relevant query interfaces such as Intelligent Process Automation (IPA) and the business registration information database. The system then calls a large model to compare the business scope information with the green and low-carbon transformation industry guidance catalog in the green finance knowledge base to determine the possible green classification. For example, if the business scope includes highly polluting content, it is directly determined as non-green. If the business scope involves green manufacturing, it is identified as a possible green classification (e.g., "new energy and clean energy equipment manufacturing"). If it does not belong to any green industry, the system returns a message indicating that the customer is not a green customer (the business scope contains content that does not belong to a green classification, the main business does not belong to a green classification, or the main business products do not meet green industry standards), and provides the reasons for the judgment. If it belongs to a green industry, the system concludes that it is potentially green (potentially green), lists possible green classifications, and proceeds to the next step.

[0103] Secondly, determine the unique classification of the green industry based on the main business revenue source and increase the confidence level of the judgment. Since the industry has already been matched based on the business scope, multiple green industry subcategories may be matched in the end. It is necessary to narrow down the classification range to a unique category based on the main business revenue source. Determine whether multiple classification results exist. If multiple classifications exist, wait for the customer to upload relevant materials proving the main business (if the customer cannot provide relevant materials, give a conclusion that it is likely green and list possible green classifications), such as corporate financial statements, etc. The system calls the large model to determine the product with the highest main business revenue, and combines it with product patent information in the public database to locate which green industry category the customer's main business belongs to. If the main business belongs to the green industry, the customer is likely to be green. Further judgment is needed to determine whether the green main business products comply with relevant national standards.

[0104] Secondly, the system determines whether the company's core green business products comply with relevant national standards, specifically whether the energy efficiency rating of these products meets the detailed technical standards for green products. The Green and Low-Carbon Transformation Industry Guidance Catalogue provides detailed regulations on energy efficiency information for some products. Only products that meet the detailed standards of the Green Industry Guidance Catalogue can be classified as green. If the identified green industry subcategory has detailed national standards in the Green and Low-Carbon Transformation Industry Guidance Catalogue, the system uses a large-scale model to determine whether the energy efficiency information of the core products meets the detailed standards of the Green Industry Guidance Catalogue. If it does, the company can be basically confirmed to belong to the green industry. If it does not, the system returns a message indicating that the customer is not a green customer and provides the reasons for the judgment.

[0105] Finally, the classification results of the large model are verified through expert rule validation. The system calls the large model and combines it with the bank's green finance policy library (the revised content of special statistics on green loans in the green finance knowledge base, and the dictionary table of statistical classification of green loans) and expert rule library (expert rules for green industry identification) to cross-judge the classification results, providing the final result, the reasons for the customer's green classification, and risk warnings for the data source. Ultimately, customers are divided into three categories: green customers, non-green customers whose business scope involves green industries, and non-green customers whose business scope does not involve any green industries.

[0106] For example, taking the identification of green projects in a project identification scenario as an example, Figure 5 A schematic flowchart illustrating the green project identification process of the green business identification method according to an embodiment of this application is shown. Figure 5 As shown, the process for determining whether a project qualifies as a green project is based on its construction content and related business materials. The green project identification process is as follows:

[0107] Users open the Smart Green Identification Assistant to enter project name, project content, and other filing information, and upload business materials such as project application forms, feasibility study reports, and environmental impact assessment reports. The system automatically retrieves project ledger information from the Green Credit Management System (GCMS), calls a large model, and determines the green classification of the project's construction content based on the green finance knowledge base's green and low-carbon transformation industry guidance catalog. If the system determines that the project is not green, it will prompt the user with the judgment result, such as the project's construction content being unrelated to the green and low-carbon industry or the construction content not conforming to relevant national standards and specifications.

[0108] If the project falls under the green category, its compliance with national standards and regulations is determined based on the project's address data. The system extracts information about the project's renovation content and address, compares it with national standards and ecological protection red line boundaries listed in the Green Finance Knowledge Base's Green and Low-Carbon Transformation Industry Guidance Catalogue, and uses a large-scale model to determine whether the project's construction / renovation complies with relevant national standards and regulations. If the project does not comply with regulations or is located within an ecological protection red line area, it is deemed not a green project, such as if the project's construction content is unrelated to the green and low-carbon industry or does not comply with relevant national standards and regulations.

[0109] If the project meets national standards, the classification results are verified through expert rule validation, confirming the classification outcome determined by the large model. The system calls the large model and combines it with the bank's green finance policy database (the revised content of the special statistics on green funds in the green finance knowledge base, and the dictionary table of green fund statistical classification) and expert rule database (expert rules for green industry identification) to cross-judge the classification results. It determines whether the project complies with relevant policies and expert rules. If it does, a final green classification result is given; otherwise, the project is deemed not to belong to the green industry, and the reasons are provided, outputting the reasons for the green classification and a data source risk warning. Ultimately, the project is divided into three categories: green projects, non-green projects containing green content, and non-green projects containing no green content.

[0110] For example, taking the identification of green IOUs in a voucher recognition scenario as an example, Figure 6 A schematic flowchart illustrating a green loan document process according to an embodiment of this application for identifying green services is provided. Figure 6As shown, firstly, based on the identified projects and customers corresponding to the loan agreement, and according to the green project / customer identifier, determine the green classification results for the project (green project, non-green project with green content, non-green project without any green content) and the green classification results for the customer (e.g., green customer, non-green customer whose business scope involves green industries, non-green customer whose business scope does not involve any green industries). Then, link and inherit the green identification results for the loan agreement. If the project associated with the payment has already been identified as green, the green classification of the associated project is reused. If it is not a project payment, it is a working capital payment, and the customer's green classification is reused. If it is neither a green project payment nor a green customer payment, a judgment needs to be made based on whether the investment direction of the funds conforms to green standards.

[0111] The first step involves the user opening the Smart Green Assistant, entering the contract number, and uploading trade contracts, invoices, and other information. The system then searches for contract information within the system based on the contract number, locating the specific project and customer information, and matching it with the corresponding green label (green category). Based on the conditions, the system determines whether the payment is for a project. If so, it further determines whether it is a green project. If it is, the green project category is directly reused. Otherwise, it determines whether the project content involves green elements. The system calls a large model based on the Green Finance Knowledge Base's Green and Low-Carbon Transformation Industry Guidance Directory to determine the green category of the invoice / contract content / purpose of funds, and then proceeds to the second step. If it is not a green project, it determines whether the customer is a green customer. If so, the green customer category is directly reused. Otherwise, it determines whether the loan agreement involves green business. If it is a entrusted payment method, the large model is called to determine the green category of the invoice / contract content / purpose of funds, and proceeds to the second step. Otherwise, it is directly determined not to be a green loan agreement.

[0112] The second step is to verify the judgment results of the large model through expert rule validation. The system calls the large model and combines it with the bank's green finance policy database (the revised content of the special statistics on green loans in the green finance knowledge base, and the statistical classification dictionary table of green loans) and expert rule database (expert rules for green industry identification) to cross-judge the classification results. It determines whether the classification conforms to the relevant policies and expert rules. If it does, a final green classification result is given; if it does not, the loan agreement is determined to be a non-green loan agreement, and the reasons are output, along with a data source risk warning. Ultimately, the loan agreements are divided into two categories: green loan agreements and non-green loan agreements.

[0113] Based on the above embodiments, the beneficial effects of the green business identification method of this application are as follows:

[0114] High interpretability: While outputting the green classification results, the system can provide detailed reasons for the judgment (such as the enterprise is mainly engaged in lithium-ion battery manufacturing and meets the requirements of Chapter xx in the Green and Low-Carbon Transformation Industry Guidance Catalogue) and data sources, which solves the problem that traditional "black box" models cannot explain.

[0115] Low maintenance costs: When green finance standards are updated, only the documents and rules in the knowledge base need to be updated. With the help of the online learning mechanism, there is no need to retrain the underlying large model, which greatly reduces the operation and maintenance costs and response time.

[0116] High recognition accuracy: The dual guarantee of general understanding of the large model and strong verification of the expert rule base effectively suppresses the illusion problem of the large model, and the output results are stable and reliable.

[0117] Business process automation: It supports automatic capture of public data and automatic parsing of unstructured documents, which greatly reduces the time account managers spend manually entering and searching for policies, and realizes cost reduction and efficiency improvement in green finance business.

[0118] Based on the above-described green service identification method, embodiments of this application also provide a green service identification device. The following will combine... Figure 7 The device is described in detail.

[0119] Figure 7 A schematic block diagram of a green service identification device according to an embodiment of this application is shown.

[0120] like Figure 7 As shown, the green service identification device 1000 of this embodiment includes a data processing module 1010, a semantic matching module 1020, and a hierarchical verification module 1030.

[0121] The data processing module 1010 is used to acquire multimodal data of the service to be identified and preprocess the multimodal data to obtain core data. In one embodiment, the data processing module 1010 can be used to execute step S210 described above, which will not be repeated here.

[0122] The semantic matching module 1020 is used to perform semantic matching between the core data and the dynamic knowledge base based on retrieval enhancement generation technology and a domain-wide model to obtain initial category results. The dynamic knowledge base is dynamically updated based on a set of green industry standards and specifications; the domain-wide model is generated by fine-tuning and training the model based on historical green business data. In one embodiment, the semantic matching module 1020 can be used to execute step S220 described above, which will not be repeated here.

[0123] The hierarchical verification module 1030 is used to perform hierarchical verification on the core data based on the expert rule base, the dynamic knowledge base, and the domain big model when the initial category result is a potential green category, in order to generate an identification result; wherein, the expert rule base is constructed based on constraint risk control rules and green identification rules. In one embodiment, the hierarchical verification module 1030 can be used to execute step S230 described above, which will not be repeated here.

[0124] According to an embodiment of this application, the hierarchical verification module 1030 includes: a scenario compliance verification submodule, used to perform scenario compliance verification on the core data based on the business identification scenario of the business to be identified, through the domain big model, to generate green classification results and compliance explanation information; and a cross-validation submodule, used to perform cross-validation on the green classification results based on the expert rule base, through the domain big model, to generate the identification results and verification explanation information.

[0125] According to embodiments of this application, the business identification scenario includes at least one of a subject identification scenario, a project identification scenario, and a voucher identification scenario. The scenario compliance verification submodule includes: a subject identification unit, used for verifying the energy efficiency of the main products of the core data based on the domain big model and the dynamic knowledge base for the subject identification scenario; a project identification unit, used for extracting project data from the core data for the project identification scenario, and verifying the project data against environmental protection standards based on the domain big model and the dynamic knowledge base; and / or a voucher identification unit, used for obtaining the subject classification result and project classification result corresponding to the voucher to be identified for the voucher identification scenario, and verifying the transaction direction of the core data through the domain big model based on the subject classification result and the project classification result.

[0126] According to an embodiment of this application, the subject identification unit includes: a main business analysis subunit, used to extract main business data from the core data, analyze the green industry classification and confidence level corresponding to the main business data, and extract main business information based on the main business data; an industry determination subunit, used to obtain intellectual property information corresponding to the main business information when the confidence level meets a first grading threshold, and determine the consistency between the intellectual property information, the main business information and the green industry classification; and a business determination subunit, used to extract product energy efficiency standard data based on the main business information and the dynamic knowledge base when the confidence level meets a second grading threshold, and compare the energy efficiency testing data corresponding to the main business information with the product energy efficiency standard data item by item through the domain big model.

[0127] According to an embodiment of this application, the project data includes report data, address data, and environmental assessment data. The project identification unit includes: an industry information comparison subunit, used to parse the construction indicator information corresponding to the report data and compare the construction indicator information with the green industry information in the dynamic knowledge base through the domain big model; a spatial overlay analysis subunit, used to perform spatial overlay analysis on the address data and the ecological protection red line database through the domain big model; and an environmental protection analysis subunit, used to extract environmental protection information from the environmental assessment data and compare the environmental protection information with the environmental standard information in the dynamic knowledge base through the domain big model.

[0128] According to an embodiment of this application, the voucher identification unit includes: an industry investment verification subunit, used to determine the greenness of the project classification result; if the project classification result is not a green project, extracting the transaction voucher information of the core data and verifying the industry investment of the transaction voucher information through the domain big model; a subject determination subunit, used to determine the greenness of the subject classification result if the industry investment verification fails; and a business investment verification subunit, used to extract the transaction business information of the core data if the subject classification result is not a green subject, verifying the business investment of the transaction business information through the domain big model, and verifying the consistency between the fund payment object and the transaction object of the transaction voucher information.

[0129] According to an embodiment of this application, the semantic matching module 1020 includes: a retrieval submodule, used to retrieve green standard rule data associated with the core data in the dynamic knowledge base based on the retrieval enhancement generation technology; and a semantic fuzzy matching submodule, used to perform semantic fuzzy matching between the core data and the green standard rule data through the domain big model, and determine the initial category result based on the semantic fuzzy matching result and the dynamic knowledge base.

[0130] According to an embodiment of this application, the data processing module 1010 includes: a structured data extraction module, used to extract structured data from the multimodal data based on keywords; an unstructured data parsing module, used to parse the unstructured data of the multimodal data using optical character recognition technology to obtain parsed data; a key data extraction module, used to extract key data from the structured data and the parsed data based on the business recognition scenario; and a data standardization module, used to standardize the key data to obtain the core data.

[0131] According to embodiments of this application, any multiple modules of the data processing module 1010, semantic matching module 1020, and hierarchical verification module 1030 can be merged into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the data processing module 1010, semantic matching module 1020, and hierarchical verification module 1030 can be at least partially implemented as a hardware circuit, such as a field-programmable gate array, a programmable logic array, a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit, or any other reasonable means of integrating or packaging the circuit, or implemented in software, hardware, or firmware, or in any appropriate combination of any of these three implementation methods. Alternatively, at least one of the data processing module 1010, semantic matching module 1020, and hierarchical verification module 1030 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0132] Figure 8 A block diagram of an electronic device suitable for implementing a green business identification method according to an embodiment of this application is shown schematically.

[0133] like Figure 8 As shown, an electronic device 1200 according to an embodiment of this application includes a processor 1201, which can perform various appropriate actions and processes according to a program stored in a read-only memory 1202 or a program loaded from a storage portion 1208 into a random access memory 1203. The processor 1201 may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a dedicated microprocessor. The processor 1201 may also include onboard memory for caching purposes. The processor 1201 may include a single processing unit or multiple processing units for executing different steps of the method flow according to an embodiment of this application.

[0134] Random access memory 1203 stores various programs and data required for the operation of electronic device 1200. Processor 1201, read-only memory 1202, and random access memory 1203 are interconnected via bus 1204. Processor 1201 executes various steps of the method flow according to embodiments of this application by executing programs in read-only memory 1202 and / or random access memory 1203. It should be noted that the programs may also be stored in one or more memories other than read-only memory 1202 and random access memory 1203. Processor 1201 may also execute various steps of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0135] According to embodiments of this application, the electronic device 1200 may further include an input / output interface 1205, which is also connected to the bus 1204. The electronic device 1200 may also include one or more of the following components connected to the input / output interface 1205: an input section 1206 including a keyboard, mouse, etc.; an output section 1207 including a cathode ray tube, liquid crystal display, etc., and a speaker, etc.; a storage section 1208 including a hard disk, etc.; and a communication section 1209 including a network interface card, such as a local area network card, modem, etc. The communication section 1209 performs communication processing via a network such as the Internet. A drive 1210 is also connected to the input / output interface 1205 as needed. A removable medium 1211, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 1210 as needed so that computer programs read from it can be installed into the storage section 1208 as needed.

[0136] Embodiments of this application also provide a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0137] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof. In embodiments of this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include the read-only memory 1202, and / or random access memory 1203, and / or one or more memories other than read-only memory 1202 and random access memory 1203 described above.

[0138] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods provided in the embodiments of this application.

[0139] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 1209, and / or installed from the removable medium 1211. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0140] In embodiments of this application, the computer program can be downloaded and installed from a network via communication section 1209, and / or installed from removable medium 1211. When the computer program is executed by processor 1201, it performs the functions defined in the system of this application embodiment. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0141] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0142] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0143] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A method for identifying green services, characterized in that, The method includes: Acquire multimodal data of the service to be identified, and preprocess the multimodal data to obtain core data; Based on retrieval enhancement generation technology, the core data is semantically matched with a dynamic knowledge base using a domain-wide model to obtain initial category results; wherein, the dynamic knowledge base is dynamically updated based on a set of green industry standards and specifications; the domain-wide model is generated by fine-tuning and training a large model based on historical green business data; and If the initial category result is a potential green category, the core data is subjected to hierarchical verification based on the expert rule base, the dynamic knowledge base, and the domain big model to generate an identification result; wherein, the expert rule base is constructed based on constraint risk control rules and green identification rules.

2. The method according to claim 1, characterized in that, The process of performing hierarchical verification on the core data based on the expert rule base, the dynamic knowledge base, and the domain-wide model to generate recognition results includes: Based on the business identification scenario of the business to be identified, the core data is subjected to scenario compliance verification through the domain-wide model to generate green classification results and compliance explanation information; and Based on the expert rule base, the green classification results are cross-validated using the domain-wide model to generate the identification results and verification explanation information.

3. The method according to claim 2, characterized in that, The business identification scenario includes at least one of entity identification scenario, project identification scenario, and credential identification scenario. Based on the business identification scenario of the business to be identified, the core data undergoes scenario compliance verification through the domain-wide model, including: For the subject identification scenario, the energy efficiency of the main products is verified based on the domain big model and the dynamic knowledge base. For the project identification scenario, the core project data is extracted, and based on the domain-wide model and the dynamic knowledge base, the project data is verified against environmental protection standards; and / or For the aforementioned credential recognition scenario, the subject classification result and project classification result corresponding to the credential to be recognized are obtained, and the transaction direction verification of the core data is performed through the domain big model based on the subject classification result and the project classification result.

4. The method according to claim 3, characterized in that, The process of verifying the energy efficiency of main products based on the domain-wide model and the dynamic knowledge base includes: Extract the main business data from the core data, analyze the green industry classification and confidence level corresponding to the main business data, and extract the main business information based on the main business data; Under the condition that the confidence level meets the first classification threshold, the intellectual property information corresponding to the main business information is obtained, and the consistency between the intellectual property information, the main business information and the green industry classification is determined; Under the condition that the confidence level meets the second grade threshold, based on the main business information and the dynamic knowledge base, product energy efficiency standard data is extracted, and through the domain big model, the energy efficiency test data corresponding to the main business information is compared with the product energy efficiency standard data item by item.

5. The method according to claim 3, characterized in that, in, The project data includes report data, address data, and environmental assessment data. Based on the domain-wide model and the dynamic knowledge base, the project data undergoes environmental standard verification, including: The construction indicator information corresponding to the report data is analyzed, and the construction indicator information is compared with the green industry information in the dynamic knowledge base through the domain big model; Using the aforementioned large-scale model, the address data and the ecological protection red line database are spatially overlaid and analyzed; and Environmental information is extracted from the environmental assessment data, and then compared with environmental standard information in the dynamic knowledge base using the domain-wide model.

6. The method according to claim 3, characterized in that, The step of verifying the transaction direction of the core data using the domain-wide model based on the subject classification results and the project classification results includes: The project classification results are determined to be green. If the project classification results are not green, the transaction voucher information of the core data is extracted, and the industry investment direction of the transaction voucher information is verified through the domain big model. If the industry investment orientation verification fails, the greenness of the subject classification result will be determined; and If the subject classification result is not a green subject, the transaction business information of the core data is extracted. The business direction of the transaction business information is verified through the domain big model, and the consistency between the fund payment object and the transaction object of the transaction voucher information is verified.

7. The method according to any one of claims 1 to 6, characterized in that, The retrieval-enhanced generation technology uses a domain-wide model to semantically match the core data with a dynamic knowledge base to obtain initial category results, including: Based on the aforementioned retrieval enhancement generation technology, green standard rule data associated with the core data is retrieved from the dynamic knowledge base; and Using the domain-wide model, the core data is semantically fuzzy matched with the green standard rule data, and the initial category result is determined based on the semantic fuzzy matching result and the dynamic knowledge base.

8. The method according to any one of claims 2 to 6, characterized in that, The preprocessing of the multimodal data to obtain core data includes: Based on keywords, extract structured data from the multimodal data; The unstructured data of the multimodal data is analyzed using optical character recognition technology to obtain analyzed data; Based on the aforementioned business identification scenario, key data is extracted from the structured data and the parsed data; and Standardize the key data to obtain the core data.

9. A green business identification device, characterized in that, The device includes: The data processing module is used to acquire multimodal data of the business to be identified and preprocess the multimodal data to obtain core data; A semantic matching module is used to perform semantic matching between the core data and a dynamic knowledge base based on retrieval enhancement generation technology and a domain-wide model to obtain initial category results. The dynamic knowledge base is dynamically updated based on a set of green industry standards and specifications. The domain-wide model is generated by fine-tuning and training the model using historical green business data. The hierarchical verification module is used to perform hierarchical verification on the core data based on the expert rule base, the dynamic knowledge base, and the domain big model when the initial category result is a potential green category, so as to generate a recognition result; wherein, the expert rule base is constructed based on constraint risk control rules and green recognition rules.

10. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 8.

12. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 8.

Images