Analysis report generation method and device, medium and program product

Abstract

The invention provides an analysis report generation method and device, equipment, a storage medium and a program product, can be applied to the fields of artificial intelligence and big data, and relates to application of a large model in a financial science and technology scene. The method comprises the steps that to-be-evaluated parameters and a plurality of to-be-executed service nodes are acquired based on service requirements, the to-be-evaluated parameters at least comprise one of a to-be-evaluated industry, the name of a to-be-evaluated enterprise and to-be-evaluated indexes, and each to-be-executed service node is used for executing one type of analysis report generation operation based on the to-be-evaluated parameters; based on the to-be-evaluated parameters and the plurality of to-be-executed service nodes, obtaining different types of analysis reports corresponding to the to-be-executed service nodes by using a large model; and integrating the analysis reports of all the to-be-executed service nodes to obtain a target analysis report.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and big data, and specifically to the application of large-scale models in fintech scenarios. More specifically, it relates to a method, apparatus, device, medium, and program product for generating analytical reports. Background Technology

[0002] In the due diligence phase for corporate clients, the financial industry needs to conduct extensive research on clients' industry policies, corporate information, financial statements, legal proceedings, and other information. The investigation reports involve a wide range of diverse data sources. Currently, industry analysis reports mostly rely on manual retrieval or web crawling. This not only requires a lot of time to collect data but also makes it easy to miss important information. Moreover, manually written analysis reports also suffer from problems such as inconsistent formatting and untimely data updates, which seriously affect the quality and efficiency of report generation. Summary of the Invention

[0003] In view of the above problems, this application provides analytical report generation methods, apparatus, equipment, media, and program products that improve the quality and efficiency of analytical reports.

[0004] According to the first aspect of this application, a method for generating an analysis report is provided, comprising: obtaining parameters to be evaluated and multiple business nodes to be executed based on business needs, wherein the parameters to be evaluated include at least one of the following: the industry to be evaluated, the name of the company to be evaluated, and the indicator to be evaluated; each business node to be executed is used to perform a type of analysis report generation operation based on the parameters to be evaluated; obtaining different types of analysis reports corresponding to each business node to be executed using a large model based on the parameters to be evaluated and the multiple business nodes to be executed; and integrating the analysis reports of all business nodes to be executed to obtain a target analysis report.

[0005] According to an embodiment of this application, based on the parameters to be evaluated and multiple business nodes to be executed, a large model is used to obtain different types of analysis reports corresponding to each business node to be executed, including: for any business node to be executed, retrieving information from a database according to the parameters to be evaluated and the type of analysis report corresponding to the current business node to be executed, wherein the retrieval information is related to the parameters to be evaluated and the current business node to be executed, and the database includes intermediate data generated during the training of the large model; inputting the retrieval information and its corresponding prompt words into the large model to obtain the analysis report of the current business node to be executed, wherein the prompt words are obtained based on a prompt word template associated with the current business node to be executed.

[0006] According to an embodiment of this application, based on the parameters to be evaluated and the type of analysis report corresponding to the current business node to be executed, retrieval information is obtained from the database, including: obtaining preliminary information from the database based on the parameters to be evaluated and the type of analysis report corresponding to the current business node to be executed, and determining the value assessment dimension of the preliminary information, wherein the preliminary information is related to the parameters to be evaluated and the current business node to be executed, and the value assessment dimension includes at least one of the following: the timeliness of the preliminary information, the relevance of the preliminary information to the parameters to be evaluated, and the relevance of the preliminary information to the current business node to be executed; assigning weights to the preliminary information based on the value assessment dimension; and using the preliminary information whose weights meet a preset weight threshold as retrieval information.

[0007] According to an embodiment of this application, the method further includes: determining the dependency relationship between each business node to be executed based on business requirements, wherein the dependency relationship indicates the connection information between the previous business node to be executed and the current business node to be executed; determining the execution priority of each business node to be executed based on the dependency relationship; and obtaining the analysis report of each business node to be executed sequentially according to the execution priority.

[0008] According to an embodiment of this application, the analysis reports of each business node to be executed are obtained sequentially according to the execution priority, including: during the execution of operations at each business node to be executed, obtaining current retrieval information based on the parameters to be evaluated and the type of analysis report corresponding to the current business node to be executed; verifying the correctness of dependencies based on the current retrieval information; if the dependencies are incorrect or the business requirements change, pausing the operations performed by the current business node to be executed and adjusting the dependencies; updating the execution priority based on the adjusted dependencies; and obtaining the analysis reports of each business node to be executed sequentially according to the updated execution priority.

[0009] According to an embodiment of this application, the retrieval information and its corresponding prompt words are input into a large model to obtain an analysis report of the current business node to be executed. This includes: for any business node to be executed, inputting the retrieval information and prompt words into the large model to obtain an initial analysis report of the current business node to be executed. Each business node to be executed corresponds to a prompt word template, which includes system-level prompts, task-level prompts, constraint prompts built based on business requirements, and inference step prompts built based on associated business nodes to be executed. System-level prompts are used to constrain the roles and output requirements of the large model, and task-level prompts are used to constrain the inference tasks of the large model. The first metadata information of the initial analysis report is extracted, including element information and element data. The system retrieves the source information of the information; compares the first-level metadata information with the prompt words, calculates the evaluation indicators of the initial analysis report, wherein the evaluation indicators include at least one of the following: completeness of element information, completeness of reasoning steps, and reliability of source information; if any evaluation indicator does not meet its own preset threshold, adjusts the prompt words; inputs the adjusted prompt words into the large model and re-acquires the initial analysis report; based on the re-acquired initial analysis report, iterates the calculation of evaluation indicators, adjustment of prompt words, and re-acquisition of the initial analysis report until preset conditions are met, wherein preset conditions include reaching a preset number of iterations or the current evaluation indicators meeting their respective preset thresholds; and uses the initial analysis report that meets the preset conditions as the analysis report of the current business node to be executed.

[0010] According to an embodiment of this application, the initial analysis report is obtained in the following manner: Multiple text fragments and secondary metadata information of the text fragments are obtained from the search information. The secondary metadata information includes keywords, source information, business rules, and a knowledge graph. The knowledge graph is obtained based on entity extraction from the multiple text fragments. The business rules include at least one of the following: risk control rules, credit granting rules, and approval rules. The prompts corresponding to the search information are parsed to construct the architecture of the analysis report. Multiple text fragments, keywords, source information, and the knowledge graph are integrated to obtain structured information. The structured information is matched with the business rules to obtain matching results. The structured information and matching results are written into the corresponding parts of the architecture to obtain the initial analysis report.

[0011] A second aspect of this application provides an analysis report generation apparatus, comprising: an evaluation information acquisition module, used to acquire evaluation parameters and multiple execution business nodes based on business needs, wherein the evaluation parameters include at least one of the following: the industry to be evaluated, the name of the enterprise to be evaluated, and the evaluation indicator, and each execution business node is used to perform a type of analysis report generation operation based on the evaluation parameters; a node analysis report acquisition module, used to acquire different types of analysis reports corresponding to each execution business node using a large model based on the evaluation parameters and multiple execution business nodes; and a target analysis report acquisition module, used to integrate the analysis reports of all execution business nodes to obtain a target analysis report.

[0012] A third aspect of this application provides an electronic device 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.

[0013] A fourth aspect of this application also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0014] The fifth aspect of this application also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method. Attached Figure Description

[0015] 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:

[0016] Figure 1 The illustrations depict application scenarios of the analysis report generation method, apparatus, device, medium, and program product according to embodiments of this application.

[0017] Figure 2 A flowchart illustrating an analysis report generation method according to an embodiment of this application is shown schematically;

[0018] Figure 3 The illustration shows a schematic diagram of a method for obtaining an analysis report of a business node to be executed according to an embodiment of this application;

[0019] Figure 4 This illustration schematically shows a method for obtaining retrieval information according to an embodiment of this application;

[0020] Figure 5 This illustration schematically shows a method for obtaining an analysis report of a business node to be executed based on a large model according to an embodiment of this application;

[0021] Figure 6 This illustration shows a schematic diagram of a method for obtaining the execution order of each business node to be executed according to an embodiment of this application;

[0022] Figure 7 This illustration schematically shows a method for obtaining analysis reports of each business node to be executed based on the execution order according to an embodiment of this application;

[0023] Figure 8 This schematically illustrates a structural block diagram of an analysis report generation apparatus according to an embodiment of this application; and

[0024] Figure 9 A block diagram schematically illustrates an electronic device suitable for implementing an analysis report generation method according to an embodiment of this application. Detailed Implementation

[0025] 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.

[0026] 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 the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0027] 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.

[0028] 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.).

[0029] In the technical solution of this application, the 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.

[0030] In scenarios involving automated decision-making using personal information, the methods, devices, and systems provided in this application all offer users corresponding entry points for choosing to agree to or reject the automated decision-making results. If the user chooses to reject, the process proceeds to the expert decision-making stage. Here, "automated decision-making" refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests, or economic, health, and credit status through computer programs, and then making a decision. Here, "expert decision-making" refers to the activity of making decisions by personnel who specialize in a particular field, possess specialized experience, knowledge, and skills, and have reached a certain level of professional expertise.

[0031] The embodiments of this application provide an analysis report generation method, which aims to solve the problems of poor quality and low efficiency in existing report generation methods. Based on business needs, the method determines the business nodes to be executed, and each business node is used to perform a type of analysis report generation operation. By obtaining the analysis reports of each business node to be executed, a high-quality target analysis report is obtained.

[0032] Figure 1 The diagram illustrates an application scenario of the analysis report generation method according to an embodiment of this application.

[0033] like Figure 1 As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 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 third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0034] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0035] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0036] 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, the second terminal device 102, and the third terminal device 103 (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. For example, users can send information to be evaluated to server 105 through the first terminal device 101, the second terminal device 102, and the third terminal device 103. Server 105 processes the information and feeds back the generated target analysis report to the terminal devices. The information to be evaluated may include the industry to be evaluated, the name of the company to be evaluated, and the indicators to be evaluated.

[0037] It should be noted that the analysis report generation method provided in this application embodiment can generally be executed by server 105. Correspondingly, the analysis report generation device provided in this application embodiment can generally be located in server 105. The analysis report generation method provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the analysis report generation device provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0038] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0039] The following will be based on Figure 1 The described scene, through Figures 2-7 The method for generating analysis reports according to embodiments of this application will be described in detail.

[0040] Figure 2 A flowchart illustrating an analysis report generation method according to an embodiment of this application is shown schematically.

[0041] like Figure 2 As shown, the analysis report generation method 200 of this embodiment includes operations S210 to S230.

[0042] In operation S210, based on business requirements, parameters to be evaluated and multiple business nodes to be executed are obtained. The parameters to be evaluated include at least one of the following: the industry to be evaluated, the name of the company to be evaluated, and the indicator to be evaluated. Each business node to be executed is used to perform a type of analysis report generation operation based on the parameters to be evaluated.

[0043] When operating S220, based on the parameters to be evaluated and multiple business nodes to be executed, a large model is used to obtain different types of analysis reports corresponding to each business node to be executed.

[0044] By operating S230, the analysis reports of all pending business nodes are integrated to obtain the target analysis report.

[0045] For example, users can input evaluation parameters and select execution nodes on the interface of the visual analysis process orchestration engine according to business needs. The large model generates analysis reports corresponding to each execution node based on the evaluation parameters and the selected execution nodes. Finally, all analysis reports are integrated to obtain the target analysis report. The visual analysis process orchestration engine then converts the target analysis report into visual elements and displays them to the user through its interface. Furthermore, these visual elements can be various types of charts and standardized visual reports. This report generation method not only allows users to retrieve relevant information about interested companies or industries in real time on the visual analysis process orchestration engine interface and generate corresponding analysis reports, but also quickly generates standardized analysis reports that conform to industry norms, thus improving report generation quality.

[0046] For example, the business requirement is to analyze a company's credit approval status from 2021 to 2023. Based on this business requirement, the parameters to be evaluated can be the company's name or unified social credit code, the indicators to be evaluated (such as credit approval status), the industry the company is involved in, and the years to be evaluated (2021 to 2023), etc. The types of analysis reports involved in this business requirement may include a summary of the company's development, an interpretation report of credit approval policies, and an analysis report of credit approval status. Therefore, the business nodes to be executed based on this business requirement can include business nodes used to generate these types of analysis reports. Based on these parameters to be evaluated and the selected business nodes to be executed, the large model can be used to generate analysis reports corresponding to each business node to be executed, thereby obtaining the target analysis report.

[0047] In some embodiments, during operation S210, the analysis report type includes at least one of the following: industry development analysis, policy and regulation interpretation, enterprise overview analysis, communication strategy recommendations, and financial service solutions.

[0048] In some embodiments, during operation S220, a large model can refer to a deep learning model with a large number of model parameters. A large model typically contains hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters. Large models can include large-scale language models (LLMs), large visual models, multimodal large models, and so on. The large model involved in the embodiments of this disclosure can be a general-purpose large model, or it can be an expert large model obtained after fine-tuning based on requirements; the embodiments of this disclosure do not limit this.

[0049] In some embodiments, after obtaining the analysis reports of each business node to be executed in operation S230, the analysis reports are structured and integrated according to the preset target analysis report template, and the analysis reports are integrated into one report to obtain the target analysis report. The target analysis report can be displayed to the user through the visual analysis process orchestration engine interface for the user to review or modify.

[0050] According to the embodiments of this application, multiple business nodes to be executed are determined based on business needs, and different types of analysis reports for each business node to be executed are obtained using a large model. This allows for the evaluation of customer or enterprise information from multiple dimensions. This intelligent method of obtaining analysis reports can improve report generation efficiency and avoid information omissions and subjectivity issues caused by manual report writing, thereby improving the quality of report generation.

[0051] Figure 3 The illustration shows a schematic diagram of a method for obtaining an analysis report of a business node to be executed according to an embodiment of this application.

[0052] like Figure 3 As shown, the method for obtaining the analysis report of the business node to be executed in this embodiment includes operations S310 to S320.

[0053] In operation S310, for any business node to be executed, retrieval information is obtained from the database based on the parameters to be evaluated and the type of analysis report corresponding to the current business node to be executed. The retrieval information is related to the parameters to be evaluated and the current business node to be executed, respectively. The database includes intermediate data generated during the training of the large model.

[0054] When operating S320, the retrieved information and its corresponding prompt words are input into the large model to obtain the analysis report of the current business node to be executed. The prompt words are obtained based on the prompt word template associated with the current business node to be executed.

[0055] In some embodiments, during operation S310, assuming the parameters to be evaluated are a company name, company development prospects, report format, and report depth, the analysis report includes an analysis report on the development of the industry the company is involved in, an interpretation of relevant industry policies, and a summary analysis of the company, information related to the company's industry, relevant industry policies, the company's basic information, and information related to the report format are retrieved from the database to obtain the search information. Further, assuming the company's industry is new energy battery manufacturing, the search information should include the development trend and market size of the new energy battery manufacturing industry, subsidy policies and environmental protection requirements for the new energy battery manufacturing industry, the company's operating status and technological level, basic information about the company's competitors in the new energy battery manufacturing industry, and relevant information on various types of analysis report formats. For customer acquisition scenarios, the analysis report also includes communication strategy suggestions and financial service solutions. The type of analysis report varies depending on the business scenario.

[0056] In some embodiments, during operation S310, the intermediate data generated during the large model training process includes professional terms in various fields, historical keywords obtained based on professional terms, source information of historical keywords, business rules involved in historical evaluation materials, and historical knowledge graphs constructed based on historical evaluation materials. Historical evaluation materials may include basic information of customers or enterprises, financial statements, and industrial policies, etc. The business rules may be: if a customer is a high-level customer and the customer's recent transaction amount is greater than a preset transaction threshold, a manual review mechanism is triggered to manually review the credit risk of the customer. Alternatively, if an enterprise's historical transaction records show no bad transaction behavior, but bad transaction behavior has frequently occurred recently, a risk control strategy matching the enterprise can be selected from a preset risk control strategy library to control the risk of the enterprise.

[0057] The training process of the large model is as follows: Historical assessment materials are used as training samples. These samples come from a wide range of sources, including customer credit records, transaction logs, contract texts, internal company reports, industry research reports, policies and regulations of various industries, news and information related to various industries or customers, and financial databases. During training, the large model first preprocesses the training samples, including data cleaning, noise reduction, removal of outliers and duplicates, and standardization. Then, based on a pre-defined knowledge base, it identifies the corresponding business terms in the preprocessed data. This knowledge base includes mapping relationships between business terms and data in various fields, product manuals for various industries, and mapping relationships between various business rules and data. Based on the identified business terms, it extracts key information such as historical keywords from the training samples and obtains the source information of this key information. Based on key information, business rules matching the training samples are determined from a pre-set knowledge base. Entities are extracted from the training samples to obtain a historical knowledge graph. This historical knowledge graph includes information such as entities, entity attributes, and entity sources. For example, an entity might be "credit card pre-authorization," with the attribute "validity period of 30 days" and the source "business manual." Another example is "company name," with the attribute "emerging enterprise" and the source "internal company report." Next, long texts in the training samples are segmented and vectorized to obtain vectorized text. Corresponding historical prompts are set based on the training samples. Using the vectorized text, key information, business rules matching the training samples, and the historical knowledge graph, combined with the historical prompts and existing analysis reports corresponding to historical evaluation materials as annotations, the general large language model is fine-tuned to obtain a trained large model. The vectorized text, key information, business rules matching the training samples, and historical knowledge graph from the large model training process are all stored in a database for subsequent retrieval. Furthermore, the purpose of vectorizing the segmented text data is to convert it into a format suitable for fine-tuning the large model.

[0058] The fine-tuning process of the large model is as follows: A historical architecture for the analysis report is constructed based on historical prompts. For example, if historical prompts include the analysis year and the target company's competitors, the final generated analysis report should include these two sections. Vectorized text, key information, source information of the key information, and historical knowledge graphs are integrated to obtain historical structured information. This historical structured information is matched with the business rules corresponding to the training samples to obtain historical matching results. The historical matching results and historical structured information are written into the corresponding parts of the historical architecture, and the analysis report is output. Based on the output analysis report and the labeled existing analysis reports, the loss function is calculated. The model parameters are adjusted according to the loss function. Based on the adjusted model parameters, the above fine-tuning process is repeated until the loss function converges. This fine-tuning method can significantly improve the large model's semantic understanding, logical reasoning, information extraction, and compliance expression capabilities in different scenarios, and enhance its responsiveness to prompts.

[0059] According to the embodiments of this application, based on the type of analysis report of the parameter to be evaluated and the business node to be executed, relevant retrieval information can be obtained from the database. This allows for the rapid and accurate acquisition of information related to the parameter to be evaluated on the business node to be executed. By inputting the retrieval information and its prompts into the large model, the efficiency of generating analysis reports can be improved.

[0060] Figure 4 The illustration shows a schematic diagram of a method for obtaining retrieval information according to an embodiment of this application.

[0061] like Figure 4 As shown, the information retrieval method of this embodiment includes operations S410 to S430.

[0062] In operation S410, preliminary information is obtained from the database based on the type of analysis report corresponding to the parameter to be evaluated and the current business node to be executed, and the value assessment dimensions of the preliminary information are determined. The preliminary information is related to the parameter to be evaluated and the current business node to be executed, and the value assessment dimensions include at least one of the following: the timeliness of the preliminary information, the relevance of the preliminary information to the parameter to be evaluated, and the relevance of the preliminary information to the current business node to be executed.

[0063] When operating S420, weights are assigned to preliminary information based on the value assessment dimension.

[0064] In operation S430, preliminary information whose weights meet the preset weight threshold is used as retrieval information.

[0065] For example, the parameters to be evaluated are the company name and its credit risk situation over the past three months. The analysis report types include company overview analysis report, company operating status analysis report, and company credit risk analysis report. If the preliminary information includes the company's credit risk situation over the past year, the credit risk situation of its competitors, and basic information about the customers served by the company, since users are not concerned about the company's credit risk situation in the previous nine months, a lower weight is assigned to the credit risk situation in the previous nine months. This reflects the timeliness of the preliminary information as a value assessment dimension. Similarly, since users are not concerned about the situation of competitors, a lower weight is assigned to the credit risk situation of competitors. This reflects the relevance of the preliminary information to the parameters to be evaluated as a value assessment dimension. Furthermore, since the basic information of the customers served by the company has a weak correlation with the company's credit risk analysis report, a lower weight is assigned to the basic information of the customers served by the company. This reflects the relevance of the preliminary information to the business nodes to be executed as a value assessment dimension. Finally, preliminary information that meets the preset weight threshold is used as retrieval information. By assigning weights to the preliminary information and obtaining retrieval information, the compliance and accuracy of the analysis report can be greatly improved in the subsequent analysis report generation stage.

[0066] According to embodiments of this application, by assigning weights to the initially retrieved information and using the initial information whose weights meet a preset weight threshold as the retrieved information, the accuracy of the retrieved information can be improved, thereby enhancing the accuracy of the analysis report.

[0067] Figure 5 The illustration shows a schematic diagram of a method for obtaining an analysis report of a business node to be executed based on a large model according to an embodiment of this application.

[0068] like Figure 5 As shown, the method for obtaining the analysis report of the business node to be executed based on the large model in this embodiment includes operations S510 to S570.

[0069] In operation S510, for any business node to be executed, the retrieved information and the prompt words are input into the large model to obtain the initial analysis report of the current business node to be executed. Each business node to be executed corresponds to a prompt word template. The prompt word template includes system-level prompts, task-level prompts, constraint prompts built based on business requirements, and inference step prompts built based on associated business nodes to be executed. System-level prompts are used to constrain the roles and output requirements of the large model, and task-level prompts are used to constrain the inference tasks of the large model.

[0070] In operation S520, the first metadata information of the initial analysis report is extracted, which includes feature information and source information of the feature information.

[0071] In operation S530, the first metadata information is compared with the prompt words to calculate the evaluation index of the initial analysis report. The evaluation index includes at least one of the following: the completeness of the element information, the completeness of the reasoning steps, and the reliability of the source information.

[0072] When operating S540, if an evaluation indicator does not meet its preset threshold, the prompt word will be adjusted.

[0073] When operating the S550, input the adjusted prompts into the large model and retrieve the initial analysis report again.

[0074] In operation S560, based on the reacquired initial analysis report, the evaluation index calculation, prompt word adjustment, and initial analysis report reacquisition are iterated until the preset conditions are met. The preset conditions include reaching a preset number of iterations or the current evaluation index meeting its respective preset threshold.

[0075] When operating S570, the initial analysis report that meets the preset conditions is used as the analysis report for the current business node to be executed.

[0076] In some embodiments, during operation S510, system-level prompts may include: setting the role of the large model, clarifying the analysis objective (e.g., generating a rigorous due diligence report for a specific company), clarifying output requirements (e.g., requiring the output analysis report to have a clear structure and rigorous logic), or constraining the compliance of the analysis report (e.g., requiring the analysis report to comply with relevant regulations of credit approval policies); task-level prompts may be precise instructions set for specific business nodes to be executed, to clarify the specific operations that the large model needs to perform, such as extracting information on key competitors and comparing their advantages and disadvantages, interpreting the specific impact of a certain policy on a certain industry, etc., and may also set specific instructions based on the search information and the parameters to be evaluated; constraint prompts may be "provide an analysis report only based on the provided search information" or "output..." The analysis report must conform to a specific format requirement, or it can specify "avoid generating search information or content not mentioned in the parameters to be evaluated." The inference step prompts provide the steps required for the large model to perform inference. For example, the inference step prompts could be "first analyze industry trends, second assess the company's positioning, and finally identify the company's risk points and provide risk control suggestions." These inference step prompts essentially integrate inference thought chains. Different inference thought chains can be set for different business scenarios (such as due diligence, risk control, and marketing), or thought chains can be set based on business needs to guide the large model in deep inference. For example, in the process of obtaining policy analysis reports, an inference thought chain such as "identify the policy - obtain the policy's scope of impact - identify company risks - provide response strategies" can be set to guide the large model in deep inference. Furthermore, the prompt word templates can be reused by other pending business nodes, and users can add parameters according to their actual needs. For example, parameters such as company name and industry code can be added to the prompt word templates.

[0077] In some embodiments, during operation S530, the semantic features of the prompt word are compared with the semantic features of the element information to determine whether the element information fully covers the content indicated by the prompt word, that is, to assess the completeness of the element information. The reasoning steps given in the prompt word are compared with the reasoning steps in the first metadata information to determine the completeness of the reasoning steps in the first metadata information. Based on the source information in the first metadata information, the reliability of the source information is determined. If the source information of the element information in the first metadata information is missing or incorrect, then the source information is unreliable and has low reliability.

[0078] According to the embodiments of this application, the analysis report generated based on the large model is further adjusted by modifying the prompt words to improve the accuracy of the analysis report and make it more in line with business needs.

[0079] In some embodiments, the initial analysis report is obtained by: obtaining multiple text fragments and secondary metadata information of the multiple text fragments from the search information, wherein the secondary metadata information includes keywords, source information, business rules, and a knowledge graph, the knowledge graph being obtained based on entity extraction from the multiple text fragments, and the business rules including at least one of the following: risk control rules, credit granting rules, and approval rules; parsing the prompt words corresponding to the search information to construct the architecture of the analysis report; integrating multiple text fragments, keywords, source information, and knowledge graph to obtain structured information; matching the structured information with the business rules to obtain matching results; and writing the structured information and matching results into the corresponding parts of the architecture to obtain the initial analysis report.

[0080] For example, for any business node to be executed, the process of generating an analysis report in each iteration of the large model is as follows, taking the generation of the initial analysis report as an example: Suppose that the analysis report type to be generated for a certain business node to be executed is a policy and regulation interpretation report for the financial industry. The retrieved information includes policy text fragments in the financial industry. Multiple vectorized text fragments and their corresponding secondary metadata information are obtained from these policy text fragments. For example, a text fragment could be "regulating the business scope of third-party institutions". Its corresponding secondary metadata information includes keywords (third-party institutions, business scope), source information (e.g., a certain policy document), business rules (e.g., risk control rules), and the knowledge graph involving the business scope of third-party institutions. The prompt words corresponding to this retrieval information could be: "You are a financial industry policy analyst. Based on the provided policy text fragments, summarize the main policies that have affected the financial industry in the past year and analyze the specific impact of each policy. You are required to provide an analysis report by following the steps of identifying key policies, extracting core clauses, and identifying related enterprise businesses. The analysis report should include three parts: policy name and release time, core content of the policy, and impact on enterprises." Based on the prompts, the structure of the analysis report can be obtained. The analysis report for the business node to be executed should include three main sections: key policies, core terms, and related enterprise businesses. The policy-related sections should include the policy name and release date, the core content of the policy, and the policy's impact on the enterprise. By integrating text fragments, keywords, source information, and knowledge graphs, structured information is obtained. The structured information is then matched with risk control rules to obtain matching results. The structured information and matching results are written into the corresponding parts of the analysis report structure to obtain the initial analysis report for the business node to be executed.

[0081] According to the embodiments of this application, the architecture of the analysis report based on prompt words can ensure that the structure of the analysis report is clear, that each module in the analysis report matches the user's needs, and that structured information and rule matching results are obtained based on text fragments and their secondary metadata information, thereby improving the reliability and interpretability of the analysis report.

[0082] Figure 6 The illustration shows a schematic diagram of a method for obtaining the execution order of each business node to be executed according to an embodiment of this application.

[0083] like Figure 6 As shown, the method for obtaining the execution order of each business node to be executed in this embodiment includes operations S610 to S630.

[0084] When operating S610, the dependencies between each business node to be executed are determined according to business requirements. The dependencies indicate the connection information between the previous business node to be executed and the current business node to be executed.

[0085] When operating S620, the execution priority of each business node to be executed is determined based on the dependency relationship.

[0086] When operating the S630, the analysis reports of each pending business node are obtained sequentially according to the execution priority.

[0087] For example, if the business requirement is to analyze the market risk of a certain industry, and obtaining the market risk report requires a policy analysis report, then the business node for obtaining the industry's policy analysis report must be executed first, followed by the business node for obtaining the industry's market risk analysis report. In other words, the obtained industry policy analysis report is used as input for the business node for obtaining the industry's market risk analysis report. In this way, based on the dependencies between the business nodes, each business node is called sequentially to obtain the analysis report. For other business requirements, based on the dependencies between the business nodes, each business node is called sequentially to obtain the corresponding analysis report.

[0088] According to the embodiments of this application, by clarifying the dependencies between each business node to be executed, determining the execution priority of each business node to be executed, and associating the output of the previous business node to be executed with the input of the current business node to be executed, the accuracy of the analysis report corresponding to each business node to be executed can be improved. This dependency between the business nodes to be executed also enhances the logic of the target analysis report.

[0089] Figure 7 The illustration shows a schematic diagram of a method for obtaining analysis reports of each business node to be executed based on the execution order according to an embodiment of this application.

[0090] like Figure 7 As shown, the method for obtaining the analysis report of each business node to be executed based on the execution order in this embodiment includes operations S710 to S750.

[0091] When operating S710, during the execution of operations on each business node to be executed, the current retrieval information is obtained based on the parameters to be evaluated and the type of analysis report corresponding to the current business node to be executed.

[0092] When operating S720, the correctness of the dependencies is verified based on the current retrieval information.

[0093] When operating S730, if the dependency relationship is incorrect or the business requirements change, the operation performed by the currently pending business node will be paused and the dependency relationship will be adjusted.

[0094] When operating S740, update the execution priority based on the adjusted dependencies.

[0095] When operating the S750, the analysis reports of each business node to be executed are obtained sequentially according to the updated execution priority.

[0096] For example, the required analysis reports include a macro analysis report for a certain industry, a competitive capability analysis report for a company within that industry, and the company's development plan report. Based on business needs, the dependencies are: first obtain the macro analysis report; then, based on the macro analysis report, obtain the competitive capability analysis report for the company within that industry; and finally, based on the competitive capability analysis report, obtain the company's development rules report. If, during the process of calling the pending business node used to generate the macro analysis report, the retrieved information shows "a significant adjustment in market demand structure or a significant change in industry policies," then it may be necessary to first obtain the company's preliminary development plan report based on this change after obtaining the macro analysis report, then generate the competitive capability analysis report based on the preliminary development plan report, and finally generate the final development plan report based on the competitive capability analysis report. Conversely, if, during the process of calling the pending business node used to generate the competitive capability analysis report, the retrieved information shows "the company is vigorously developing businesses in other industries," but the macro analysis report shows "the industry has a very good development prospect, and the policy support for the industry is strong, with consumer demand for the industry growing rapidly," this indicates that the company's development direction is different from... The direction indicated in the macro analysis report is inconsistent and detrimental to the company's sustainable development. Therefore, the execution priority needs to be adjusted to obtain the development plan report most beneficial to the company's sustainable development. The adjusted execution order should be: first, obtain the macro analysis report; then, based on the macro analysis report, obtain the preliminary development plan report; based on the preliminary development plan report, analyze the company's competitiveness; and finally, obtain the company's final development plan report. If business requirements change during the process of calling the pending business node used to generate the competitiveness analysis report, for example, if the user also wants to analyze the company's collaboration capabilities with its downstream companies, then the parameters to be evaluated may include "analysis of the company's collaboration capabilities with its downstream companies." In this case, based on the new parameters to be evaluated and the current pending business node, retrieve the search information, which includes historical collaboration information between the company and its downstream companies. Although the competitiveness analysis report may include some content related to collaboration capabilities, the content may not be detailed enough. Therefore, the competitiveness analysis report can be input into the next pending business node to generate a collaboration capability analysis report, and then the company's development plan report can be obtained based on the collaboration capability analysis report and the competitiveness analysis report.

[0097] According to the embodiments of this application, during the execution of each business node to be executed, the dependency relationship of each business node to be executed is judged based on the current search information. If there is a problem with the dependency relationship or the business requirements change, the execution priority is adjusted in a timely manner, that is, the execution order of each business node to be executed is adjusted. This dynamic adjustment method further improves the accuracy and logic of the analysis report, thereby improving the quality and accuracy of the target analysis report.

[0098] Based on the above-described method for generating analysis reports, this application also provides an apparatus for generating analysis reports. The following will be combined with... Figure 8 The device is described in detail.

[0099] Figure 8 A schematic block diagram of an analysis report generation apparatus according to an embodiment of this application is shown.

[0100] like Figure 8 As shown, the analysis report generation device 800 of this embodiment includes an information acquisition module 810 to be evaluated, a node analysis report acquisition module 820, and a target analysis report acquisition module 830.

[0101] The evaluation information acquisition module 810 is used to acquire evaluation parameters and multiple execution business nodes based on business needs. The evaluation parameters include at least one of the following: the industry to be evaluated, the name of the company to be evaluated, and the evaluation indicator. Each execution business node is used to perform a type of analysis report generation operation based on the evaluation parameters. In one embodiment, the evaluation information acquisition module 810 can be used to execute the operation S210 described above, which will not be repeated here.

[0102] The node analysis report acquisition module 820 is used to acquire different types of analysis reports corresponding to each business node to be executed based on the parameters to be evaluated and multiple business nodes to be executed, using a large model. In one embodiment, the node analysis report acquisition module 820 can be used to perform the operation S220 described above, which will not be repeated here.

[0103] The target analysis report acquisition module 830 is used to integrate the analysis reports of all business nodes to be executed to obtain a target analysis report. In one embodiment, the target analysis report acquisition module 830 can be used to perform the operation S230 described above, which will not be repeated here.

[0104] According to embodiments of this application, the analysis report generation device can be used to evaluate customer or enterprise information from multiple dimensions, improve report generation efficiency, avoid information omissions and subjectivity issues caused by manual report writing, and thus improve report generation quality.

[0105] In some embodiments, the node analysis report acquisition module 820 is specifically used for: for any business node to be executed, retrieving information from the database according to the parameters to be evaluated and the type of analysis report corresponding to the current business node to be executed, wherein the retrieval information is related to the parameters to be evaluated and the current business node to be executed, and the database includes intermediate data generated during the training of the large model; inputting the retrieval information and its corresponding prompt words into the large model to obtain the analysis report of the current business node to be executed, wherein the prompt words are obtained based on the prompt word template associated with the current business node to be executed.

[0106] In some embodiments, the node analysis report acquisition module 820 is further configured to: acquire preliminary information from the database according to the type of analysis report corresponding to the parameter to be evaluated and the current business node to be executed, and determine the value assessment dimension of the preliminary information, wherein the preliminary information is related to the parameter to be evaluated and the current business node to be executed, and the value assessment dimension includes at least one of the following: the timeliness of the preliminary information, the relevance of the preliminary information to the parameter to be evaluated, and the relevance of the preliminary information to the current business node to be executed; assign weights to the preliminary information based on the value assessment dimension; and use the preliminary information whose weights meet the preset weight threshold as the retrieval information.

[0107] In some embodiments, the node analysis report acquisition module 820 is further configured to: for any business node to be executed, input the retrieved information and the prompt words into the large model to obtain the initial analysis report of the current business node to be executed, wherein each business node to be executed corresponds to a prompt word template, the prompt word template includes system-level prompts, task-level prompts, constraint prompts constructed based on business requirements, and inference step prompts constructed based on associated business nodes to be executed, the system-level prompts are used to constrain the roles and output requirements of the large model, and the task-level prompts are used to constrain the inference tasks of the large model; extract the first metadata information of the initial analysis report, wherein the first metadata information includes element information and source information of the element information; compare with the first metadata information; and extract the first metadata information of the initial analysis report. The system uses metadata and prompts to calculate evaluation metrics for the initial analysis report. These metrics include at least one of the following: completeness of element information, completeness of reasoning steps, and reliability of source information. If any evaluation metric fails to meet its preset threshold, the prompts are adjusted. The adjusted prompts are then input into the large model to re-acquire the initial analysis report. Based on the re-acquired initial analysis report, the calculation of evaluation metrics, prompt adjustment, and re-acquisition of the initial analysis report are iterated until preset conditions are met. These preset conditions include reaching a preset number of iterations or each evaluation metric meeting its respective preset threshold. The initial analysis report that meets the preset conditions is then used as the analysis report for the current pending business node.

[0108] In some embodiments, the node analysis report acquisition module 820 is further configured to: acquire multiple text fragments and secondary metadata information of the multiple text fragments from the retrieved information, wherein the secondary metadata information includes keywords, source information, business rules, and a knowledge graph, the knowledge graph being obtained based on entity extraction from the multiple text fragments, and the business rules including at least one of the following: risk control rules, credit granting rules, and approval rules; parse the prompt words corresponding to the retrieved information to construct the architecture of the analysis report; integrate the multiple text fragments, keywords, source information, and knowledge graph to obtain structured information; match the structured information with the business rules to obtain matching results; and write the structured information and matching results into the corresponding parts of the architecture to obtain an initial analysis report.

[0109] In some embodiments, the analysis report generation device 800 is further configured to: determine the dependency relationships between each business node to be executed according to business requirements, wherein the dependency relationships indicate the connection information between the previous business node to be executed and the current business node to be executed; determine the execution priority of each business node to be executed according to the dependency relationships; and obtain the analysis reports of each business node to be executed sequentially according to the execution priority.

[0110] In some embodiments, the analysis report generation device 800 is further configured to: during the execution of operations at each business node to be executed, obtain current retrieval information based on the parameters to be evaluated and the type of analysis report corresponding to the current business node to be executed; verify the correctness of dependencies based on the current retrieval information; if the dependencies are incorrect or the business requirements change, pause the operations performed by the current business node to be executed and adjust the dependencies; update the execution priority based on the adjusted dependencies; and obtain the analysis reports of each business node to be executed sequentially according to the updated execution priority.

[0111] According to embodiments of this application, any multiple modules among the evaluation information acquisition module 810, node analysis report acquisition module 820, and target analysis report acquisition module 830 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 evaluation information acquisition module 810, node analysis report acquisition module 820, and target analysis report acquisition module 830 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any one of the three implementation methods or a suitable combination of any of them. Alternatively, at least one of the evaluation information acquisition module 810, the node analysis report acquisition module 820, and the target analysis report acquisition module 830 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0112] Figure 9 A block diagram schematically illustrates an electronic device suitable for implementing an analysis report generation method according to an embodiment of this application.

[0113] like Figure 9 As shown, an electronic device 900 according to an embodiment of this application includes a processor 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage portion 908 into a random access memory (RAM) 903. The processor 901 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 901 may also include onboard memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0114] RAM 903 stores various programs and data required for the operation of electronic device 900. Processor 901, ROM 902, and RAM 903 are interconnected via bus 904. Processor 901 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 902 and / or RAM 903. It should be noted that the programs may also be stored in one or more memories other than ROM 902 and RAM 903. Processor 901 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0115] According to embodiments of this application, the electronic device 900 may further include an input / output (I / O) interface 905, which is also connected to a bus 904. The electronic device 900 may also include one or more of the following components connected to the input / output (I / O) interface 905: an input section 906 including a keyboard, mouse, etc.; an output section 907 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the input / output (I / O) interface 905 as needed. A removable medium 911, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 910 as needed so that computer programs read from it can be installed into the storage section 908 as needed.

[0116] This application also provides 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.

[0117] 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 (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In 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 ROM 902 and / or RAM 903 and / or one or more memories other than ROM 902 and RAM 903 described above.

[0118] 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 enables the computer system to implement the analysis report generation method provided in the embodiments of this application.

[0119] When the computer program is executed by the processor 901, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0120] 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 downloaded and installed via the communication section 909, and / or installed from a removable medium 911. 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.

[0121] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, it performs the functions defined in the system of this application embodiment. According to the embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0122] 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. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming 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).

[0123] 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.

[0124] 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 generating analysis reports, characterized in that, The method includes: Based on business needs, parameters to be evaluated and multiple business nodes to be executed are obtained. The parameters to be evaluated include at least one of the following: the industry to be evaluated, the name of the company to be evaluated, and the indicators to be evaluated. Each business node to be executed is used to perform a type of analysis report generation operation based on the parameters to be evaluated. Based on the parameters to be evaluated and the multiple business nodes to be executed, different types of analysis reports corresponding to each business node to be executed are obtained using a large model. The analysis reports of all pending business nodes are integrated to obtain the target analysis report.

2. The method according to claim 1, characterized in that, Based on the parameters to be evaluated and the multiple business nodes to be executed, a large model is used to obtain different types of analysis reports corresponding to each business node to be executed, including: For any business node to be executed, retrieval information is obtained from the database according to the parameters to be evaluated and the type of the analysis report corresponding to the current business node to be executed. The retrieval information is related to the parameters to be evaluated and the current business node to be executed, respectively. The database includes intermediate data generated during the training of the large model. The search information and its corresponding prompt words are input into the large model to obtain the analysis report of the current business node to be executed, wherein the prompt words are obtained based on the prompt word template associated with the current business node to be executed.

3. The method according to claim 2, characterized in that, The step of retrieving information from the database based on the parameters to be evaluated and the type of the analysis report corresponding to the current business node to be executed includes: Based on the parameters to be evaluated and the type of the analysis report corresponding to the current business node to be executed, preliminary information is obtained from the database, and the value assessment dimensions of the preliminary information are determined. The preliminary information is related to the parameters to be evaluated and the current business node to be executed, respectively. The value assessment dimensions include at least one of the following: the timeliness of the preliminary information, the correlation between the preliminary information and the parameters to be evaluated, and the correlation between the preliminary information and the current business node to be executed. Based on the aforementioned value assessment dimensions, weights are assigned to the preliminary information; The preliminary information whose weights satisfy a preset weight threshold is used as the retrieval information.

4. The method according to claim 2, characterized in that, The method further includes: Based on the business requirements, the dependencies between each business node to be executed are determined, wherein the dependencies indicate the connection information between the previous business node to be executed and the current business node to be executed. Based on the dependencies, the execution priority of each business node to be executed is determined; According to the execution priority, the analysis reports of each business node to be executed are obtained sequentially.

5. The method according to claim 4, characterized in that, The step of sequentially obtaining the analysis reports of each pending business node according to the execution priority includes: During the execution of operations at each pending business node, current retrieval information is obtained based on the parameters to be evaluated and the type of the analysis report corresponding to the current pending business node; Based on the current search information, verify the correctness of the dependency relationship; If the dependency relationship is incorrect or the business requirements change, the operation performed by the currently pending business node is paused and the dependency relationship is adjusted. Based on the adjusted dependencies, the execution priority is updated; According to the updated execution priority, the analysis reports of each business node to be executed are obtained sequentially.

6. The method according to claim 2, characterized in that, The step of inputting the retrieval information and its corresponding prompts into the large model to obtain the analysis report of the current business node to be executed includes: For any pending business node, the search information and the prompt words are input into the large model to obtain the initial analysis report of the current pending business node. Each pending business node corresponds to a prompt word template, which includes system-level prompts, task-level prompts, constraint prompts constructed based on the business requirements, and inference step prompts constructed based on associated pending business nodes. The system-level prompts are used to constrain the roles and output requirements of the large model, and the task-level prompts are used to constrain the inference tasks of the large model. Extract the first metadata information from the initial analysis report, wherein the first metadata information includes feature information and the source information of the feature information; By comparing the first metadata information with the prompt words, the evaluation index of the initial analysis report is calculated, wherein the evaluation index includes at least one of the following: the completeness of the element information, the completeness of the reasoning steps, and the reliability of the source information; If the evaluation indicator does not meet its preset threshold, the prompt word will be adjusted. Input the adjusted prompts into the large model and re-obtain the initial analysis report; Based on the reacquired initial analysis report, the calculation of the evaluation index, the adjustment of the prompt words, and the reacquisition of the initial analysis report are iterated until a preset condition is reached. The preset condition includes reaching a preset number of iterations or the current evaluation index meeting its respective preset threshold. The initial analysis report that meets the preset conditions will be used as the analysis report for the current business node to be executed.

7. The method according to claim 6, characterized in that, The initial analysis report was obtained through the following methods: Multiple text fragments and second metadata information of the multiple text fragments are obtained from the search information. The second metadata information includes keywords, source information, business rules and knowledge graph. The knowledge graph is obtained based on entity extraction of the multiple text fragments. The business rules include at least one of the following: risk control rules, credit granting rules and approval rules. Parse the prompts corresponding to the search information to construct the architecture of the analysis report; By integrating the multiple text fragments, the keywords, the source information, and the knowledge graph, structured information is obtained; The structured information is matched with the business rules to obtain a matching result; The structured information and the matching results are written into the corresponding parts of the architecture to obtain the initial analysis report.

8. An analytical report generation device, characterized in that, The device includes: The module for obtaining information to be evaluated is used to obtain parameters to be evaluated and multiple business nodes to be executed based on business needs. The parameters to be evaluated include at least one of the following: the industry to be evaluated, the name of the company to be evaluated, and the indicators to be evaluated. Each business node to be executed is used to perform a type of analysis report generation operation based on the parameters to be evaluated. The node analysis report acquisition module is used to acquire different types of analysis reports corresponding to each business node to be executed based on the parameters to be evaluated and the multiple business nodes to be executed, using a large model. The target analysis report acquisition module is used to integrate the analysis reports of all business nodes to be executed to obtain the target analysis report.

9. 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 7.

10. 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 7.

11. 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 7.

Images