Enterprise exhaustion report intelligent generation method and system based on large model

Abstract

The invention discloses an enterprise exhaustion report intelligent generation method and system based on a large model, and relates to the technical field of artificial intelligence application. The demand understanding module analyzes a complete call demand and a key field in a text input by a user, and presets a template of a complete call report; the intelligent reasoning module constructs execution plans of other modules according to user exhaustion demands and report task characteristics, and formulates report task targets and output standards; the multi-source data scheduling module is in butt joint with a structured database of each administrative department related to an enterprise, and converts a natural language query requirement into an SQL statement capable of directly calling the database; the public opinion analysis module retrieves related news public opinions and performs emotional tendency analysis by using a large model; the intelligent integration module carries out integration according to a preset template of the exhaustion call report to give a comprehensive conclusion; the interactive generation module performs detail adjustment on the exhaustion report based on the real-time feedback of the user to obtain a final exhaustion report; the working efficiency is remarkably improved.

Description

Technical Field

[0001] This invention discloses a method and system for intelligent generation of enterprise due diligence reports based on large models, which relates to the field of artificial intelligence application technology. Background Technology

[0002] Current due diligence methods primarily rely on manual collection of corporate business registration, financial, and legal data, which is then manually organized and analyzed using Excel or Word to generate reports. This approach is inefficient, prone to data omissions or subjective biases, and struggles to track public opinion dynamics in real time. Knowledge graph-based corporate due diligence systems use knowledge graph technology to link corporate data, but report generation relies on preset rules, resulting in insufficient flexibility, an inability to understand natural language requirements, and the need for manual definition of data association logic. This makes them ill-suited for complex and ever-changing due diligence scenarios, and their data processing capabilities are insufficient, exhibiting poor intelligence and flexibility. Furthermore, existing tools only provide data query functions and cannot automatically generate analytical conclusions that meet regulatory requirements. Public opinion risk monitoring and report updates are lagging, and manual revisions are costly. Summary of the Invention

[0003] This invention addresses the problems of existing technologies by providing a method and system for intelligent generation of corporate due diligence reports based on a large-scale model. Through automated data collection, analysis, and report generation, it reduces the existing manual due diligence process from hours or days to minutes, significantly improving work efficiency. This invention is particularly suitable for commercial banks, investment institutions, and other financial-related fields, and can be widely applied to scenarios such as corporate credit approval, investment due diligence, and M&A evaluation, driving the development of financial due diligence processes towards intelligence and automation.

[0004] The specific solution proposed in this invention is as follows:

[0005] This invention provides an intelligent generation system for enterprise due diligence reports based on a large model, including a demand understanding module, an intelligent reasoning module, a multi-source data scheduling module, a public opinion analysis unit, an intelligent integration module, and an interactive generation module;

[0006] Requirements Understanding Module: Based on Natural Language Processing (NLP) methods, this module parses the due diligence requirements and key fields in the user's input text and pre-sets a template for the due diligence report. The due diligence requirements include the name of the company being investigated, the format of the due diligence report, and the content requirements for each section.

[0007] Intelligent reasoning module: Based on user due diligence needs and the characteristics of reporting tasks, it constructs execution plans for other modules and sets reporting task objectives and output standards;

[0008] Multi-source data scheduling module: dynamically connects to the structured databases of various administrative departments related to the enterprise, and converts natural language query requirements into SQL statements that can be directly called from the database based on the key fields extracted by parsing;

[0009] Public opinion analysis module: Using online search tools, relevant news and public opinion are retrieved from hot news platforms and financial websites. Large models are used to conduct sentiment analysis and output public opinion analysis results and risk assessment reports, generating corporate reputation profiles.

[0010] Intelligent integration module: Integrates the structured data and public opinion analysis results according to the preset template of the due diligence report; automatically fills in the enterprise information overview, financial status analysis, risk assessment and public opinion analysis in the due diligence report, and gives a comprehensive conclusion;

[0011] Interactive generation module: Based on real-time user feedback, the due diligence report can be fine-tuned and revised through an interactive interface to obtain the final due diligence report. Version control of the due diligence report is also performed, and the due diligence report can be shared according to user needs.

[0012] Furthermore, the requirement understanding module of the intelligent enterprise due diligence report generation system based on a large model parses the due diligence requirements and key fields in the user input text using Natural Language Processing (NLP) methods, including:

[0013] The requirements understanding module uses a domain-adapted pre-trained language model to perform deep semantic encoding on the user input text, and then constructs a task instruction tree to structurally represent the due diligence requirements. It uses a hybrid neural network to process the input text. During text preprocessing, it standardizes financial terms based on regular expressions and uses SentencePiece for sub-word segmentation to identify financial terms.

[0014] Furthermore, the intelligent reasoning module of the enterprise due diligence report intelligent generation system based on a large model is based on a multi-agent framework and, in conjunction with a preset due diligence report template, designs the due diligence report generation process.

[0015] The due diligence report generation process involves optimizing the scheduling and execution strategies of other modules. Among these, the execution plan for each task is dynamically generated based on a multi-agent framework. Based on the task dependencies and priorities, the execution order of tasks is reasonably arranged. The dependencies between tasks are modeled as a directed acyclic graph, where nodes represent tasks and edges represent the dependencies between tasks. A topological sorting algorithm is used to schedule the tasks.

[0016] Furthermore, the multi-source data scheduling module of the intelligent enterprise due diligence report generation system based on a large model uses NL2SQL to transform natural language query requirements into SQL query statements, including:

[0017] Obtain natural language query requirements.

[0018] We understand the query target through semantic analysis of large models, and extract the target fields and filtering conditions.

[0019] Perform SQL template matching: Select the appropriate SQL template based on the natural language query requirements;

[0020] Generate SQL query statements: populate placeholders in the SQL template, generate complete SQL query statements, retrieve the generated SQL query statements, and ensure the correctness of the statement syntax.

[0021] This invention also provides a method for intelligently generating enterprise due diligence reports based on a large model. The method utilizes an intelligent enterprise due diligence report generation system to generate enterprise due diligence reports. The system includes a requirements understanding module, an intelligent reasoning module, a multi-source data scheduling module, a public opinion analysis unit, an intelligent integration module, and an interactive generation module.

[0022] The requirements understanding module uses Natural Language Processing (NLP) methods to parse the due diligence requirements and key fields in the user's input text and preset the template for the due diligence report. The due diligence requirements include the name of the due diligence company, the format of the due diligence report, and the content requirements of each section.

[0023] Based on user due diligence needs and reporting task characteristics, the intelligent reasoning module constructs execution plans for other modules and sets reporting task objectives and output standards.

[0024] The multi-source data scheduling module dynamically connects to the structured databases of various administrative departments related to the enterprise, and transforms natural language query requirements into SQL statements that can be directly called from the database based on the key fields extracted by parsing.

[0025] The public opinion analysis module uses online search tools to retrieve relevant news and public opinion on trending news platforms and financial websites. It uses a large model to conduct sentiment analysis and outputs public opinion analysis results and risk assessment reports to generate corporate reputation profiles.

[0026] The intelligent integration module integrates the structured data and public opinion analysis results according to the preset template of the due diligence report; it automatically fills in the enterprise information overview, financial status analysis, risk assessment and public opinion analysis in the due diligence report, and gives a comprehensive conclusion.

[0027] The interactive generation module allows users to make detailed adjustments to the due diligence report based on real-time feedback through the interactive interface, revise the due diligence report, obtain the final due diligence report, and perform version control of the due diligence report. At the same time, the due diligence report can be shared according to user needs.

[0028] Furthermore, the intelligent generation method for enterprise due diligence reports based on a large model uses a requirements understanding module to parse due diligence requirements and key fields in user input text using Natural Language Processing (NLP) methods, including:

[0029] The requirements understanding module uses a domain-adapted pre-trained language model to perform deep semantic encoding on the user input text, and then constructs a task instruction tree to structurally represent the due diligence requirements. It uses a hybrid neural network to process the input text. During text preprocessing, it standardizes financial terms based on regular expressions and uses SentencePiece for sub-word segmentation to identify financial terms.

[0030] Furthermore, the intelligent generation method for enterprise due diligence reports based on a large model utilizes an intelligent reasoning module based on a multi-agent framework, combined with a pre-set due diligence report template, to design the due diligence report generation process.

[0031] The due diligence report generation process involves optimizing the scheduling and execution strategies of other modules. Among these, the execution plan for each task is dynamically generated based on a multi-agent framework. Based on the task dependencies and priorities, the execution order of tasks is reasonably arranged. The dependencies between tasks are modeled as a directed acyclic graph, where nodes represent tasks and edges represent the dependencies between tasks. A topological sorting algorithm is used to schedule the tasks.

[0032] Furthermore, the intelligent generation method for enterprise due diligence reports based on a large model uses NL2SQL through a multi-source data scheduling module to convert natural language query requirements into SQL query statements, including:

[0033] Obtain natural language query requirements.

[0034] We understand the query target through semantic analysis of large models, and extract the target fields and filtering conditions.

[0035] Perform SQL template matching: Select the appropriate SQL template based on the natural language query requirements;

[0036] Generate SQL query statements: populate placeholders in the SQL template, generate complete SQL query statements, retrieve the generated SQL query statements, and ensure the correctness of the statement syntax.

[0037] The advantages of this invention are:

[0038] This invention achieves automation and intelligence in the due diligence process by integrating large models, data interfaces, and online retrieval tools. The system of this invention can automatically acquire and integrate data, improving the efficiency of data analysis and report generation.

[0039] This invention automatically converts natural language query requirements into efficient SQL statements by calling SQL data, accurately calling key database fields, reducing the workload of manually writing query statements, and improving the accuracy and speed of data retrieval;

[0040] This invention integrates online retrieval tools to retrieve and conduct due diligence on relevant news and public opinion from multiple data sources such as news platforms and social media in real time. It also uses the semantic understanding capabilities of large models to analyze public opinion sentiment and obtain analysis results on public opinion risks, thereby enhancing the risk assessment function of the due diligence report.

[0041] This invention provides an interactive modification function between the user and the system. After the report is generated, the user can input modification requests in natural language, and the system will respond and adjust the report content. This avoids problems such as low interpretability caused by large model generation, improves the accuracy of due diligence reports, and meets the financial industry's demand for high accuracy. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the system application process of the present invention. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0044] Example 1

[0045] This invention provides an intelligent generation system for enterprise due diligence reports based on a large model, including a demand understanding module, an intelligent reasoning module, a multi-source data scheduling module, a public opinion analysis unit, an intelligent integration module, and an interactive generation module;

[0046] Requirements Understanding Module: Based on Natural Language Processing (NLP) methods, this module parses the due diligence requirements and key fields in the user's input text and pre-sets a template for the due diligence report. The due diligence requirements include the name of the company being investigated, the format of the due diligence report, and the content requirements for each section.

[0047] The requirements understanding module can employ a multi-layered NLP architecture for intelligent processing during the due diligence requirements acquisition process. For example, it first performs deep semantic encoding on the input text using a domain-adapted pre-trained language model, then constructs a task instruction tree to structurally represent the due diligence requirements. A hybrid neural network architecture is used to process the input text. In the text preprocessing layer, a financial terminology standardization method based on regular expressions is used, such as "Q2" → "Second Quarter," and SentencePiece is used for sub-word segmentation to solve the problem of professional terminology recognition. In the multi-task learning layer, a dependency tree can be constructed based on the Eisner algorithm, and a rule engine can be applied to transform business logic relationships.

[0048] Intelligent Reasoning Module: Based on the user's due diligence needs and the characteristics of the reporting task, it constructs the execution plan for other modules and sets the reporting task objectives and output standards.

[0049] The intelligent reasoning module is based on a multi-agent framework and, in conjunction with a pre-set due diligence report template, designs a due diligence report generation process.

[0050] The due diligence report generation process involves optimizing the scheduling and execution strategies of other modules. Among these, the execution plan for each task is dynamically generated based on a multi-agent framework. Based on the task dependencies and priorities, the execution order of tasks is reasonably arranged. The dependencies between tasks are modeled as a directed acyclic graph, where nodes represent tasks and edges represent the dependencies between tasks. A topological sorting algorithm is used to schedule the tasks.

[0051] In real-world environments, intelligent reasoning modules typically work in conjunction with deep learning servers, cloud computing platforms, or computing clusters, significantly improving task planning efficiency through parallel computing technology. For example, an intelligent reasoning module automatically generates a task execution plan based on user-inputted due diligence requirements. Its workflow is as follows: First, it receives the user's due diligence requirements; then, it designs a complete task execution process by combining pre-defined laws, regulations, industry standards, and due diligence specifications. The generated execution plan includes multiple tasks such as data collection, public opinion analysis, and case information extraction, with clear task objectives and execution order set for each task. To ensure the rationality of the execution plan, the module needs to accurately determine the dependencies between tasks and scientifically arrange the execution order. By employing an intelligent agent framework and advanced reasoning algorithms, the module can dynamically adjust task execution strategies and continuously optimize each step in the report generation process, thereby achieving the final goal more efficiently. This dynamic optimization mechanism allows the system to flexibly adjust the execution path according to actual conditions, significantly improving overall work efficiency.

[0052] Multi-source data scheduling module: dynamically connects to the structured databases of various administrative departments related to the enterprise, and transforms natural language query requirements into SQL statements that can be directly invoked from the database based on the key fields extracted by parsing.

[0053] The multi-source data scheduling module uses NL2SQL to convert natural language query requirements into SQL query statements, including:

[0054] Obtain natural language query requirements.

[0055] We understand the query target through semantic analysis of large models, and extract the target fields and filtering conditions.

[0056] Perform SQL template matching: Select the appropriate SQL template based on the natural language query requirements;

[0057] Generate SQL query statements: populate placeholders in the SQL template, generate complete SQL query statements, retrieve the generated SQL query statements, and ensure the correctness of the statement syntax.

[0058] Public opinion analysis module: Using online search tools, relevant news and public opinion are retrieved from trending news platforms and financial websites. Large-scale models are used to conduct sentiment analysis, and the results of the public opinion analysis and risk assessment report are output to generate a corporate reputation profile.

[0059] The public opinion analysis module identifies the sentiment trends (positive, negative, and neutral) of the retrieved public opinion information related to the due diligence target, and scores the risk based on click-through rates and comment volume. It then outputs a public opinion summary and risk assessment report, clearly defining risk information related to the company's reputation and market behavior. It can also build a sentiment dictionary adapted to the financial industry, including positive, negative, and neutral terms. Deep learning models LSTM or BERT are used to classify the public opinion content.

[0060] The intelligent integration module integrates the structured data and public opinion analysis results according to the preset template of the due diligence report; it automatically fills in the enterprise information overview, financial status analysis, risk assessment, and public opinion analysis in the due diligence report, and provides a comprehensive conclusion. During data integration and report generation, automatic formatting ensures the report meets formatting requirements, automatically adjusting the format of the report content, including chapters, tables, and paragraphs. It identifies the chapter structure of the report and determines the layout of each part. It automatically adjusts font, line spacing, table format, etc., according to the report template.

[0061] Interactive generation module: Based on real-time user feedback, the due diligence report can be fine-tuned and revised through an interactive interface to obtain the final due diligence report. Version control of the due diligence report is also implemented. The due diligence report can be shared according to user needs, such as the final report can be output as a Word document conforming to industry standards, supporting export, sharing, and printing.

[0062] Example 2

[0063] This invention also provides a method for intelligently generating enterprise due diligence reports based on a large model. The method utilizes an intelligent enterprise due diligence report generation system to generate enterprise due diligence reports. The system includes a requirements understanding module, an intelligent reasoning module, a multi-source data scheduling module, a public opinion analysis unit, an intelligent integration module, and an interactive generation module.

[0064] The requirements understanding module uses Natural Language Processing (NLP) methods to parse the due diligence requirements and key fields in the user's input text and preset the template for the due diligence report. The due diligence requirements include the name of the due diligence company, the format of the due diligence report, and the content requirements of each section.

[0065] Based on user due diligence needs and reporting task characteristics, the intelligent reasoning module constructs execution plans for other modules and sets reporting task objectives and output standards.

[0066] The multi-source data scheduling module dynamically connects to the structured databases of various administrative departments related to the enterprise, and transforms natural language query requirements into SQL statements that can be directly called from the database based on the key fields extracted by parsing.

[0067] The public opinion analysis module uses online search tools to retrieve relevant news and public opinion on trending news platforms and financial websites. It uses a large model to conduct sentiment analysis and outputs public opinion analysis results and risk assessment reports to generate corporate reputation profiles.

[0068] The intelligent integration module integrates the structured data and public opinion analysis results according to the preset template of the due diligence report; it automatically fills in the enterprise information overview, financial status analysis, risk assessment and public opinion analysis in the due diligence report, and gives a comprehensive conclusion.

[0069] The interactive generation module allows users to make detailed adjustments to the due diligence report based on real-time feedback through the interactive interface, revise the due diligence report, obtain the final due diligence report, and perform version control of the due diligence report. At the same time, the due diligence report can be shared according to user needs.

[0070] The above method utilizes the information interaction and execution process between the modules within the system of the present invention to solve the technical problem of the present invention. It is based on the same concept as the system embodiment of the present invention, and the specific details can be found in the description of the system embodiment of the present invention, which will not be repeated here.

[0071] Similarly, the advantages of the method of the present invention are:

[0072] By integrating large models, data interfaces, and online retrieval tools, this invention automates and intelligentizes the due diligence process, enabling automatic data acquisition and integration, and improving the efficiency of data analysis and report generation.

[0073] By using the system of this invention to call SQL data, natural language query requirements are automatically converted into efficient SQL statements, and key fields of the database are accurately called, reducing the workload of manually writing query statements and improving the accuracy and speed of data retrieval;

[0074] By integrating online retrieval tools through the system of this invention, it is possible to retrieve and conduct due diligence on relevant news and public opinion from multiple data sources such as news platforms and social media in real time, and use the semantic understanding capabilities of large models to conduct sentiment analysis of public opinion, thereby obtaining analysis results on public opinion risks and enhancing the risk assessment function of due diligence reports.

[0075] The system allows users to interact with it and make modifications. After the report is generated, users can input their modification requests in natural language, and the system will respond and adjust the report content. This avoids problems such as low interpretability caused by large model generation, improves the accuracy of due diligence reports, and meets the financial industry's demand for high accuracy.

[0076] It should be noted that not all steps and modules in the above processes and system structures are mandatory; some steps or modules can be omitted as needed. The execution order of the steps is not fixed and can be adjusted as required. The system structures described in the above embodiments can be physical or logical structures. That is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities, or they may be implemented by certain components in multiple independent devices.

[0077] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. An intelligent enterprise due diligence report generation system based on a large model, characterized by: It includes a demand understanding module, an intelligent reasoning module, a multi-source data scheduling module, a public opinion analysis unit, an intelligent integration module, and an interactive generation module; Requirements Understanding Module: Based on Natural Language Processing (NLP) methods, this module parses the due diligence requirements and key fields in the user's input text and pre-sets a template for the due diligence report. The due diligence requirements include the name of the company being investigated, the format of the due diligence report, and the content requirements for each section. Intelligent reasoning module: Based on user due diligence needs and the characteristics of reporting tasks, it constructs execution plans for other modules and sets reporting task objectives and output standards; Multi-source data scheduling module: dynamically connects to the structured databases of various administrative departments related to the enterprise, and converts natural language query requirements into SQL statements that can be directly called from the database based on the key fields extracted by parsing; Public opinion analysis module: Using online search tools, relevant news and public opinion are retrieved from hot news platforms and financial websites. Large models are used to conduct sentiment analysis and output public opinion analysis results and risk assessment reports, generating corporate reputation profiles. Intelligent integration module: Integrates the structured data and public opinion analysis results according to the preset template of the due diligence report; automatically fills in the enterprise information overview, financial status analysis, risk assessment and public opinion analysis in the due diligence report, and gives a comprehensive conclusion; Interactive generation module: Based on real-time user feedback, the due diligence report can be fine-tuned and revised through an interactive interface to obtain the final due diligence report. Version control of the due diligence report is also performed, and the due diligence report can be shared according to user needs.

2. The intelligent generation system for enterprise due diligence reports based on a large model as described in claim 1, characterized in that the requirements... The understanding module uses Natural Language Processing (NLP) methods to parse due diligence requirements and key fields from user input text, including: The requirements understanding module uses a domain-adapted pre-trained language model to perform deep semantic encoding on the user input text, and then constructs a task instruction tree to structurally represent the due diligence requirements. It uses a hybrid neural network to process the input text. During text preprocessing, it standardizes financial terms based on regular expressions and uses SentencePiece for sub-word segmentation to identify financial terms.

3. The intelligent generation system for enterprise due diligence reports based on a large model as described in claim 1, characterized in that: The intelligent reasoning module, based on a multi-agent framework and incorporating pre-set due diligence report templates, designs a due diligence report generation process. The due diligence report generation process involves optimizing the scheduling and execution strategies of other modules. Among these, the execution plan for each task is dynamically generated based on a multi-agent framework. Based on the task dependencies and priorities, the execution order of tasks is reasonably arranged. The dependencies between tasks are modeled as a directed acyclic graph, where nodes represent tasks and edges represent the dependencies between tasks. A topological sorting algorithm is used to schedule the tasks.

4. The intelligent generation system for enterprise due diligence reports based on a large model as described in claim 1, characterized by multi-source... The data scheduling module uses NL2SQL to translate natural language query requirements into SQL query statements, including: Obtain natural language query requirements. We understand the query target through semantic analysis of large models, and extract the target fields and filtering conditions. Perform SQL template matching: Select the appropriate SQL template based on the natural language query requirements; Generate SQL query statements: populate placeholders in the SQL template, generate complete SQL query statements, retrieve the generated SQL query statements, and ensure the correctness of the statement syntax.

5. A method for intelligently generating enterprise due diligence reports based on a large model, characterized by: The system utilizes an intelligent enterprise due diligence report generation system to generate enterprise due diligence reports. This system includes a requirements understanding module, an intelligent reasoning module, a multi-source data scheduling module, a public opinion analysis unit, an intelligent integration module, and an interactive generation module. The requirements understanding module uses Natural Language Processing (NLP) methods to parse the due diligence requirements and key fields in the user's input text and preset the template for the due diligence report. The due diligence requirements include the name of the due diligence company, the format of the due diligence report, and the content requirements of each section. Based on user due diligence needs and reporting task characteristics, the intelligent reasoning module constructs execution plans for other modules and sets reporting task objectives and output standards. The multi-source data scheduling module dynamically connects to the structured databases of various administrative departments related to the enterprise, and transforms natural language query requirements into SQL statements that can be directly called from the database based on the key fields extracted by parsing. The public opinion analysis module uses online search tools to retrieve relevant news and public opinion on trending news platforms and financial websites. It uses a large model to conduct sentiment analysis and outputs public opinion analysis results and risk assessment reports to generate corporate reputation profiles. The intelligent integration module integrates the structured data and public opinion analysis results according to the preset template of the due diligence report; it automatically fills in the enterprise information overview, financial status analysis, risk assessment and public opinion analysis in the due diligence report, and gives a comprehensive conclusion. The interactive generation module allows users to make detailed adjustments to the due diligence report based on real-time feedback through the interactive interface, revise the due diligence report, obtain the final due diligence report, and perform version control of the due diligence report. At the same time, the due diligence report can be shared according to user needs.

6. The intelligent generation method for enterprise due diligence reports based on a large model according to claim 5, characterized in that: The requirements understanding module uses Natural Language Processing (NLP) methods to parse user input text, identifying due diligence requirements and key fields, including: The requirements understanding module uses a domain-adapted pre-trained language model to perform deep semantic encoding on the user input text, and then constructs a task instruction tree to structurally represent the due diligence requirements. It uses a hybrid neural network to process the input text. During text preprocessing, it standardizes financial terms based on regular expressions and uses SentencePiece for sub-word segmentation to identify financial terms.

7. The intelligent generation method for enterprise due diligence reports based on a large model according to claim 5, characterized in that: Based on a multi-agent framework and using a pre-defined due diligence report template, an intelligent reasoning module is designed to generate due diligence reports. The due diligence report generation process involves optimizing the scheduling and execution strategies of other modules. Among these, the execution plan for each task is dynamically generated based on a multi-agent framework. Based on the task dependencies and priorities, the execution order of tasks is reasonably arranged. The dependencies between tasks are modeled as a directed acyclic graph, where nodes represent tasks and edges represent the dependencies between tasks. A topological sorting algorithm is used to schedule the tasks.

8. The intelligent generation method for enterprise due diligence reports based on a large model according to claim 5, characterized in that: The NL2SQL module, through its multi-source data scheduling module, transforms natural language query requirements into SQL query statements, including: Obtain natural language query requirements. We understand the query target through semantic analysis of large models, and extract the target fields and filtering conditions. Perform SQL template matching: Select the appropriate SQL template based on the natural language query requirements; Generate SQL query statements: populate placeholders in the SQL template, generate complete SQL query statements, retrieve the generated SQL query statements, and ensure the correctness of the statement syntax.

Images