Data analysis method and system based on digital enterprise management

Abstract

This application relates to the fields of digitalization, big data and artificial intelligence technology, and in particular provides a data analysis method and system based on digital enterprise management. This application combines artificial intelligence and natural language processing technology, and provides enterprises with an efficient, accurate and comprehensive digital enterprise management report analysis method through efficient semantic analysis, precise key semantic optimization, comprehensive operational risk analysis and intuitive risk labeling, which helps enterprises improve operational efficiency and reduce operational risks. In addition, this digital enterprise management technology can be applied to e-commerce, remote office, smart city, digital twin and other related enterprise operation scenarios.

Description

Technical Field

[0001] The present application relates to the fields of digitalization, big data and artificial intelligence technology, and in particular to a data analysis method and system based on digital enterprise management. Background Art

[0002] In the digital age, big data, artificial intelligence and cloud computing technologies have become important tools for enterprise management, each of which has unique application value.

[0003] Big data technology can collect, store and process massive amounts of data, including structured data and unstructured data such as text, images and audio. In digital enterprise management, big data technology can help enterprises understand market dynamics, customer needs and internal operations more comprehensively. Through in-depth analysis of these data, enterprises can discover the laws and trends hidden in the data and provide a scientific basis for decision-making.

[0004] Artificial intelligence technology, especially machine learning and deep learning algorithms, can realize automated prediction, classification and optimization based on big data. In enterprise management, artificial intelligence technology can be applied to customer classification, sales forecasting, risk management and many other aspects. By training models to identify patterns in data, artificial intelligence can help enterprises more accurately predict market trends, optimize resource allocation and improve operational efficiency.

[0005] Cloud computing technology provides a strong infrastructure support for the application of big data and artificial intelligence. The cloud computing platform is elastic and scalable, and can provide flexible computing resources and storage space according to the actual needs of enterprises. Through cloud computing, enterprises can access and process data anytime and anywhere, and realize real-time information sharing and collaborative work. In addition, cloud computing can also reduce the IT costs of enterprises and improve data security.

[0006] However, with the advancement of enterprise digitalization, digital enterprise management reports have become an important basis for enterprise decision-making. These reports usually contain a large amount of text information, and this information is scattered in multiple text blocks. Although big data, artificial intelligence and cloud computing technologies have been widely used in enterprise management, there are still some challenges in processing these complex digital enterprise management reports.

[0007] For example, the large amount of text information in these reports makes information extraction and analysis complicated and time-consuming. Traditional report analysis methods mainly rely on manual reading and summarization, which is not only inefficient but also prone to errors when processing large amounts of data. Therefore, although enterprises already have strong technical support, it still cannot meet the needs of modern enterprises for fast and accurate decision-making. Summary of the invention

[0008] In order to improve the above problems, the present application provides a data analysis method and system based on digital enterprise management.

[0009] The present application embodiment provides a data analysis method based on digital enterprise management, which is applied to a data analysis system. The method includes:

[0010] Obtaining a target digital enterprise management report to be analyzed, wherein the target digital enterprise management report includes at least two management report text blocks;

[0011] Performing semantic analysis on the target digital enterprise management report to obtain the text block description semantics of each management report text block in the target digital enterprise management report;

[0012] Performing a key semantic optimization operation based on the text block description semantics of each management report text block in the target digital enterprise management report to obtain a key semantic optimization vector of the target digital enterprise management report, wherein the key semantic optimization vector is used to reflect the value feature-related data between each management report text block in the target digital enterprise management report;

[0013] Based on the text block description semantics of each management report text block in the target digital enterprise management report and the value feature-involved data represented by the key semantic optimization vector of the target digital enterprise management report, an enterprise operational risk decision analysis is performed on the target digital enterprise management report to obtain an enterprise operational risk decision analysis viewpoint of the target digital enterprise management report. The enterprise operational risk decision analysis viewpoint is used to represent the potential operational risk labels of each management report text block in the target digital enterprise management report.

[0014] Preferably, the key semantic optimization vector of the target digital enterprise management report is obtained based on the target semantic parsing network; the step of obtaining the target semantic parsing network includes:

[0015] Obtaining a neural network learning sample; wherein the neural network learning sample includes a text block description semantics of a target management report text block in a first digital enterprise management report, a text block description semantics of a target management report text block in a second digital enterprise management report, a text block description semantics of a certified management report text block in the second digital enterprise management report, and a priori annotations of the second digital enterprise management report, wherein the priori annotations are used to characterize whether there is an involvement state between the certified management report text block and the target management report text block;

[0016] Using an initial semantic parsing network, a key semantic optimization operation is performed based on the text block description semantics of the target management report text block in the first digital enterprise management report, the text block description semantics of the target management report text block in the second digital enterprise management report, and the text block description semantics of the authenticated management report text block in the second digital enterprise management report, to obtain a key semantic optimization vector of the second digital enterprise management report, wherein the key semantic optimization vector of the second digital enterprise management report is used to characterize whether there is an involvement state between the target management report text block and the authenticated management report text block in the second digital enterprise management report;

[0017] Based on the condition of minimizing the difference between the prior annotations of the second digital enterprise management report and the content corresponding to the key semantic optimization vector of the second digital enterprise management report, the initial semantic parsing network is debugged by a neural network to obtain the target semantic parsing network.

[0018] Preferably, the initial semantic parsing network is used to perform a key semantic optimization operation based on the text block description semantics of the target management report text block in the first digital enterprise management report, the text block description semantics of the target management report text block in the second digital enterprise management report, and the text block description semantics of the authenticated management report text block in the second digital enterprise management report to obtain a key semantic optimization vector of the second digital enterprise management report, including:

[0019] Performing text semantic integration on the text block description semantics of the target management report text block in the first digital enterprise management report and the text block description semantics of each management report text block in the second digital enterprise management report to obtain the text block integrated semantics of the second digital enterprise management report;

[0020] A key semantic optimization vector of the second digital enterprise management report is generated according to the text block integration semantics and the text block description semantics of the target management report text block in the second digital enterprise management report.

[0021] Preferably, the method further comprises:

[0022] Acquire the first digital enterprise management report, wherein the potential operational risk tag of the target management report text block in the first digital enterprise management report is a target management risk tag;

[0023] Determine an authenticated management risk tag that has an involvement status with the target management risk tag, and obtain a management report text block under the authenticated management risk tag to obtain the authenticated management report text block;

[0024] The second digital enterprise management report is outputted based on the first digital enterprise management report and the authenticated management report text block.

[0025] Preferably, the second digital enterprise management report is generated based on a target NLP text report output network, and the step of obtaining the target NLP text report output network includes:

[0026] Using an initial NLP text report output network, determining a common management risk label that is involved with the target management risk label, and obtaining a management report text block under the common management risk label to obtain a common management report text block;

[0027] Performing digital enterprise management report output processing based on the first digital enterprise management report and the common management report text block to obtain an NLP enterprise management output report;

[0028] Acquire a certified digital enterprise management report, wherein the certified digital enterprise management report includes the target management report text block and management report text blocks that are involved with the target management report text block;

[0029] Based on the condition of maximizing the commonality value between the NLP enterprise management output report and the authenticated digital enterprise management report, the initial NLP text report output network is debugged by a neural network to obtain the target NLP text report output network.

[0030] Preferably, the neural network debugging of the initial NLP text report output network based on the condition of maximizing the commonality value between the NLP enterprise management output report and the authenticated digital enterprise management report to obtain the target NLP text report output network includes:

[0031] Acquire a priori annotations of the word and sentence units of the certified digital enterprise management report, wherein the a priori annotations of the word and sentence units are used to characterize potential operational risk labels corresponding to management report text blocks to which each certified word and sentence unit in the certified digital enterprise management report belongs;

[0032] Determine a potential operational risk label corresponding to a management report text block to which each word unit in the NLP enterprise management output report belongs, wherein the word unit in the NLP enterprise management output report has a one-to-one matching relationship with the authenticated word unit in the authenticated digital enterprise management report;

[0033] Based on the condition of maximizing the commonality value between the potential operational risk label corresponding to the management report text block to which each word and sentence unit in the NLP enterprise management output report belongs and the potential operational risk label corresponding to the management report text block to which the corresponding authenticated word and sentence unit belongs, the initial NLP text report output network is subjected to neural network debugging to obtain the target NLP text report output network.

[0034] Preferably, the method further comprises:

[0035] Acquire the first digital enterprise management report and the certified NLP enterprise management output report, wherein the potential operational risk tag of the target management report text block in the first digital enterprise management report is a target management risk tag, and the certified NLP enterprise management output report includes the target management report text block and a management report text block having an involvement status with the target management report text block;

[0036] Determining, based on the text block description semantics of the target management report text block in the first digital enterprise management report, an authenticated management risk tag having an involvement status with the target management risk tag;

[0037] Obtaining the management report text block under the authenticated management risk tag, obtaining the authenticated management report text block, and outputting the second digital enterprise management report based on the first digital enterprise management report and the authenticated management report text block;

[0038] The neural network debugging of the initial semantic parsing network based on the condition of minimizing the difference between the prior annotation of the second digital enterprise management report and the content corresponding to the key semantic optimization vector of the second digital enterprise management report to obtain the target semantic parsing network includes:

[0039] Based on the conditions of minimizing the difference between the prior annotations of the second digital enterprise management report and the contents corresponding to the key semantic optimization vectors of the second digital enterprise management report, and maximizing the commonality value between the second digital enterprise management report and the certified NLP enterprise management output report, the initial semantic parsing network is neural network debugged to obtain the target semantic parsing network.

[0040] Preferably, the method further comprises:

[0041] Acquire a target report semantic relationship network, wherein the target report semantic relationship network is used to represent a semantic delivery chain corresponding to the target digital enterprise management report;

[0042] Implementing a first feature embedding process of a logical description level on the target report semantic relationship network to obtain a first report semantic relationship network of a first logical description level, wherein the first logical description level is smaller than a basic logical description level of the target report semantic relationship network;

[0043] Implementing a second feature embedding process of the sentiment description level on the first report semantic relationship network to obtain a second report semantic relationship network, wherein the first sentiment description level of the second report semantic relationship network is greater than the original sentiment description level of the target report semantic relationship network, and the sentiment feature recognition degree of the second report semantic relationship network is greater than the original feature recognition degree of the target report semantic relationship network;

[0044] Implementing the third feature embedding process of the sentiment description level on the second report semantic relationship network to obtain a third report semantic relationship network, wherein the second sentiment description level of the third report semantic relationship network is consistent with the original sentiment description level, and the sentiment feature recognition degree of the third report semantic relationship network is consistent with the sentiment feature recognition degree of the second report semantic relationship network;

[0045] Implementing a fourth feature embedding process of a logical description level on the third report semantic relationship network to obtain a fourth report semantic relationship network of a second logical description level, wherein the second logical description level is consistent with the basic logical description level;

[0046] The fourth report semantic relationship network is used to construct a structured text tree for the target digital enterprise management report.

[0047] Preferably, the second feature embedding processing of the sentiment description level is performed on the first report semantic relationship network to obtain the second report semantic relationship network, including: performing the first feature embedding processing of the semantic depth level and the semantic coverage level on the first report semantic relationship network to obtain a first embedded semantic relationship network, wherein the semantic depth level of the first embedded semantic relationship network is greater than the original semantic depth level of the target report semantic relationship network, the semantic coverage level of the first embedded semantic relationship network is greater than the original semantic coverage level of the target report semantic relationship network, and the sentiment feature recognition degree of the first embedded semantic relationship network is greater than the original feature recognition degree; based on the first embedded semantic relationship network, the second report semantic relationship network is obtained;

[0048] Then the third feature embedding processing of the sentiment description level is implemented on the second report semantic relationship network to obtain the third report semantic relationship network, including: implementing the second feature embedding processing of the semantic depth level and the semantic coverage level on the second report semantic relationship network to obtain a second embedded semantic relationship network, wherein the semantic depth level of the second embedded semantic relationship network is consistent with the original semantic depth level, the semantic coverage level of the second embedded semantic relationship network is consistent with the original semantic coverage level, and the sentiment feature recognition degree of the second embedded semantic relationship network is consistent with the sentiment feature recognition degree of the first embedded semantic relationship network; based on the second embedded semantic relationship network, the third report semantic relationship network is obtained.

[0049] An embodiment of the present application provides a data analysis system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the above method.

[0050] An embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. The computer program implements the above method when running.

[0051] In recent years, semantic parsing technology has been widely used. Semantic parsing can automatically extract key information from text and convert it into a computer-understandable format, which is convenient for subsequent data analysis and decision support. However, existing semantic parsing methods still have some challenges when processing digital enterprise management reports.

[0052] First, digital enterprise management reports usually contain multiple management report text blocks, which may have complex associations and dependencies. Simply performing independent semantic analysis on each text block often fails to accurately capture these relationships, thus affecting the accuracy of the analysis.

[0053] Secondly, traditional semantic parsing methods often have problems with information redundancy and loss of key information when processing large amounts of text data. This is mainly because these methods do not take into account the value characteristics of the data involved between text blocks, that is, the actual importance and mutual influence of each text block in corporate operations.

[0054] Therefore, in order to more effectively parse and process digital enterprise management reports and improve the accuracy and efficiency of decision-making, the embodiment of the present application proposes a new method. The method first obtains a target digital enterprise management report containing at least two management report text blocks, and then extracts the descriptive semantics of each text block through semantic parsing technology. Then, based on these descriptive semantics, a key semantic optimization operation is performed to obtain a key semantic optimization vector, which can reflect the value feature-related data between each text block in the report. Finally, using this vector and the descriptive semantics of the text block, the report is analyzed for enterprise operational risk decisions, potential operational risks are quickly and accurately identified, and presented in the form of intuitive risk labels. The above method not only improves the efficiency and accuracy of report analysis, but also provides strong data support for enterprise decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A flowchart of a data analysis method based on digital enterprise management provided in an embodiment of the present application.

[0056] Figure 2 A structural diagram of a data analysis system 200 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to better understand the above technical scheme, the technical scheme of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0058] Figure 1 A data analysis method based on digital enterprise management is shown, which is applied to a data analysis system. The method includes the following steps 110 to 140.

[0059] Step 110: Obtain a target digital enterprise management report to be analyzed, wherein the target digital enterprise management report includes at least two management report text blocks.

[0060] Among them, the target digital enterprise management report to be analyzed is a comprehensive, digital enterprise management document that records in detail the information and data of the enterprise in various operational aspects. This report can be generated regularly to reflect the operating status of the enterprise in a certain period of time. In this application scenario, the report contains content from multiple business areas such as financial status, sales performance, supply chain management, human resources and market analysis. Enterprise decision makers or management will rely on this report to understand the overall operation of the enterprise and make decisions. For example, through the financial data in the report, decision makers can understand the profitability and cost control of the enterprise; through sales performance data, they can judge market trends and customer needs.

[0061] Furthermore, management report text blocks are independent parts of digital enterprise management reports, and each text block focuses on describing a specific aspect of enterprise operations, such as finance, sales, supply chain, etc. These text blocks usually contain detailed data, analysis, and insights, aiming to help readers deeply understand the status quo and trends of this business area. In this application scenario, the data analysis system performs detailed semantic analysis on each text block to extract key information and value features for subsequent risk decision analysis.

[0062] In detail, in step 110, the first task of the data analysis system is to obtain the target digital enterprise management report to be analyzed. This report is crucial for subsequent data analysis and risk decision-making because it contains multi-faceted data and information about the enterprise's operations. The data analysis system obtains this report from the enterprise's data center or cloud storage through a secure data transmission channel, such as an encrypted file transfer protocol or a secure FTP server. This target digital enterprise management report is not a simple document, but a complex data structure containing at least two management report text blocks. These text blocks may cover information in key business areas such as the company's financial status, sales performance, supply chain management, human resources status, and market analysis. Each text block presents data in a specific format and structure so that the data analysis system can read and parse it accurately. After obtaining the report, the data analysis system performs a preliminary integrity check to ensure that all necessary text blocks are included and that the data format is as expected. Therefore, the successful completion of step 110 is the basis for subsequent semantic parsing and key semantic optimization operations, and is also the key to ensuring the accuracy of the final risk decision analysis.

[0063] Step 120: semantically parse the target digital enterprise management report to obtain the text block description semantics of each management report text block in the target digital enterprise management report.

[0064] Semantic parsing refers to the in-depth analysis and understanding of text through natural language processing technology, so as to obtain the specific meaning and contextual relationship contained in the text. In this process, the system identifies elements such as vocabulary, phrases, sentence structure, etc. in the text and converts them into a format that can be understood by the computer. In this way, complex information originally expressed in human language, such as opinions, facts, intentions, etc., can be accurately captured and converted into structured data for subsequent data analysis and application. Semantic parsing technology is of great significance for processing unstructured data and improving the accuracy of information retrieval and data mining.

[0065] Furthermore, the descriptive semantics of a text block refers to the detailed meaning and description of the content of a specific text block extracted during the semantic parsing process. Each text block has its own unique theme and content. Through semantic parsing technology, we can obtain the specific information expressed by these text blocks, such as factual descriptions, data presentations, and point of view elaborations. These parsed descriptive semantics not only reflect the basic content of the text block, but also include the logical relationships within the text block, key information, and possible implicit intentions or trends. These descriptive semantics are an important basis for subsequent data analysis, information retrieval, and decision support.

[0066] In detail, in step 120, the data analysis system will perform a key task: semantically parse the target digital enterprise management report. The purpose of this process is to deeply understand the meaning and content of each management report text block in the report, so as to extract the text block description semantics of each text block. First, the data analysis system scans the entire report and identifies each independent text block. These text blocks can include multiple aspects such as financial status, sales performance, supply chain management, human resources, etc., and each text block contains key information and data in this business field. Next, the data analysis system uses advanced natural language processing technology to perform detailed semantic analysis on each text block. This process includes multiple steps such as vocabulary analysis, syntactic analysis, and semantic role labeling, aiming to accurately capture the key information and intentions in the text block. For example, in the financial status text block, the system can identify key indicators such as "income", "cost", and "profit", and analyze the association and trend between them. Through semantic analysis, the data analysis system can generate text block description semantics for each text block. These description semantics not only include the basic content of the text block, but also reflect the logical relationship, data characteristics, and potential trends within the text block. For example, for a sales performance text block, the system can extract key information such as sales growth trends, major sales channels, and customer satisfaction, and save this information in a structured form. It can be seen that step 120 is crucial for subsequent key semantic optimization operations and enterprise operation risk decision analysis. Only by accurately understanding the meaning and content of each text block can the data analysis system perform effective data analysis and risk assessment.

[0067] Step 130: Perform key semantic optimization operations based on the text block description semantics of each management report text block in the target digital enterprise management report to obtain a key semantic optimization vector of the target digital enterprise management report. The key semantic optimization vector is used to reflect the value feature-involved data between each management report text block in the target digital enterprise management report.

[0068] Among them, the key semantic optimization operation refers to further processing and optimizing the extracted text block description semantics based on semantic analysis, in order to emphasize and highlight the key information that has a significant impact on the enterprise's operational decisions. This operation can include reorganizing the description semantics, redistributing weights, strengthening relevance, etc., in order to better understand the inherent connections and value characteristics between text blocks, and how these characteristics affect the overall operation status and risks of the enterprise. Through the key semantic optimization operation, the data analysis system can more accurately identify the key information and risk points in the report, and provide more powerful data support for subsequent enterprise operational risk decision analysis.

[0069] Furthermore, the key semantic optimization vector is a data structure generated by the data analysis system after completing the key semantic optimization operation. This vector integrates the key semantic information of each management report text block in the report and represents it in a numerical form. The key semantics of each text block is converted into a point or a set of values ​​in the vector space, which reflects the relevance and importance between the text blocks and their contribution to the overall value of the enterprise. In this way, the originally complex text information is converted into a mathematical form that is easier for computers to handle, facilitating more in-depth data analysis and mining.

[0070] Thirdly, value feature-involved data refers to data closely related to the enterprise's operational value and risks that are identified and extracted during the key semantic optimization process. These data can include specific values ​​and proportions of key business areas such as financial indicators, market trends, customer satisfaction, and supply chain efficiency. These data not only reflect the current operating status of the enterprise, but also predict future development trends and potential risks. By analyzing these value feature-involved data, enterprises can more accurately evaluate their own operational efficiency and market competitiveness, thereby making more informed decisions.

[0071] In detail, in step 130, the data analysis system will perform a key semantic optimization operation, which is an important part of the entire analysis process. The data analysis system has obtained the text block description semantics of each management report text block in the target digital enterprise management report through the semantic analysis of the previous step. Now, it needs to process these description semantics more finely to reveal the deep connection and value characteristics between the text blocks. The data analysis system will first conduct an in-depth analysis of the description semantics of each text block to identify the key information and data therein. This information may include specific financial data, market trend analysis, supply chain management status, etc., which are all value characteristics that are crucial to the operation of the enterprise. Next, the data analysis system performs a key semantic optimization operation. The purpose of this operation is to strengthen the key semantics that have an important impact on the overall operation of the enterprise, while weakening or ignoring those secondary or redundant information. In this way, the system can more accurately capture the intrinsic connection and value involvement between the various text blocks in the report. After completing the key semantic optimization operation, the data analysis system generates a key semantic optimization vector. This vector is a multi-dimensional data structure, in which each dimension represents the key semantic information of a specific text block in the report. The numerical representation of the vector enables the system to easily perform subsequent mathematical operations and data analysis. This key semantic optimization vector not only reflects the correlation between the various text blocks, but more importantly, it highlights the value characteristics of the data involved in the various management report text blocks in the report. These data are of great significance for evaluating the overall operating status of the enterprise, identifying potential risks, and formulating effective decision-making strategies. For example, through the comparative analysis of financial indicators and market trend data in the vector, the enterprise can promptly discover the impact of market changes on the financial status, and thus adjust the business strategy to deal with potential risks.

[0072] Step 140: Based on the text block description semantics of each management report text block in the target digital enterprise management report and the value feature-involved data represented by the key semantic optimization vector of the target digital enterprise management report, an enterprise operational risk decision analysis is performed on the target digital enterprise management report to obtain an enterprise operational risk decision analysis viewpoint of the target digital enterprise management report. The enterprise operational risk decision analysis viewpoint is used to represent the potential operational risk labels of each management report text block in the target digital enterprise management report.

[0073] Among them, enterprise operational risk decision analysis is a process of comprehensive risk assessment and strategy formulation for the operation and management of an enterprise. It involves in-depth analysis of multiple dimensions such as the internal and external environment, market conditions, financial conditions, supply chain management, and human resources of the enterprise, with the aim of identifying potential risk factors that may affect the operation of the enterprise and providing decision support for the strategic planning and daily operations of the enterprise. This analysis is usually based on a large amount of data and information, including but not limited to financial reports, market analysis reports, operational data, etc., to ensure the accuracy and comprehensiveness of the analysis.

[0074] Secondly, the enterprise operational risk decision analysis viewpoints are the conclusions or opinions drawn from the enterprise operational risk decision analysis. These viewpoints are a comprehensive assessment of the enterprise's operating conditions, including financial stability, market competitiveness, operational efficiency, supply chain management effectiveness, etc. These viewpoints not only reflect the current operating conditions of the enterprise, but also predict the risks and challenges that may be faced in the future, and accordingly put forward corresponding risk response strategies and decision-making suggestions.

[0075] Furthermore, potential operational risk tags are keywords or phrases used to identify and classify various potential operational risks during the enterprise operational risk decision-making analysis process. These tags help enterprises quickly identify and understand different types of risks, such as financial risks, market risks, supply chain risks, etc., and formulate corresponding countermeasures based on the nature and severity of these risks. By labeling various risks, enterprises can monitor and manage these risks more effectively and ensure the sound operation of the enterprise.

[0076] In detail, in step 140, the data analysis system will conduct an enterprise operation risk decision analysis, which is the core link of the entire analysis process. The data analysis system will have obtained the text block description semantics of each management report text block in the target digital enterprise management report through the previous steps, and generated a key semantic optimization vector. These data and information provide a solid foundation for the subsequent risk decision analysis. The data analysis system will first conduct an in-depth analysis of the operating conditions reflected by each text block based on the text block description semantics of each management report text block. For example, the description semantics of the financial status text block may reveal information such as the profitability and cost control of the enterprise; the sales performance text block may reflect market trends, changes in customer demand, etc. Next, the data analysis system will combine the value feature involved data represented by the key semantic optimization vector to conduct a more comprehensive operation risk analysis. These data not only include the correlation information between text blocks, but also highlight the key factors that have an important impact on the overall value of the enterprise. For example, the system can find that the abnormal fluctuation of a financial indicator is closely related to a specific market trend or supply chain management problem. After in-depth analysis and comparison, the data analysis system forms an analysis point of view on the enterprise operation risk decision. These viewpoints are a comprehensive assessment and summary of the potential operational risks in the enterprise management report. They clearly and unambiguously point out the risk points that the enterprise may face now and in the future, such as tight capital chain, intensified market competition, unstable supply chain, etc. Finally, the data analysis system will label these potential operational risks to facilitate the subsequent monitoring and management of the enterprise. These potential operational risk labels not only help enterprises quickly identify risk types, but also provide important references for the formulation of risk response strategies. For example, for the risk label of "tight capital chain", enterprises may need to consider adjusting financial strategies, optimizing cash flow management, and other ways to reduce risks.

[0077] Combining the above content, in a complete application scenario, the data analysis system is conducting an in-depth analysis of a target digital enterprise management report (the report comes from a medium-sized enterprise that hopes to identify potential operational risks through report analysis).

[0078] First, the data analysis system obtains the target digital enterprise management report to be analyzed. This report contains multiple management report text blocks, covering different business areas such as financial status, sales performance, supply chain management, human resources and market analysis.

[0079] Next, the data analysis system uses natural language processing technology to perform semantic analysis on the report. The system recognizes and understands the meaning and content of each text block. For example, the financial status text block describes the company's assets, liabilities, and profit status; the sales performance text block details the product's sales volume, sales channels, and customer feedback.

[0080] Then, the data analysis system performs key semantic optimization operations on the semantics of the parsed text block descriptions. The system extracts key information from each text block through an algorithm and generates a key semantic optimization vector. This vector not only reflects the correlation between each text block, but also highlights the data involved in the value features in the report, such as which business areas have a greater impact on the overall operation of the enterprise and which areas may have potential risks.

[0081] Finally, the data analysis system conducted an enterprise operational risk decision analysis on the digital enterprise management report based on the semantics of the text block description and the key semantic optimization vector. The system combined historical data and industry trends to evaluate the potential operational risks of each text block and gave corresponding risk labels. For example, the system may find that the debt ratio mentioned in the financial status text block is high, so it marks it as "high financial risk"; or it may identify a trend of declining sales in the sales performance text block, and give a label of "risk of declining sales performance".

[0082] This design can accurately and efficiently understand the potential risks in various management areas, provide strong data support for decision makers, and help companies adjust their operating strategies in a timely manner and avoid risks.

[0083] In another complete application scenario, the data analysis system obtains a target digital enterprise management report, which contains two management report text blocks: text block A and text block B.

[0084] Through semantic parsing, the data analysis system generates text block description semantics for each text block. The description semantics of text block A can be: "Sales increased by 10% in the past quarter, new customers increased by 15%, and market feedback was positive." The description semantics of text block B can be: "Production costs increased by 5%, raw material prices fluctuated greatly, and supply chain management faced challenges." These description semantics capture the key information and main points in the text blocks.

[0085] Next, the data analysis system performs a key semantic optimization operation to convert these descriptive semantics into key semantic optimization vectors. These vectors represent the relevance and importance between text blocks in numerical form. The key semantic optimization vector of text block A is: [0.8, 0.6, 0.2], where the numbers may represent the importance of different dimensions such as sales growth, customer growth, and market feedback. The key semantic optimization vector of text block B may be: [0.4, 0.3, 0.9], representing the criticality of production costs, raw material prices, and supply chain management. These vectors not only reflect the information within a single text block, but also imply the value feature-involved data between text blocks, such as the potential correlation between sales growth and rising production costs.

[0086] Based on the semantics of the text block description and the key semantic optimization vector, the data analysis system conducts risk decision analysis for enterprise operations. For text block A, due to the positive sales and customer growth and good market feedback, it can be marked as "low risk" and a recommendation to continue to maintain the market strategy is given. For text block B, due to rising production costs and fluctuations in raw material prices, it can be marked as "medium-high risk" and a recommendation to strengthen supply chain management and cost control is given.

[0087] Ultimately, these analytical perspectives form potential operational risk labels, such as “Low Risk – Market Growth” and “Medium-High Risk – Cost Management Challenges,” which provide companies with clear risk indications and decision support.

[0088] It can be seen that by applying the embodiment of the present application, the target digital enterprise management report to be analyzed can be deeply semantically parsed, so as to accurately extract the text block description semantics of each management report text block. This efficient parsing capability enables enterprises to quickly understand and grasp the key information in the report, and provides an accurate data basis for subsequent risk decision analysis. Through the key semantic optimization operation, the key semantic optimization vector of the target digital enterprise management report is obtained, which not only reflects the inherent connection between each text block in the report, but also highlights the value characteristics that have an important impact on the operation of the enterprise. In this way, the analysis results are more accurate, which helps enterprises to discover potential risk points and opportunities. Based on the text block description semantics and key semantic optimization vectors, a comprehensive enterprise operation risk decision analysis is performed on the target digital enterprise management report. In this way, the operating status of the enterprise can be comprehensively evaluated, potential risks can be discovered in time, and targeted risk response strategies can be provided for the enterprise. Through the enterprise operation risk decision analysis, the potential operation risk labels of each management report text block are obtained. These labels show the risk points that the enterprise may face in an intuitive and concise way, which is convenient for the enterprise to make decisions quickly and reduce operational risks.

[0089] In summary, the embodiments of the present application provide enterprises with an efficient, accurate and comprehensive digital enterprise management report analysis method through efficient semantic parsing, precise key semantic optimization, comprehensive operational risk analysis and intuitive risk labeling, which helps enterprises improve operational efficiency and reduce operational risks.

[0090] In some optional embodiments, the key semantic optimization vector of the target digital enterprise management report is obtained based on the target semantic parsing network; the step of obtaining the target semantic parsing network includes: obtaining a neural network learning sample; wherein the neural network learning sample includes a text block description semantics of a target management report text block in the first digital enterprise management report, a text block description semantics of a target management report text block in the second digital enterprise management report, a text block description semantics of a certified management report text block in the second digital enterprise management report, and a priori annotations of the second digital enterprise management report, wherein the priori annotations are used to characterize whether there is an involvement state between the certified management report text block and the target management report text block; using an initial semantic parsing network, based on the target management report in the first digital enterprise management report The text block description semantics of the target management report text block in the second digital enterprise management report, the text block description semantics of the target management report text block in the second digital enterprise management report, and the text block description semantics of the certified management report text block in the second digital enterprise management report are subjected to key semantic optimization operations to obtain a key semantic optimization vector of the second digital enterprise management report, where the key semantic optimization vector of the second digital enterprise management report is used to characterize whether there is an involvement state between the target management report text block and the certified management report text block in the second digital enterprise management report; based on the condition of minimizing the difference between the prior annotations of the second digital enterprise management report and the content corresponding to the key semantic optimization vector of the second digital enterprise management report, the initial semantic parsing network is subjected to neural network debugging to obtain the target semantic parsing network.

[0091] In this embodiment, first, neural network learning samples are obtained, which is the basis for training the neural network. These learning samples include the text block description semantics of the target management report text block in the first digital enterprise management report and the second digital enterprise management report. In particular, the second digital enterprise management report also includes the text block description semantics of the authenticated management report text block, as well as an important element-a priori annotation. This priori annotation is used to indicate whether there is a certain involvement state between the authenticated management report text block and the target management report text block, that is, whether there is a certain association or dependency relationship between them.

[0092] Next, an initial semantic parsing network is used for processing. This network will perform key semantic optimization operations based on the text block description semantics of the target management report text block in the first and second digital enterprise management reports, and the text block description semantics of the certified management report text block in the second digital enterprise management report. The purpose of this operation is to extract the key semantic information between these text blocks, especially the correlation between them. Through this operation, a key semantic optimization vector can be obtained, which is specifically used to characterize the involvement status between the target management report text block and the certified management report text block in the second digital enterprise management report.

[0093] Then, based on an important principle - minimizing the difference between the prior annotations of the second digital enterprise management report and the content corresponding to the key semantic optimization vector obtained earlier, the initial semantic parsing network is debugged. This process is actually optimizing the parameters and structure of the network to more accurately capture the involvement status between text blocks. After such debugging, an accurate target semantic parsing network can be finally obtained.

[0094] This target semantic parsing network can not only deeply understand the text information in the digital enterprise management report, but also accurately identify the correlation and dependency between different text blocks. This is very valuable for enterprises because it can help them understand the content of the report more comprehensively, discover the connection between different parts, and make more informed decisions.

[0095] In this way, digital enterprise management reports can be parsed and processed more effectively, the readability and usability of the reports can be improved, and more accurate and comprehensive data support can be provided for enterprise decision-making.

[0096] In some further optional embodiments, the initial semantic parsing network is used to perform key semantic optimization operations based on the text block description semantics of the target management report text block in the first digital enterprise management report, the text block description semantics of the target management report text block in the second digital enterprise management report, and the text block description semantics of the authenticated management report text block in the second digital enterprise management report to obtain the key semantic optimization vector of the second digital enterprise management report, including: performing text semantic integration on the text block description semantics of the target management report text block in the first digital enterprise management report and the text block description semantics of each management report text block in the second digital enterprise management report to obtain the text block integrated semantics of the second digital enterprise management report; generating the key semantic optimization vector of the second digital enterprise management report based on the text block integrated semantics and the text block description semantics of the target management report text block in the second digital enterprise management report.

[0097] In this embodiment, the detailed process of using the initial semantic parsing network to perform key semantic optimization operations is as follows.

[0098] First, the system performs text semantic integration on the text block description semantics of the target management report text block in the first digital enterprise management report and the text block description semantics of each management report text block in the second digital enterprise management report. This process can be seen as fusing the information in different text blocks together to form a more comprehensive and richer semantic expression. Specifically, the system can use natural language processing (NLP) technology to extract keywords, phrases, and sentences from each text block and analyze their grammatical structure and semantic relationships. Then, the system fuses this extracted information together to form an integrated text block integrated semantics.

[0099] For example, the target management report text block in the first digital enterprise management report describes a sales growth plan, while the target management report text block in the second digital enterprise management report describes a marketing campaign, and the certified management report text block describes a successful marketing campaign case. When performing text semantic integration, the system combines the relevant information in these three text blocks to form a comprehensive semantic expression of sales growth, marketing promotion, and successful marketing campaign cases.

[0100] Next, the system generates a key semantic optimization vector based on the text block integrated semantics and the text block description semantics of the target management report text block in the second digital enterprise management report. This vector is actually a multi-dimensional numerical representation that captures the key semantic information in the text and converts it into a format that can be understood and operated by computers. Specifically, the system can use machine learning algorithms, such as word embedding technology, to convert the text block integrated semantics and the text block description semantics of the target management report text block into vector representations.

[0101] For a more intuitive understanding, the key semantic optimization vector can be exemplified as a five-dimensional vector: [0.8, -0.2, 0.5, 0.3, -0.1]. Each dimension of this vector represents a specific semantic feature in the text, and the value indicates the importance of this feature in the text.

[0102] Finally, through this key semantic optimization vector, the system can more accurately capture the involvement status between the target management report text block and the certified management report text block in the second digital enterprise management report. This provides enterprises with deeper data insights and decision support.

[0103] In summary, this embodiment can help enterprises to more comprehensively understand the content in the digital enterprise management report and discover the correlation and dependency between different text blocks. This not only improves the readability and usability of the report, but also provides more accurate and comprehensive data support for enterprise decision-making.

[0104] Under a preferred design idea, the method also includes: obtaining the first digital enterprise management report, the potential operational risk label of the target management report text block in the first digital enterprise management report is the target management risk label; determining the certified management risk label that has an involvement status with the target management risk label, and obtaining the management report text block under the certified management risk label to obtain the certified management report text block; outputting the second digital enterprise management report based on the first digital enterprise management report and the certified management report text block.

[0105] Based on this design idea, the method also includes a series of important steps to generate a second digital enterprise management report.

[0106] First, the system obtains the first digital enterprise management report, which contains multiple target management report text blocks, and these text blocks are marked with target management risk tags. These risk tags are used to indicate the possible operational risks in each text block.

[0107] For example, a target management report text block may describe the company's sales strategy and be marked as "high market risk." In other words, the sales strategy may be affected by market fluctuations and has certain operational risks.

[0108] Next, the system determines the certified management risk labels that have an involvement status with the target management risk label. Here, "involvement status" means that there is a certain correlation or mutual influence between the two risk labels. The system may determine this involvement status by analyzing historical data, expert advice, or risk models.

[0109] Continuing with the sales strategy example above, the system may find that "high market risk" is related to the certified management risk label "unstable supply chain". This is because market fluctuations can affect the stability of the supply chain, which in turn has a negative impact on the sales strategy.

[0110] After determining the implicated status, the system retrieves the management report text blocks corresponding to the identified management risk labels, which contain detailed information about the identified risks. In the above example, the system finds the text block describing the "supply chain instability" risk.

[0111] Finally, the system combines the information from the first digital enterprise management report and the certified management report text blocks to output the second digital enterprise management report. This report not only contains the original target management report text blocks, but also adds the certified risk information related to these text blocks. This makes the report more comprehensive and can help enterprises better understand and manage operational risks.

[0112] For example, in the final output of the second digital enterprise management report, next to the text block about sales strategy there can be a risk warning about supply chain stability and corresponding countermeasure recommendations.

[0113] In this way, the system can provide enterprises with more comprehensive and in-depth risk management support. Enterprises can formulate more accurate risk response strategies based on these reports, thereby reducing operational risks and improving operational efficiency. At the same time, the above method also helps to enhance the risk awareness and risk management capabilities within the enterprise, laying a solid foundation for the long-term development of the enterprise.

[0114] Under another preferred design idea, the second digital enterprise management report is generated based on the target NLP text report output network, and the step of obtaining the target NLP text report output network includes: using the initial NLP text report output network, determining the common management risk label that has an involvement status with the target management risk label, and obtaining the management report text block under the common management risk label to obtain the common management report text block; implementing digital enterprise management report output processing based on the first digital enterprise management report and the common management report text block to obtain the NLP enterprise management output report; obtaining the certified digital enterprise management report, the certified digital enterprise management report includes the target management report text block, and the management report text block that has an involvement status with the target management report text block; based on the condition of maximizing the commonality value between the NLP enterprise management output report and the certified digital enterprise management report, performing neural network debugging on the initial NLP text report output network to obtain the target NLP text report output network.

[0115] In another preferred design idea, the generation of the second digital enterprise management report is based on a specific network - the target NLP text report output network. In order to obtain this target NLP text report output network, the following key steps are required.

[0116] First, the system uses an initial NLP text report output network. This network can be just a basic model at the beginning, which needs to be trained and optimized to adapt to specific tasks.

[0117] Next, the system determines the common management risk tags that have an involved status with the target management risk tag. Here, "common management risk tags" refer to those risk tags that are associated with or similar to the target management risk tags. The system may identify these common risks by analyzing a large amount of historical data or using machine learning methods. Once these common risk tags are determined, the system will obtain the management report text blocks related to these tags to form common management report text blocks.

[0118] For example, if the target management risk tag is "supply chain disruption", then the common management risk tag may be "logistics delay" or "raw material shortage", because these risks may lead to supply chain disruption. The system finds text blocks related to these common risk tags from a large number of reports.

[0119] Then, the system uses the initial NLP text report output network, combined with the first digital enterprise management report and the common management report text block, to perform the output processing of the digital enterprise management report. This process can be seen as fusing the original report and the common risk information together to generate a new NLP enterprise management output report. This report contains both the original target management report text block and the common risk information related to it.

[0120] To further improve the performance of the network, the system also obtains certified digital enterprise management reports. These reports are expert-certified or practically verified and contain target management report text blocks and other management report text blocks with relevant status to these text blocks. These reports provide accurate reference standards for the network.

[0121] Finally, the system debugs the neural network of the initial NLP text report output network based on the condition of maximizing the commonality value between the NLP enterprise management output report and the certified digital enterprise management report. This debugging process is to make the report output by the network closer to the certified report, thereby improving the accuracy and credibility of the report. Through this debugging process, the system finally obtains the target NLP text report output network.

[0122] The above-mentioned target NLP text report output network can generate more comprehensive and accurate digital enterprise management reports, helping enterprises to better understand and manage operational risks. At the same time, since common risk information is incorporated into the report, enterprises can more comprehensively grasp the risks they may face and make more informed decisions. In addition, through neural network debugging and optimization, the accuracy and credibility of the report have also been significantly improved, which is of great significance to the enterprise's risk management and decision support.

[0123] In the next step, based on the condition of maximizing the commonality value between the NLP enterprise management output report and the certified digital enterprise management report, the initial NLP text report output network is debugged by a neural network to obtain the target NLP text report output network, including: obtaining a priori annotations of the sentence units of the certified digital enterprise management report, the sentence unit prior annotations are used to characterize the potential operational risk labels corresponding to the management report text blocks to which each certified sentence unit in the certified digital enterprise management report belongs; determining the potential operational risk labels corresponding to the management report text blocks to which each sentence unit in the NLP enterprise management output report belongs, the sentence units in the NLP enterprise management output report and the certified sentence units in the certified digital enterprise management report having a one-to-one matching relationship; based on the condition of maximizing the commonality value between the potential operational risk labels corresponding to the management report text blocks to which each sentence unit in the NLP enterprise management output report belongs and the potential operational risk labels corresponding to the management report text blocks to which the corresponding certified sentence units belong, the initial NLP text report output network is debugged by a neural network to obtain the target NLP text report output network.

[0124] In the next step, the system will debug the neural network of the initial NLP text report output network to obtain the target NLP text report output network. This process is mainly based on maximizing the commonality value between the NLP enterprise management output report and the certified digital enterprise management report.

[0125] First, the system obtains the prior annotations of the word and sentence units of the certified digital enterprise management report. These prior annotations are used to characterize the potential operational risk labels corresponding to the management report text blocks to which each certified word and sentence unit in the report belongs. In simple terms, each word and sentence unit is labeled with a risk label, indicating the operational risk that this word and sentence unit may have.

[0126] For example, a sentence in a certified report, “Raw material costs have increased significantly recently”, could be labeled as “cost control risk”.

[0127] Next, the system determines the potential operational risk labels corresponding to the management report text blocks to which each word unit in the NLP enterprise management output report belongs. The key here is that the word units in the NLP enterprise management output report have a one-to-one match with the certified word units in the certified report. This means that the system compares similar word units in the two reports and labels them with the same or similar risk labels.

[0128] Continuing with the above example, if the NLP output report also contains a description of rising raw material costs, the system will also label it as a “cost control risk”.

[0129] Finally, based on the condition of maximizing the commonality between the potential operational risk labels corresponding to the management report text block to which each sentence unit in the NLP enterprise management output report belongs and the potential operational risk labels corresponding to the management report text block to which the corresponding certified sentence unit belongs, the system performs neural network debugging on the initial NLP text report output network. This process is to make the risk labels of the NLP output report as consistent as possible with the risk labels of the certified report, thereby improving the accuracy and reliability of the NLP output report.

[0130] During the debugging process, the system can adjust the weights and parameters of the network so that similar word units can get the same or similar risk labels in the two reports. Through repeated iterations and optimization, the system will eventually obtain a target NLP text report output network that can generate an NLP enterprise management output report that is highly consistent with the certified report.

[0131] Such a design can not only improve the accuracy and reliability of NLP enterprise management output reports, but also help enterprises to have a more comprehensive understanding of operational risks and make more informed decisions. At the same time, since the above method is debugged based on the prior annotations of the certified reports, it can also effectively utilize existing knowledge and experience to improve the performance and efficiency of the NLP text report output network.

[0132] Under an optional design concept, the method also includes: obtaining the first digital enterprise management report and the certified NLP enterprise management output report, the potential operational risk label of the target management report text block in the first digital enterprise management report is the target management risk label, and the certified NLP enterprise management output report includes the target management report text block, and the management report text block that has an involvement status with the target management report text block; based on the text block description semantics of the target management report text block in the first digital enterprise management report, determining the certified management risk label that has an involvement status with the target management risk label; obtaining the management report text block under the certified management risk label, obtaining the certified management report text block, and outputting the second digital enterprise management report based on the first digital enterprise management report and the certified management report text block.

[0133] Based on this, based on the condition of minimizing the difference between the prior annotations of the second digital enterprise management report and the content corresponding to the key semantic optimization vector of the second digital enterprise management report, the neural network debugging of the initial semantic parsing network to obtain the target semantic parsing network includes: based on the condition of minimizing the difference between the prior annotations of the second digital enterprise management report and the content corresponding to the key semantic optimization vector of the second digital enterprise management report, and maximizing the commonality value between the second digital enterprise management report and the certified NLP enterprise management output report, the neural network debugging of the initial semantic parsing network to obtain the target semantic parsing network.

[0134] Based on this design idea, the system implementation method covers several key steps, aiming to generate a more accurate second digital enterprise management report and further optimize the semantic parsing network.

[0135] First, the system obtains the first digital enterprise management report and the certified NLP enterprise management output report. The first digital enterprise management report contains target management report text blocks, which are marked with target management risk tags to indicate specific operational risks. The certified report contains not only these target management report text blocks, but also other management report text blocks that are involved.

[0136] To illustrate, in the First Digital Enterprise Management Report, there is a target management report text block describing "sales have dropped significantly recently" and is labeled as "market risk". In addition to this information, the certified report may also contain related text blocks such as "inventory backlog" or "customer churn".

[0137] Next, the system determines the authenticated management risk tags that have an involved status with the target management risk tag based on the text block description semantics of the target management report text block in the first digital enterprise management report. This usually involves in-depth analysis and understanding of the text block to determine which risk tags are interrelated.

[0138] Continuing with the above example, the system can analyze that "sales volume drops significantly" is related to risk tags such as "inventory overstock" and "customer churn".

[0139] Subsequently, the system obtains the management report text blocks under these authenticated management risk tags to form a set of authenticated management report text blocks, which will provide important information for generating the second report.

[0140] After obtaining this information, the system combines the first report with the certified management report text block to output a second digital enterprise management report. This report will be more comprehensive because it not only contains the original target management report text block, but also incorporates other risk information related to it.

[0141] In order to further improve the parsing accuracy of the report, the system will also perform neural network debugging on the initial semantic parsing network based on the differences between the prior annotations of the second report and the contents corresponding to the key semantic optimization vectors, as well as the commonality values ​​between the second report and the certified report.

[0142] The "prior annotation" in the embodiment of the present application can be understood as a pre-explanation or annotation of each part of the report, and the "key semantic optimization vector" is a numerical representation of the key information in the report. The system optimizes the semantic parsing network by minimizing the difference between the content corresponding to the prior annotation and the key semantic optimization vector, and maximizing the commonality value between the second report and the certified report.

[0143] For example, if a block of text in the second report is a priori annotated as "high cost risk", but the content corresponding to its key semantic optimization vector differs from the description of the actual cost risk, the system will adjust the semantic parsing network to enable it to more accurately understand and interpret this block of text.

[0144] Similarly, if there is a high degree of consistency between the second report and the certified report regarding the description of a risk, the system will use this information to optimize the semantic parsing network so that it can more accurately identify such risks in future parsing.

[0145] In this way, the system can not only generate more comprehensive and accurate enterprise management reports, but also continuously improve its semantic parsing capabilities, thereby providing enterprises with more effective risk management support.

[0146] In some other embodiments, the method further includes: obtaining a target report semantic relationship network, wherein the target report semantic relationship network is used to represent the semantic delivery chain corresponding to the target digital enterprise management report; performing a first feature embedding process at a logical description level on the target report semantic relationship network to obtain a first report semantic relationship network at a first logical description level, wherein the first logical description level is smaller than a basic logical description level of the target report semantic relationship network; performing a second feature embedding process at a sentiment description level on the first report semantic relationship network to obtain a second report semantic relationship network, wherein the first sentiment description level of the second report semantic relationship network is larger than the original sentiment description level of the target report semantic relationship network, and the sentiment feature recognition of the second report semantic relationship network The degree is greater than the original feature recognition degree of the target report semantic relationship network; the third feature embedding processing of the sentiment description level is performed on the second report semantic relationship network to obtain a third report semantic relationship network, wherein the second sentiment description level of the third report semantic relationship network is consistent with the original sentiment description level, and the sentiment feature recognition degree of the third report semantic relationship network is consistent with the sentiment feature recognition degree of the second report semantic relationship network; the fourth feature embedding processing of the logical description level is performed on the third report semantic relationship network to obtain a fourth report semantic relationship network of the second logical description level, wherein the second logical description level is consistent with the basic logical description level; the fourth report semantic relationship network is used to construct a structured text tree for the target digital enterprise management report.

[0147] In other embodiments, the method executed by the system further enhances the processing and analysis capabilities of the report semantic relationship network to more accurately understand and structure the content of the digital enterprise management report.

[0148] First, the system obtains the semantic relationship network of the target report. This network is the semantic expression of the digital enterprise management report. It depicts the relationship between the concepts and entities in the report, forming a complex semantic transmission chain. This chain can be regarded as a graph structure, in which the nodes represent concepts or entities, and the edges represent the relationship between them.

[0149] In order to conduct a deeper analysis of this complex semantic relationship, the system will first simplify it. This is the purpose of the first feature embedding process, which reduces the logical description level to obtain a first report semantic relationship network of the first logical description level. The "logical description level" here can be understood as the complexity of describing the relationship or attribute of things. By reducing the level, the system can focus on more core and direct semantic relationships.

[0150] However, while simplifying the logical hierarchy, the system also needs to enhance the recognition of the emotional color in the report. Therefore, the system will then perform a second feature embedding process, which focuses on improving the emotional description level and feature recognition. This means that the system not only pays attention to the objective factual description in the report, but also begins to pay attention to the emotional tendencies contained therein, such as positive, negative or neutral. Through this processing, the second report semantic relationship network can capture the information in the report more comprehensively.

[0151] After enhancing the ability to describe emotions, the system then performs a third feature embedding process, this time to restore the emotional description level to a state consistent with the original level while maintaining the previously improved feature recognition. The purpose of this step is to ensure that the overall structure of the semantic relationship network is consistent with the original report while maintaining sensitivity to emotional color.

[0152] Finally, the system performs the fourth feature embedding process, restores the logical description level to the basic level, and obtains a fourth report semantic relationship network at the second logical description level. This network simplifies the complex semantic relationships and retains the emotional color and key features in the report, providing a solid foundation for the subsequent structured text tree construction.

[0153] Through this processing flow, the system can more accurately parse and understand the content of digital enterprise management reports. This not only helps enterprises better grasp the key information in the report, but also provides more comprehensive and in-depth data support for decision-making. It can be seen that the above series of feature embedding processing operations enable the system to gradually and comprehensively parse and understand enterprise management reports, from logical structure to emotional color, to the identification of key features, each step enhances the system's ability to grasp the content of the report. Ultimately, these processing operations together provide a high-quality semantic relationship network for the construction of a structured text tree, thereby improving the efficiency of enterprises' use of report data and the accuracy of decision support.

[0154] In some optional embodiments, the second feature embedding processing of the sentiment description level is performed on the first report semantic relationship network to obtain the second report semantic relationship network, including: performing first feature embedding processing of the semantic depth level and the semantic coverage level on the first report semantic relationship network to obtain a first embedded semantic relationship network, wherein the semantic depth level of the first embedded semantic relationship network is greater than the original semantic depth level of the target report semantic relationship network, the semantic coverage level of the first embedded semantic relationship network is greater than the original semantic coverage level of the target report semantic relationship network, and the sentiment feature recognition degree of the first embedded semantic relationship network is greater than the original feature recognition degree; based on the first embedded semantic relationship network, the second report semantic relationship network is obtained.

[0155] Then the third feature embedding processing of the sentiment description level is implemented on the second report semantic relationship network to obtain the third report semantic relationship network, including: implementing the second feature embedding processing of the semantic depth level and the semantic coverage level on the second report semantic relationship network to obtain a second embedded semantic relationship network, wherein the semantic depth level of the second embedded semantic relationship network is consistent with the original semantic depth level, the semantic coverage level of the second embedded semantic relationship network is consistent with the original semantic coverage level, and the sentiment feature recognition degree of the second embedded semantic relationship network is consistent with the sentiment feature recognition degree of the first embedded semantic relationship network; based on the second embedded semantic relationship network, the third report semantic relationship network is obtained.

[0156] In some optional embodiments, the system performs a series of refined processing on the first report semantic relationship network to obtain deeper semantic analysis and sentiment recognition.

[0157] First, the system implements the first feature embedding processing of the semantic depth level and semantic coverage level on the first report semantic relationship network. The "semantic depth level" here can be understood as the complexity of the concepts, information and logic involved in the report content, while the "semantic coverage level" refers to the scope or breadth covered by the report content. Through the first feature embedding processing, the system improves these two levels, thereby obtaining the first embedded semantic relationship network.

[0158] Specifically, the semantic depth level of the first embedded semantic relationship network is raised to a higher level than the original semantic depth level of the target report semantic relationship network. This means that the system can dig deeper into the meaning and contextual relationships in the report. At the same time, the semantic coverage level of the first embedded semantic relationship network also exceeds the original semantic coverage level, indicating that the system can now understand and analyze a wider range of content in the report.

[0159] In addition, the sentiment feature recognition of the first embedded semantic relationship network has also been improved, exceeding the original feature recognition, which means that the system can now not only understand the basic semantics of the report, but also more accurately identify and interpret the emotional color and tendency.

[0160] After completing the construction of the first embedded semantic relationship network, the system obtains the second report semantic relationship network based on this enhanced network. This network not only retains the semantic structure of the original report, but also incorporates deeper and more extensive semantic and sentiment analysis.

[0161] Next, the system further embedded the features of the semantic relationship network of the second report. This time, the system adjusted the semantic depth level and semantic coverage level to restore them to the same state as the original level. The purpose of this is to ensure that the overall structure of the semantic relationship network is consistent with the original report while maintaining a deep understanding of the report content, avoiding semantic distortion caused by over-parsing.

[0162] Through this series of feature embedding processing, the system obtains the second embedded semantic relationship network. This network incorporates richer sentiment analysis and recognition capabilities while maintaining the original semantic structure. Finally, the system obtains the third report semantic relationship network based on the second embedded semantic relationship network. This design enables the system to understand the semantic content of the report more deeply, while accurately identifying and interpreting the emotional color therein. This not only helps to improve the company's ability to interpret report data, but also provides more comprehensive and in-depth information support for decision-making. By enhancing the depth and breadth of semantics and improving the recognition of emotional features, the system provides companies with a more powerful and intelligent semantic analysis tool, thereby promoting data-driven decision optimization and efficiency improvement.

[0163] In other independent embodiments, the first report semantic relationship network is subjected to the first feature embedding processing of the semantic depth level and the semantic coverage level to obtain the first embedded semantic relationship network, including: the first target feature embedding processing of the semantic depth level is subjected to the first report semantic relationship network to obtain the first target report semantic relationship network; the second target feature embedding processing of the semantic coverage level is subjected to the first target report semantic relationship network to obtain the second target report semantic relationship network; or, the second target feature embedding processing of the semantic coverage level is subjected to the first report semantic relationship network to obtain the third target report semantic relationship network; the first target feature embedding processing of the semantic depth level is subjected to the third target report semantic relationship network to obtain the fourth target report semantic relationship network.

[0164] Then, based on the first embedded semantic relationship network, obtaining the second report semantic relationship network includes: on the basis that the first embedded semantic relationship network includes the second target report semantic relationship network, using the second target report semantic relationship network as the second report semantic relationship network; or, on the basis that the first embedded semantic relationship network includes the fourth target report semantic relationship network, using the fourth target report semantic relationship network as the second report semantic relationship network.

[0165] In some other independent embodiments, the system processes the first report semantic relationship network in a more detailed and flexible manner. Specifically, the system can perform feature embedding processing through two different paths to obtain the first embedded semantic relationship network.

[0166] First, the system can perform the first target feature embedding processing at the semantic depth level on the first report semantic relationship network. This step is mainly to increase the depth and complexity of the semantic relationship, so that the system can understand the semantic content in the report more deeply. In this way, the system can obtain the first target report semantic relationship network, which retains the basic semantic structure of the original report while increasing the ability to express deep semantic relationships.

[0167] Next, the system can perform the second target feature embedding processing at the semantic coverage level on the first target report semantic relationship network. The purpose of this step is to expand the coverage of semantic relationships so that the system can capture more semantic information in the report. After this processing, the system obtains the second target report semantic relationship network, which not only has a deep level of semantic understanding ability, but also can widely cover various parts of the report.

[0168] In addition to the above path, the system can also choose another processing order. First, the system implements the second target feature embedding processing of the semantic coverage level on the first report semantic relationship network to obtain the third target report semantic relationship network. This step is mainly to enable the system to capture the semantic information in the report extensively.

[0169] Then, the system performs the first target feature embedding process on the semantic depth level of the third target report semantic relationship network. In this way, the system further increases its ability to understand deep semantic relationships on the basis of already extensively covering semantic information. Through this process, the system obtains the fourth target report semantic relationship network, which has both a wide semantic coverage and a deep understanding of the deep semantic relationships in the report.

[0170] Finally, the system obtains the second report semantic network according to the specific form of the first embedded semantic network. If the first embedded semantic network is obtained by first increasing the semantic depth and then expanding the semantic coverage (i.e., the second target report semantic network), then the system directly uses this network as the second report semantic network. Conversely, if the first embedded semantic network is obtained by first expanding the semantic coverage and then increasing the semantic depth (i.e., the fourth target report semantic network), then the system uses this network as the second report semantic network.

[0171] In this way, the system can flexibly adjust the processing path according to different needs to obtain the second report semantic relationship network that best suits the current task requirements. This flexibility and adaptability enables the system to better cope with various complex semantic analysis tasks and improves the accuracy and efficiency of semantic analysis. At the same time, because the system can more deeply understand and capture the semantic information in the report, it also provides enterprises with more comprehensive and in-depth data support, helping enterprises make more informed decisions.

[0172] In some other independent embodiments, the first report semantic relationship network is subjected to first feature embedding processing at the semantic depth level and the semantic coverage level to obtain a first embedded semantic relationship network, including: the first report semantic relationship network is subjected to first target feature embedding processing at the semantic depth level to obtain a first target report semantic relationship network; and the first report semantic relationship network is subjected to second target feature embedding processing at the semantic coverage level to obtain a third target report semantic relationship network.

[0173] Then, based on the first embedded semantic relationship network, the second report semantic relationship network is obtained, including: taking the first target report semantic relationship network and the third target report semantic relationship network as the second report semantic relationship network; or integrating the first target report semantic relationship network and the third target report semantic relationship network to obtain the second report semantic relationship network.

[0174] In other independent embodiments, the system adopts a parallel processing strategy for processing the first report semantic relationship network, that is, feature embedding processing is performed on the semantic depth level and the semantic coverage level at the same time.

[0175] First, the system implements the first target feature embedding processing at the semantic depth level on the semantic relationship network of the first report. This step is mainly to enhance the deep semantic information in the report, making the semantic relationships that may have been implicit or difficult to understand directly become clearer. For example, the report may mention "income growth". Through the semantic depth level embedding processing, the system can further dig out the specific reasons and influencing factors behind the income growth. In this way, the system obtains the semantic relationship network of the first target report, which highlights the deeper semantic content in the original report.

[0176] At the same time, the system also implements the second target feature embedding processing at the semantic coverage level for the semantic relationship network of the first report. The purpose of this processing is to expand the semantic coverage of the report and ensure that more relevant information is taken into consideration. For example, the original report may focus on a certain aspect of the business. Through the embedding processing at the semantic coverage level, the system can also include other aspects related to the business into the semantic relationship network. The semantic relationship network of the third target report processed in this way expands its semantic boundaries while maintaining the core semantics of the original report.

[0177] After completing the above two feature embedding processes, the system has two ways to obtain the second report semantic relationship network. One way is to directly use the first target report semantic relationship network and the third target report semantic relationship network as two components of the second report semantic relationship network. The advantage of this is that the independence of the two different processing results is retained, which is convenient for subsequent analysis based on different semantic requirements.

[0178] Another way is to integrate the first target report semantic relationship network and the third target report semantic relationship network to form a more comprehensive second report semantic relationship network. The integration process can be a simple merger, or it can be a fusion of information in the two networks through some algorithm (such as weighted average, voting mechanism, etc.). The second report semantic relationship network obtained in this way maintains the semantic depth and breadth while enhancing the comprehensiveness and consistency of information.

[0179] The above processing method provides enterprises with more flexibility. Enterprises can choose to directly use two independent semantic relationship networks or integrate them for analysis according to their own analysis needs. Either way can help enterprises understand the report content more comprehensively and dig out more valuable information.

[0180] In some other independent embodiments, after taking the first target report semantic relationship network and the third target report semantic relationship network as the second report semantic relationship network, the second report semantic relationship network is subjected to the second feature embedding processing of the semantic depth level and the semantic coverage level to obtain the second embedded semantic relationship network, including: subjecting the first target report semantic relationship network to the third feature embedding processing of the semantic depth level to obtain the first local embedded semantic relationship network; and subjecting the third target report semantic relationship network to the fourth feature embedding processing of the semantic coverage level to obtain the second local embedded semantic relationship network; integrating the first local embedded semantic relationship network and the second local embedded semantic relationship network to obtain the second embedded semantic relationship network.

[0181] In other independent embodiments, after the system uses the first target report semantic relationship network and the third target report semantic relationship network as the second report semantic relationship network, it further refines the integrated network. This stage of processing involves re-embedding features of the semantic depth level and the semantic coverage level to extract more accurate and comprehensive semantic information.

[0182] First, the system implements the third feature embedding processing of the semantic depth level for the first target report semantic relationship network. Since the semantic depth of the first target report semantic relationship network has been enhanced in the previous step, the third feature embedding processing in this step is to further explore the deep semantic relationships therein. For example, the system may use complex algorithms and models to identify and strengthen those semantic connections that are implicit, indirect, or require reasoning. In this way, the system obtains the first local embedded semantic relationship network, which is a network richer in semantic depth.

[0183] At the same time, the system also implemented the fourth feature embedding processing of the semantic coverage level for the third target report semantic relationship network. The purpose of this processing is to further expand the boundaries of the network on the basis of the already relatively broad semantic coverage and capture more peripheral information related to the core semantics. For example, the system may enhance the semantic coverage of the network by analyzing the co-occurrence relationship between words, contextual connections, or using external knowledge bases. After such processing, the system obtained the second local embedded semantic relationship network, which has been further improved in the breadth of semantic coverage.

[0184] Finally, the system integrates the first local embedded semantic relationship network with the second local embedded semantic relationship network to form a more comprehensive and in-depth second embedded semantic relationship network. This network not only contains the deep semantic information in the original report, but also expands its semantic boundaries, allowing the system to understand and interpret the content of the report more accurately.

[0185] Through such a processing flow, the system can provide enterprises with more sophisticated and comprehensive semantic analysis services. This helps enterprises gain deeper insights into the information in the reports, discover potential business opportunities or risks, and make more informed decisions. At the same time, because the system maintains a high degree of flexibility and scalability throughout the entire processing process, it can easily handle report data of different types and sizes, providing strong support for the company's data analysis work.

[0186] It should be noted that the differences between the above different feature embedding processes are as follows.

[0187] 1) First feature embedding process:

[0188] Processing object: first report semantic relationship network (original data);

[0189] Objective: To perform preliminary feature embedding on the original data and extract key information;

[0190] Features: This is the initial embedding process, which aims to capture the most basic feature information from the raw data.

[0191] 2) Second feature embedding processing:

[0192] Processing object: the second report semantic relationship network after the first feature embedding process;

[0193] Goal: To further enhance the depth and breadth of semantic relationships based on existing feature embedding;

[0194] Features: The second feature embedding process is a further optimization and enhancement of the previous embedding results, making the semantic relationship more complex and rich.

[0195] 3) Third feature embedding processing:

[0196] Processing object: The first target is to report the semantic relationship network (focusing on the semantic depth network);

[0197] Goal: To embed features at a deeper level in terms of semantic depth and to mine deeper semantic relationships;

[0198] Features: This processing is specifically designed to enhance semantic depth, aiming to discover those less obvious but important deep semantic connections.

[0199] 4) The fourth feature embedding process:

[0200] Processing object: The third target reports the semantic relationship network (focusing on the network of semantic coverage);

[0201] Goal: To further enhance the semantic coverage of the network while maintaining semantic breadth;

[0202] Features: The fourth feature embedding process is committed to capturing more peripheral information related to the core semantics, making the semantic coverage of the network more comprehensive.

[0203] In summary, the first feature embedding process is the preliminary embedding of the original data, the second feature embedding process is a further optimization based on the first embedding, and the third and fourth feature embedding processes are specially strengthened for semantic depth and semantic coverage respectively. These four processing steps are progressive and together constitute a comprehensive and in-depth feature embedding process.

[0204] Further, Figure 2 Schematic diagram of a data analysis system 200 provided in an embodiment of the present application. Figure 2The data analysis system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0205] Alternatively, if Figure 2 As shown, the data analysis system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present application.

[0206] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0207] Alternatively, if Figure 2 As shown, the data analysis system 200 may further include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices, specifically, may send information or data to other devices, or receive information or data sent by other devices.

[0208] Optionally, the data analysis system 200 may implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or a device deployed with a storage engine in each method of the embodiments of the present application, which will not be described in detail here for the sake of brevity.

[0209] It should be understood that the processor of the embodiment of the present application may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method embodiment can be completed by the hardware integrated logic circuit or software instructions in the processor. The above processor can be a general processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or the hardware and software modules in the decoding processor are combined and performed. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, and other mature storage media in the art. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0210] It can be understood that the memory in the embodiments of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DRRAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0211] It should be understood that the above-mentioned memory is exemplary but not restrictive. For example, the memory in the embodiments of the present application may also be static random access memory (static RAM, SRAM), dynamic random access memory (dynamic RAM, DRAM), synchronous dynamic random access memory (synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (double data rate SDRAM, DDR SDRAM), enhanced synchronous dynamic random access memory (enhanced SDRAM, ESDRAM), synchronous link dynamic random access memory (synch link DRAM, SLDRAM) and direct memory bus random access memory (Direct Rambus RAM, DRRAM), etc. That is to say, the memory in the embodiments of the present application is intended to include but not limited to these and any other suitable types of memory.

[0212] Based on the above, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program implements the above method when running.

[0213] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations.

Claims

1. A data analysis method based on digital enterprise management, characterized in that: Applied to a data analysis system, the method comprises: Obtaining a target digital enterprise management report to be analyzed, wherein the target digital enterprise management report includes at least two management report text blocks; Performing semantic analysis on the target digital enterprise management report to obtain the text block description semantics of each management report text block in the target digital enterprise management report; Performing a key semantic optimization operation based on the text block description semantics of each management report text block in the target digital enterprise management report to obtain a key semantic optimization vector of the target digital enterprise management report, wherein the key semantic optimization vector is used to reflect the value feature-related data between each management report text block in the target digital enterprise management report; According to the text block description semantics of each management report text block in the target digital enterprise management report and the value feature-involved data represented by the key semantic optimization vector of the target digital enterprise management report, an enterprise operation risk decision analysis is performed on the target digital enterprise management report to obtain an enterprise operation risk decision analysis viewpoint of the target digital enterprise management report, and the enterprise operation risk decision analysis viewpoint is used to represent the potential operation risk label of each management report text block in the target digital enterprise management report; The key semantic optimization vector of the target digital enterprise management report is obtained based on the target semantic parsing network; the step of obtaining the target semantic parsing network includes: Obtaining a neural network learning sample; wherein the neural network learning sample includes a text block description semantics of a target management report text block in a first digital enterprise management report, a text block description semantics of a target management report text block in a second digital enterprise management report, a text block description semantics of a certified management report text block in the second digital enterprise management report, and a priori annotations of the second digital enterprise management report, wherein the priori annotations are used to characterize whether there is an involvement state between the certified management report text block and the target management report text block; Using an initial semantic parsing network, a key semantic optimization operation is performed based on the text block description semantics of the target management report text block in the first digital enterprise management report, the text block description semantics of the target management report text block in the second digital enterprise management report, and the text block description semantics of the authenticated management report text block in the second digital enterprise management report, to obtain a key semantic optimization vector of the second digital enterprise management report, wherein the key semantic optimization vector of the second digital enterprise management report is used to characterize whether there is an involvement state between the target management report text block and the authenticated management report text block in the second digital enterprise management report; Based on the condition of minimizing the difference between the prior annotation of the second digital enterprise management report and the content corresponding to the key semantic optimization vector of the second digital enterprise management report, performing neural network debugging on the initial semantic parsing network to obtain the target semantic parsing network; The initial semantic parsing network is used to perform a key semantic optimization operation based on the text block description semantics of the target management report text block in the first digital enterprise management report, the text block description semantics of the target management report text block in the second digital enterprise management report, and the text block description semantics of the authenticated management report text block in the second digital enterprise management report to obtain a key semantic optimization vector of the second digital enterprise management report, including: Performing text semantic integration on the text block description semantics of the target management report text block in the first digital enterprise management report and the text block description semantics of each management report text block in the second digital enterprise management report to obtain the text block integrated semantics of the second digital enterprise management report; A key semantic optimization vector of the second digital enterprise management report is generated according to the text block integration semantics and the text block description semantics of the target management report text block in the second digital enterprise management report.

2. The method according to claim 1, characterized in that The method further comprises: Acquire the first digital enterprise management report, wherein the potential operational risk tag of the target management report text block in the first digital enterprise management report is a target management risk tag; Determine an authenticated management risk tag that has an involvement status with the target management risk tag, and obtain a management report text block under the authenticated management risk tag to obtain the authenticated management report text block; The second digital enterprise management report is outputted based on the first digital enterprise management report and the authenticated management report text block.

3. The method according to claim 2, characterized in that The second digital enterprise management report is generated based on the target NLP text report output network, and the step of obtaining the target NLP text report output network includes: Using an initial NLP text report output network, determining a common management risk label that is involved with the target management risk label, and obtaining a management report text block under the common management risk label to obtain a common management report text block; Performing digital enterprise management report output processing based on the first digital enterprise management report and the common management report text block to obtain an NLP enterprise management output report; Acquire a certified digital enterprise management report, wherein the certified digital enterprise management report includes the target management report text block and management report text blocks that are involved with the target management report text block; Based on the condition of maximizing the commonality value between the NLP enterprise management output report and the authenticated digital enterprise management report, the initial NLP text report output network is debugged by a neural network to obtain the target NLP text report output network.

4. The method according to claim 3, characterized in that: The neural network debugging of the initial NLP text report output network based on the condition of maximizing the commonality value between the NLP enterprise management output report and the authenticated digital enterprise management report to obtain the target NLP text report output network includes: Acquire a priori annotations of the word and sentence units of the certified digital enterprise management report, wherein the a priori annotations of the word and sentence units are used to characterize potential operational risk labels corresponding to management report text blocks to which each certified word and sentence unit in the certified digital enterprise management report belongs; Determine a potential operational risk label corresponding to a management report text block to which each word unit in the NLP enterprise management output report belongs, wherein the word unit in the NLP enterprise management output report has a one-to-one matching relationship with the authenticated word unit in the authenticated digital enterprise management report; Based on the condition of maximizing the commonality value between the potential operational risk label corresponding to the management report text block to which each word and sentence unit in the NLP enterprise management output report belongs and the potential operational risk label corresponding to the management report text block to which the corresponding authenticated word and sentence unit belongs, the initial NLP text report output network is subjected to neural network debugging to obtain the target NLP text report output network.

5. The method according to claim 1, characterized in that The method further comprises: Acquire the first digital enterprise management report and the certified NLP enterprise management output report, wherein the potential operational risk tag of the target management report text block in the first digital enterprise management report is a target management risk tag, and the certified NLP enterprise management output report includes the target management report text block and a management report text block having an involvement status with the target management report text block; Determining, based on the text block description semantics of the target management report text block in the first digital enterprise management report, an authenticated management risk tag having an involvement status with the target management risk tag; Obtaining the management report text block under the authenticated management risk tag, obtaining the authenticated management report text block, and outputting the second digital enterprise management report based on the first digital enterprise management report and the authenticated management report text block; The neural network debugging of the initial semantic parsing network based on the condition of minimizing the difference between the prior annotation of the second digital enterprise management report and the content corresponding to the key semantic optimization vector of the second digital enterprise management report to obtain the target semantic parsing network includes: Based on the conditions of minimizing the difference between the prior annotations of the second digital enterprise management report and the contents corresponding to the key semantic optimization vectors of the second digital enterprise management report, and maximizing the commonality value between the second digital enterprise management report and the certified NLP enterprise management output report, the initial semantic parsing network is neural network debugged to obtain the target semantic parsing network.

6. The method according to claim 1, characterized in that The method further comprises: Acquire a target report semantic relationship network, wherein the target report semantic relationship network is used to represent a semantic delivery chain corresponding to the target digital enterprise management report; Performing a first feature embedding process of a logical description level on the target report semantic relationship network to obtain a first report semantic relationship network of a first logical description level, wherein the first logical description level is smaller than a basic logical description level of the target report semantic relationship network; Implementing a second feature embedding process of the sentiment description level on the first report semantic relationship network to obtain a second report semantic relationship network, wherein the first sentiment description level of the second report semantic relationship network is greater than the original sentiment description level of the target report semantic relationship network, and the sentiment feature recognition degree of the second report semantic relationship network is greater than the original feature recognition degree of the target report semantic relationship network; Implementing the third feature embedding process of the sentiment description level on the second report semantic relationship network to obtain a third report semantic relationship network, wherein the second sentiment description level of the third report semantic relationship network is consistent with the original sentiment description level, and the sentiment feature recognition degree of the third report semantic relationship network is consistent with the sentiment feature recognition degree of the second report semantic relationship network; Implementing a fourth feature embedding process of a logical description level on the third report semantic relationship network to obtain a fourth report semantic relationship network of a second logical description level, wherein the second logical description level is consistent with the basic logical description level; The fourth report semantic relationship network is used to construct a structured text tree for the target digital enterprise management report.

7. The method according to claim 6, characterized in that The method of performing a second feature embedding process at a sentiment description level on the first report semantic relationship network to obtain a second report semantic relationship network includes: performing a first feature embedding process at a semantic depth level and a semantic coverage level on the first report semantic relationship network to obtain a first embedded semantic relationship network, wherein the semantic depth level of the first embedded semantic relationship network is greater than the original semantic depth level of the target report semantic relationship network, the semantic coverage level of the first embedded semantic relationship network is greater than the original semantic coverage level of the target report semantic relationship network, and the sentiment feature recognition degree of the first embedded semantic relationship network is greater than the original feature recognition degree; based on the first embedded semantic relationship network, obtaining the second report semantic relationship network; Then the third feature embedding processing of the sentiment description level is implemented on the second report semantic relationship network to obtain the third report semantic relationship network, including: implementing the second feature embedding processing of the semantic depth level and the semantic coverage level on the second report semantic relationship network to obtain a second embedded semantic relationship network, wherein the semantic depth level of the second embedded semantic relationship network is consistent with the original semantic depth level, the semantic coverage level of the second embedded semantic relationship network is consistent with the original semantic coverage level, and the sentiment feature recognition degree of the second embedded semantic relationship network is consistent with the sentiment feature recognition degree of the first embedded semantic relationship network; based on the second embedded semantic relationship network, the third report semantic relationship network is obtained.

8. A data analysis system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 7.

Images