Report text privacy protection method and system for digital asset evaluation

Abstract

The present application relates to the field of artificial intelligence technology, and provides a method and system for privacy protection of report text for digital asset evaluation, and specifically relates to a privacy protection method for digital asset evaluation reports based on a text privacy protection decision network. By utilizing natural language processing and deep learning technology, the embodiment of the present application can automatically mine the dense text semantics of the report, and compare it with the original dense text semantic set previously mined, so as to accurately determine the sensitivity of the report. This method not only significantly improves the efficiency and accuracy of privacy protection, but also greatly reduces the cost and risk of manual review, providing strong technical support for the digital asset security of enterprises.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a report text privacy protection method and system for digital asset evaluation. Background Art

[0002] In the digital age, digital asset evaluation reports, as carriers of important corporate information, are receiving increasing attention for their security and privacy. However, traditional privacy protection methods often rely on manual review or simple keyword filtering, which is inefficient and error-prone, and difficult to cope with the increasingly complex and changing risks of privacy leakage. Especially when dealing with a large number of reports, the burden of manual review is extremely heavy, and the accuracy and consistency of the review cannot be guaranteed. Therefore, there is an urgent need for an intelligent privacy protection method that can automatically and accurately identify sensitive information in digital asset evaluation reports and perform appropriate privacy protection. Summary of the invention

[0003] In order to improve the above problems, the present application provides a report text privacy protection method and system for digital asset evaluation.

[0004] In a first aspect of an embodiment of the present application, a report text privacy protection method for digital asset evaluation is provided, which is applied to a report text privacy protection system, and the method includes:

[0005] Obtain a target digital asset evaluation report and an original dense text semantic set; the original dense text semantics in the original dense text semantic set is a feature relationship network of privacy semantics obtained by performing dense semantic mining on a positive sample set of evaluation reports corresponding to each debugging instruction in a debugging instruction set during the debugging process of a text privacy protection decision network;

[0006] Performing dense semantic mining processing on the target digital asset evaluation report through the target dense semantic mining branch in the text privacy protection decision network to obtain the target dense text semantics of the target digital asset evaluation report;

[0007] The privacy protection judgment viewpoint of the target digital asset evaluation report is determined based on the difference in semantic features between the target dense text semantics and each original dense text semantics in the original dense text semantic set; the privacy protection judgment viewpoint characterizes whether the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection.

[0008] Optionally, the debugging steps of the text privacy protection decision network include:

[0009] Obtaining a set of past digital asset evaluation reports corresponding to each debugging instruction in the debugging instruction set; the set of past digital asset evaluation reports includes a set of positive evaluation report samples and a set of derived evaluation report samples, wherein the derived evaluation report samples in the derived evaluation report sample set are obtained by deriving the corresponding positive evaluation report samples;

[0010] Determine the target debugging instruction in the debugging instruction set, debug the basic text privacy protection decision network through the evaluation report positive sample set of the target debugging instruction, and obtain the original text privacy protection decision network and the first dense text semantic set of the target debugging instruction in the debugging process;

[0011] Determining a feature relationship network of the first dense text semantic set as the original dense text semantics in the original dense text semantic set;

[0012] Freeze the discriminant branch in the original text privacy protection decision network, debug the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set, and obtain the target dense semantic mining branch and the second dense text semantic set of each pending debugging instruction in the debugging process; the second dense text semantic set is a dense text semantic set obtained by dense semantic mining of the positive sample set of the evaluation report of the pending debugging instruction, and the pending debugging instruction set is a plurality of debugging instructions in the debugging instruction set except the target debugging instruction;

[0013] Determine a feature relationship network of each of the second dense text semantic sets as the original dense text semantics in the original dense text semantic set, and adjust the original dense text semantic set;

[0014] Among them, the discrimination branch and the target dense semantic mining branch constitute the text privacy protection decision network.

[0015] Optionally, debugging the basic text privacy protection decision network through the evaluation report positive sample set of the target debugging instruction includes:

[0016] Performing dense semantic mining processing on the evaluation report positive sample set of the target debugging instruction through the to-be-debugged dense semantic mining branch in the basic text privacy protection decision network to obtain the first dense text semantic set;

[0017] Performing discriminative training on the evaluation report positive sample set of the target debugging instruction according to the first dense text semantic set through the to-be-debugged discriminative branch in the basic text privacy protection decision network, thereby obtaining a first discriminative training viewpoint on the evaluation report positive sample set of the target debugging instruction;

[0018] The configuration weights of the dense semantic mining branch to be debugged and the discriminant branch to be debugged are improved according to the first discriminant training viewpoint until the first set debugging termination requirement is met, thereby obtaining the original dense semantic mining branch and the discriminant branch.

[0019] Optionally, debugging the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set includes:

[0020] Performing dense semantic mining processing on the evaluation report positive sample set and the evaluation report derived sample set of each of the debugging instructions to be processed through the original dense semantic mining branch to obtain a second dense text semantic set of the evaluation report positive sample set and a third dense text semantic set of the evaluation report derived sample set;

[0021] The discriminant branch performs discriminant training on the past digital asset evaluation report set according to the second dense text semantic set and performs discriminant training on the evaluation report derived sample set according to the third dense text semantic set, thereby obtaining a second discriminant training viewpoint of the evaluation report positive sample set of the pending debugging instruction and a third discriminant training viewpoint of the evaluation report derived sample set;

[0022] The configuration weight of the original dense semantic mining branch is improved according to the second discriminant training viewpoint and the third discriminant training viewpoint until the second set debugging termination requirement is met to obtain the target dense semantic mining branch.

[0023] Optionally, the original dense text semantics includes dense text mean semantics and semantic jump features;

[0024] The determining of the privacy protection judgment viewpoint of the target digital asset evaluation report based on the difference in semantic features between the target dense text semantics and each original dense text semantics in the original dense text semantics set includes:

[0025] For each original dense text semantics in the original dense text semantics set, determining a dense text semantics difference between the target dense text semantics and the dense text mean semantics in the original dense text semantics;

[0026] Determining the privacy protection judgment viewpoint of the target digital asset evaluation report according to the normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the feature weighted result of the dense text semantic difference;

[0027] The step of determining the privacy protection judgment viewpoint of the target digital asset evaluation report according to the normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the feature weighted result of the dense text semantic difference includes:

[0028] The normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the minimum feature weighted result of the dense text semantic difference are used as the privacy protection judgment viewpoint of the target digital asset evaluation report;

[0029] If the privacy protection discrimination opinion of the target digital asset evaluation report is greater than the set discrimination threshold of the original dense text semantics corresponding to the minimum feature weighted result, the target digital asset evaluation report is determined to be a sensitive evaluation report that requires privacy protection.

[0030] Based on the same inventive concept, a report text privacy protection system is also provided, comprising:

[0031] A data acquisition module is used to acquire a target digital asset evaluation report and an original dense text semantic set; the original dense text semantics in the original dense text semantic set is a feature relationship network of privacy semantics obtained by performing dense semantic mining on a positive sample set of evaluation reports corresponding to each debugging instruction in a debugging instruction set during the debugging process of a text privacy protection decision network;

[0032] A semantic mining module, used to perform dense semantic mining processing on the target digital asset evaluation report through a target dense semantic mining branch in the text privacy protection decision network to obtain target dense text semantics of the target digital asset evaluation report;

[0033] A privacy protection module is used to determine the privacy protection judgment point of view of the target digital asset evaluation report based on the difference in semantic features between the target dense text semantics and each original dense text semantics in the original dense text semantic set; the privacy protection judgment point of view characterizes whether the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection.

[0034] Optionally, the debugging steps of the text privacy protection decision network include:

[0035] Obtaining a set of past digital asset evaluation reports corresponding to each debugging instruction in the debugging instruction set; the set of past digital asset evaluation reports includes a set of positive evaluation report samples and a set of derived evaluation report samples, wherein the derived evaluation report samples in the derived evaluation report sample set are obtained by deriving the corresponding positive evaluation report samples;

[0036] Determine the target debugging instruction in the debugging instruction set, debug the basic text privacy protection decision network through the evaluation report positive sample set of the target debugging instruction, and obtain the original text privacy protection decision network and the first dense text semantic set of the target debugging instruction in the debugging process;

[0037] Determining a feature relationship network of the first dense text semantic set as the original dense text semantics in the original dense text semantic set;

[0038] Freeze the discriminant branch in the original text privacy protection decision network, debug the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set, and obtain the target dense semantic mining branch and the second dense text semantic set of each pending debugging instruction in the debugging process; the second dense text semantic set is a dense text semantic set obtained by dense semantic mining of the positive sample set of the evaluation report of the pending debugging instruction, and the pending debugging instruction set is a plurality of debugging instructions in the debugging instruction set except the target debugging instruction;

[0039] Determine a feature relationship network of each of the second dense text semantic sets as the original dense text semantics in the original dense text semantic set, and adjust the original dense text semantic set;

[0040] Wherein, the discrimination branch and the target dense semantic mining branch constitute the text privacy protection decision network;

[0041] Among them, the debugging of the basic text privacy protection decision network through the positive sample set of the evaluation report of the target debugging instruction includes: performing dense semantic mining processing on the positive sample set of the evaluation report of the target debugging instruction through the dense semantic mining branch to be debugged in the basic text privacy protection decision network to obtain the first dense text semantic set; performing discriminant training on the positive sample set of the evaluation report of the target debugging instruction according to the first dense text semantic set through the discriminant branch to be debugged in the basic text privacy protection decision network to obtain the first discriminant training viewpoint of the positive sample set of the evaluation report of the target debugging instruction; improving the configuration weights of the dense semantic mining branch to be debugged and the discriminant branch to be debugged according to the first discriminant training viewpoint until the first set debugging termination requirement is met, and the original dense semantic mining branch and the discriminant branch are obtained;

[0042] Among them, the debugging of the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set includes: performing dense semantic mining processing on the evaluation report positive sample set and the evaluation report derived sample set of each pending debugging instruction through the original dense semantic mining branch to obtain the second dense text semantic set of the evaluation report positive sample set and the third dense text semantic set of the evaluation report derived sample set; performing discriminant training on the past digital asset evaluation report set according to the second dense text semantic set and performing discriminant training on the evaluation report derived sample set according to the third dense text semantic set through the discriminant branch, to obtain the second discriminant training viewpoint of the evaluation report positive sample set of the pending debugging instruction and the third discriminant training viewpoint of the evaluation report derived sample set; improving the configuration weight of the original dense semantic mining branch according to the second discriminant training viewpoint and the third discriminant training viewpoint until the second set debugging termination requirement is met to obtain the target dense semantic mining branch.

[0043] Optionally, the original dense text semantics includes dense text mean semantics and semantic jump features; then determining the privacy protection judgment viewpoint of the target digital asset evaluation report based on the semantic feature difference between the target dense text semantics and each original dense text semantics in the original dense text semantics set includes:

[0044] For each original dense text semantics in the original dense text semantics set, determining a dense text semantics difference between the target dense text semantics and the dense text mean semantics in the original dense text semantics;

[0045] Determining the privacy protection judgment viewpoint of the target digital asset evaluation report according to the normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the feature weighted result of the dense text semantic difference;

[0046] Among them, the privacy protection judgment point of view of the target digital asset evaluation report is determined according to the normalized result of the dense text semantic difference, the mapping variables of the semantic jump feature and the feature weighted result of the dense text semantic difference, including: taking the normalized result of the dense text semantic difference, the mapping variables of the semantic jump feature and the minimum feature weighted result of the dense text semantic difference as the privacy protection judgment point of view of the target digital asset evaluation report; if the privacy protection judgment point of view of the target digital asset evaluation report is greater than the set judgment threshold of the original dense text semantics corresponding to the minimum feature weighted result, it is determined that the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection.

[0047] According to a second aspect of an embodiment of the present application, a report text privacy protection system is provided, comprising: a processor, a memory and a bus connected to the processor; the processor and the memory communicate with each other via the bus; the processor is used to call a computer program in the memory to execute the above-mentioned report text privacy protection method for digital asset evaluation.

[0048] According to a third aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which a program is stored, and when the program is executed by a processor, the above-mentioned report text privacy protection method for digital asset evaluation is implemented.

[0049] The embodiment of the present application proposes a privacy protection method for digital asset evaluation reports based on a text privacy protection decision network. By utilizing natural language processing and deep learning technology, the embodiment of the present application can automatically mine the dense text semantics of the report and compare it with the original dense text semantic set previously mined, thereby accurately determining the sensitivity of the report. This method not only significantly improves the efficiency and accuracy of privacy protection, but also greatly reduces the cost and risk of manual review, providing strong technical support for the digital asset security of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A flowchart of a report text privacy protection method for digital asset evaluation provided in an embodiment of the present application.

[0052] Figure 2 A schematic diagram of the product modules of a report text privacy protection system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] The exemplary embodiments disclosed in the present application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0054] 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.

[0055] See also Figure 1 , which is a flowchart of a report text privacy protection method for digital asset evaluation provided in an embodiment of the present application. The method is applied to a report text privacy protection system. The specific content description of the method includes steps 110 to 130.

[0056] Step 110, obtain the target digital asset evaluation report and the original dense text semantic set; the original dense text semantics in the original dense text semantic set is a feature relationship network of privacy semantics obtained by dense semantic mining of the evaluation report positive sample set corresponding to each debugging instruction in the debugging instruction set during the debugging process of the text privacy protection decision network.

[0057] In an exemplary digital asset evaluation scenario, the report text privacy protection system first obtains a target digital asset evaluation report. This report describes in detail the digital asset status of a company, including asset size, value assessment, market analysis, etc. At the same time, the system also extracts data from the original dense text semantic set established previously. This original dense text semantic set is obtained by deeply mining the positive sample set of evaluation reports corresponding to multiple debugging instructions during the debugging process of the text privacy protection decision network. It reflects the complex feature relationship network between privacy semantics.

[0058] Step 120: Perform dense semantic mining processing on the target digital asset evaluation report through the target dense semantic mining branch in the text privacy protection decision network to obtain the target dense text semantics of the target digital asset evaluation report.

[0059] Next, the report text privacy protection system uses the target dense semantic mining branch in the text privacy protection decision network to conduct an in-depth analysis of the target digital asset evaluation report. This step involves comprehensive processing of the vocabulary, syntax, and context information in the report to extract the dense text semantics of the report. Through advanced natural language processing technology, the system accurately captures the key information and potential meanings in the report and generates the target dense text semantics of the target digital asset evaluation report.

[0060] Step 130: determine the privacy protection judgment point of view of the target digital asset evaluation report based on the difference in semantic features between the target dense text semantics and each original dense text semantics in the original dense text semantic set; the privacy protection judgment point of view characterizes whether the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection.

[0061] After obtaining the target dense text semantics, the report text privacy protection system begins to compare it with each element in the original dense text semantic set. By calculating the difference in semantic features, the system can accurately identify sensitive information that may be contained in the target digital asset evaluation report. This information may involve key content such as the company's financial status, strategic planning, customer data, etc., and once leaked, it may cause significant losses to the company.

[0062] Based on these comparison results, the system generates a privacy protection judgment point of view. This point of view clearly indicates whether the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection. If the system determines that the report contains sensitive information, it will automatically trigger further privacy protection measures, such as encryption, desensitization or access control, to ensure the security and confidentiality of digital assets.

[0063] Based on the above technical solutions, the report text privacy protection system not only improves the security and efficiency of the digital asset evaluation process, but also provides enterprises with a powerful data protection mechanism, effectively preventing the leakage and abuse of sensitive information.

[0064] In combination with the above content, the technical terms in step 110 to step 130 are adaptively explained below.

[0065] The target digital asset assessment report refers to a detailed report that requires privacy inspection and protection, and its content mainly involves the status of a company or organization's digital assets. The report can cover asset size, market valuation, development trends, potential risks and other aspects, and is an important basis for decision makers to conduct strategic planning and economic activities. In this technical solution, this report is the main object of system analysis and processing.

[0066] The original dense text semantic set is a pre-built data set that contains a large amount of dense text semantic information extracted from various evaluation reports. This information is mined from a large amount of text data using advanced text analysis technology, reflecting the deep meaning and relationships in the text. This set is used as a benchmark for comparison and analysis in the technical solution.

[0067] The original dense text semantics refers to the specific semantic information extracted from the evaluation report, which includes not only the literal meaning, but also the implicit meaning, contextual relationship and connection with other text elements in the text. This information is the basic unit that constitutes the original dense text semantic set.

[0068] The Text Privacy Protection Decision Network is a specially designed neural network system for processing and analyzing text data, especially the part involving privacy protection. The network can automatically identify sensitive information and privacy data in the text through learning and training, and make corresponding processing decisions based on preset rules and algorithms.

[0069] The debugging instruction set is a set of instructions for debugging and optimizing the text privacy protection decision network. These instructions cover multiple aspects such as network configuration, parameter adjustment, performance testing, etc., aiming to ensure that the network can run accurately and efficiently in various scenarios.

[0070] Debugging instructions are specific instructions in the debugging instruction set. Each instruction corresponds to a specific operation or setting in the network debugging process. By executing these instructions, technicians can accurately control and adjust the operation status of the network to achieve the best privacy protection effect.

[0071] The evaluation report positive sample set is a set of carefully selected evaluation report samples used to train and optimize the text privacy protection decision network. These samples can contain evaluation reports of various types and scenarios to ensure that the network can fully learn and adapt to different text features and data distributions.

[0072] Dense semantic mining is an in-depth text analysis technology that aims to extract rich and accurate semantic information from text data. By combining natural language processing, machine learning and other technical means, dense semantic mining can capture the subtle differences and complex relationships in the text, providing strong support for subsequent privacy protection work.

[0073] The privacy semantics feature relationship network is a complex network structure used to describe and represent the associations and relationships between privacy semantics in text. This network not only includes the direct connection between individual semantic elements, but also covers deeper structures and patterns, which helps the system to more comprehensively understand and process privacy information in text.

[0074] The target dense semantic mining branch is a special branch in the text privacy protection decision network, which is responsible for performing the dense semantic mining task of the target digital asset evaluation report. Through this branch, the system can conduct in-depth analysis on specific reports and extract key semantic information, laying the foundation for subsequent comparison and discrimination.

[0075] Target dense text semantics refers to the specific semantic information extracted from the target digital asset evaluation report through the target dense semantic mining branch. This information reflects the core content and potential meaning of the report and is an important basis for the system to make privacy protection judgments.

[0076] Semantic feature differences refer to the differences between the target dense text semantics and the individual elements in the original dense text semantics set. By calculating and comparing these differences, the system can accurately identify unique information and potential risk points in the target report.

[0077] The privacy protection judgment point of view is a conclusive point of view drawn by the system based on the analysis of semantic feature differences, which is used to determine whether the target digital asset evaluation report needs privacy protection. This point of view comprehensively considers multiple factors such as the sensitive information content and leakage risk in the report, and provides clear guidance for subsequent privacy protection measures.

[0078] Sensitive assessment reports that require privacy protection are those that are determined by the system to contain sensitive information and have a risk of leakage. For such reports, the system needs to take corresponding privacy protection measures to ensure that their security and confidentiality are not compromised.

[0079] In detail, in step 110, in the complex scenario of digital asset evaluation, the first task of the report text privacy protection system is to obtain key data. In this step, the system focuses on obtaining two core components: the target digital asset evaluation report and the original dense text semantic set.

[0080] The target digital asset assessment report is the core object of this privacy protection work. It records in detail the comprehensive status of a company's digital assets. From asset size to market value assessment, to in-depth market analysis, the report covers multiple aspects and provides comprehensive data support for internal and external decision makers. The system's primary task is to ensure that such a report containing a large amount of sensitive information is not leaked. The system receives this report from the company's internal server or designated data exchange platform through secure network transmission or encrypted file exchange. After receiving the report, the system performs preliminary integrity verification and format verification to ensure that the report content has not been tampered with and the format is correct, laying the foundation for subsequent analysis and processing.

[0081] At the same time, the system also accesses its built-in original dense text semantic set. This set is not created overnight, but is gradually accumulated during the long-term debugging and optimization of the text privacy protection decision network. During the debugging phase of the network, the development team will design a series of debugging instructions, each of which corresponds to a set of positive sample sets of evaluation reports. These sample sets are carefully selected from a large number of historical evaluation reports and are representative and typical.

[0082] For each sample set corresponding to the debugging instruction, the system conducts in-depth dense semantic mining. This mining process not only focuses on the surface meaning of the text, but also strives to capture the deep structure and implicit meaning in the text. Through advanced natural language processing technology and machine learning algorithms, the system can extract rich privacy semantic features from these sample sets and construct a complex feature relationship network. This feature relationship network is the core component of the original dense text semantic set. It reflects the intrinsic connection and interaction between different privacy semantics, and provides valuable reference information for subsequent privacy protection work. When the system obtains this semantic set, it will perform necessary updates and maintenance to ensure that the semantic features in the set are always up-to-date and can adapt to the ever-changing privacy protection needs.

[0083] It can be seen that step 110 is not only a simple process of obtaining data, but also a key link to lay a solid foundation for subsequent privacy protection work. By obtaining the target digital asset evaluation report and the original dense text semantic set, the report text privacy protection system has been fully prepared for the subsequent in-depth analysis.

[0084] Furthermore, in step 120, after successfully acquiring the target digital asset evaluation report and the original dense text semantic set, the report text privacy protection system enters the next key step: performing dense semantic mining on the target report. This step is accomplished with the help of the target dense semantic mining branch in the text privacy protection decision network.

[0085] First, the system inputs the target digital asset assessment report into the text privacy protection decision network. This network is specially designed and trained for handling text privacy protection tasks, and it has powerful text analysis and processing capabilities. Inside the network, there is a dedicated target dense semantic mining branch, which is responsible for extracting the deep semantic information of the report.

[0086] When the report is input into the network, the target dense semantic mining branch will start working. It will first pre-process the report, including word segmentation, stop word removal, part-of-speech tagging, etc., in order to more accurately capture the key information in the text. Then, this branch will use deep learning techniques such as convolutional neural networks (CNN) or long short-term memory networks (LSTM) to perform deep mining on the processed text.

[0087] During the mining process, the target dense semantic mining branch comprehensively considers the vocabulary, syntactic structure and contextual information in the report to extract the dense text semantics that best represents the report topic. This semantic information not only includes the surface meaning of the report, but also covers the implicit meaning and potential relationships in the text.

[0088] Through this series of mining and processing, the system can eventually generate the target dense text semantics of the target digital asset evaluation report. These semantic information are an important basis for subsequent privacy protection judgment. They will be used to compare and analyze with the original dense text semantic set to determine whether the report contains sensitive information and formulate corresponding privacy protection measures accordingly.

[0089] It is worth mentioning that the processing power and accuracy of the target dense semantic mining branch are constantly improving. With the continuous learning and optimization of the system, it can more accurately capture the key information in the text and provide more reliable support for privacy protection. This also reflects the wide application and far-reaching impact of artificial intelligence technology in the fields of text analysis and privacy protection.

[0090] Finally, in step 130, after completing the dense semantic mining of the target digital asset evaluation report and obtaining the target dense text semantics, the report text privacy protection system enters the key link of determining the sensitivity of the report. The core of this step is to compare the semantic feature differences of each element in the target dense text semantics and the original dense text semantics.

[0091] The system first calls its built-in algorithm to vectorize each element in the target dense text semantics and the original dense text semantics set, which is convenient for mathematical comparison and calculation. These vectors not only capture the surface meaning of the text, but also contain deep semantic relationships.

[0092] Then, the system calculates the similarity or difference between the target dense text semantic vector and each vector in the original dense text semantic set one by one. This process may involve the application of metrics such as cosine similarity and Euclidean distance, which aims to quantify the closeness or difference between the semantics of two texts.

[0093] After obtaining a series of similarity or difference values, the system sets one or more thresholds to determine whether the target digital asset evaluation report contains sensitive information. These thresholds may be based on historical data, expert judgment, or machine learning algorithms, and they represent the system's tolerance for sensitive information.

[0094] Next, the system compares the calculated similarity or difference values ​​with these thresholds. If a value exceeds the set sensitivity threshold, the system will consider that the target digital asset assessment report contains sensitive information that is highly similar to an element in the original dense text semantic set.

[0095] Based on this series of comparisons and judgments, the system will eventually form a privacy protection judgment point of view. This point of view is a clear conclusion that indicates whether the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection. If the system believes that the report contains sensitive information, then this point of view will trigger subsequent privacy protection measures.

[0096] These measures may include, but are not limited to, encrypting reports, desensitizing sensitive information, setting stricter access control permissions, etc. All of these measures are aimed at ensuring the security and confidentiality of digital assets and preventing the leakage and abuse of sensitive information.

[0097] Through precise judgment and effective protection in this step, the report text privacy protection system provides enterprises with a powerful data protection shield, making the digital asset evaluation process both efficient and secure.

[0098] The embodiment of the present application achieves accurate privacy protection judgment of the target report by obtaining the target digital asset evaluation report and the original dense text semantic set previously mined during the debugging process of the text privacy protection decision network. The target dense semantic mining branch in the text privacy protection decision network is used to deeply analyze the semantic content of the target report, and then obtain its dense text semantics. By comparing the target dense text semantics with the elements in the original dense text semantic set, the differences in semantic features are accurately captured, so as to accurately determine whether the target report contains sensitive information and whether privacy protection is required. This method not only improves the intelligent level of privacy protection, but also greatly enhances the security and privacy protection capabilities of digital asset evaluation reports, effectively prevents the leakage of sensitive information, and provides a solid guarantee for enterprise data security.

[0099] In some preferred embodiments, the debugging step of the text privacy protection decision network includes: obtaining a set of past digital asset evaluation reports corresponding to each debugging instruction in the debugging instruction set; the past digital asset evaluation report set includes an evaluation report positive sample set and an evaluation report derivative sample set, and the evaluation report derivative samples in the evaluation report derivative sample set are obtained by deriving the corresponding evaluation report positive samples; determining the target debugging instruction in the debugging instruction set, debugging the basic text privacy protection decision network through the evaluation report positive sample set of the target debugging instruction, and obtaining the original text privacy protection decision network and the first dense text semantic set of the target debugging instruction during the debugging process; determining the feature relationship network of the first dense text semantic set as the original dense text semantics in the original dense text semantic set; freezing the original text privacy protection decision network. The discriminant branch debugs the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set, and obtains the target dense semantic mining branch and the second dense text semantic set of each pending debugging instruction in the debugging process; the second dense text semantic set is a dense text semantic set obtained by dense semantic mining of the positive sample set of the evaluation report of the pending debugging instruction, and the pending debugging instruction set is a plurality of debugging instructions in the debugging instruction set except the target debugging instruction; determine the feature relationship network of each second dense text semantic set as the original dense text semantics in the original dense text semantic set, and adjust the original dense text semantic set; wherein the discriminant branch and the target dense semantic mining branch constitute the text privacy protection decision network.

[0100] Based on this embodiment, the report text privacy protection system executes the debugging steps of its text privacy protection decision network. This process involves multiple fine links, which together ensure that the network can accurately identify and protect the privacy information in the digital asset valuation report.

[0101] First, the system obtains the past digital asset evaluation report sets corresponding to each debugging instruction in the debugging instruction set. These report sets include evaluation report positive sample sets and evaluation report derivative sample sets. The positive sample set contains positive, standard evaluation reports, while the derivative sample set is obtained by making certain changes or modifications based on the positive samples to increase the generalization ability of the network.

[0102] Next, the system determines the target debugging instruction in the debugging instruction set and uses the evaluation report positive sample set of the instruction to debug the basic text privacy protection decision network. The purpose of this step is to allow the network to initially learn how to identify and process the privacy information in the evaluation report. After debugging is completed, the system obtains the original text privacy protection decision network and the first dense text semantic set related to this process.

[0103] Then, the system analyzes the feature relationship network of the first dense text semantic set, which is the deep structure of privacy semantics, and takes it as part of the original dense text semantic set. This step is crucial for the accuracy of subsequent privacy protection.

[0104] After completing the initial debugging, the system enters the next stage: freezing the discriminant branch in the original text privacy protection decision network. This is done to keep the learned discriminant ability unchanged during the subsequent debugging process, and focus on optimizing the semantic mining ability of the network.

[0105] Next, the system uses the past digital asset evaluation report set of the pending debugging instruction set to further debug the original dense semantic mining branch. The pending debugging instruction set here refers to the other instructions in the debugging instruction set except the target debugging instruction. In this process, the system generates a second dense text semantic set, which is the result of deep semantic mining of the positive sample set of evaluation reports for these instructions.

[0106] The system then analyzes the feature relationship network of each second dense text semantic set again and incorporates it into the original dense text semantic set, thereby continuously adjusting and improving this set. This step ensures that the network can continue to learn and improve to cope with more diverse evaluation reports and privacy situations.

[0107] Finally, after this series of meticulous debugging steps, the text privacy protection decision network of the report text privacy protection system was constructed and improved. This network consists of a discriminant branch and a target dense semantic mining branch, which can efficiently and accurately identify and process privacy information in digital asset evaluation reports.

[0108] Through this series of implementation steps, the report text privacy protection system realizes the intelligent identification and protection of privacy information in digital asset evaluation reports. This not only greatly improves the efficiency and accuracy of privacy protection, but also reduces the cost and risk of manual review. At the same time, through continuous learning and optimization, the system can better adapt to the complex and changing risks of privacy leakage, providing strong technical support for the data security of enterprises. In summary, this technical solution effectively protects the core information and asset security of enterprises through intelligent privacy protection methods.

[0109] In the next step, the basic text privacy protection decision network is debugged through the positive sample set of the evaluation report of the target debugging instruction, including: performing dense semantic mining processing on the positive sample set of the evaluation report of the target debugging instruction through the dense semantic mining branch to be debugged in the basic text privacy protection decision network to obtain the first dense text semantic set; performing discriminant training on the positive sample set of the evaluation report of the target debugging instruction according to the first dense text semantic set through the discriminant branch to be debugged in the basic text privacy protection decision network to obtain the first discriminant training point of view of the positive sample set of the evaluation report of the target debugging instruction; improving the configuration weights of the dense semantic mining branch to be debugged and the discriminant branch to be debugged according to the first discriminant training point of view until the first set debugging termination requirement is met, and the original dense semantic mining branch and the discriminant branch are obtained.

[0110] In this embodiment, the report text privacy protection system will carefully debug the basic text privacy protection decision network through the evaluation report positive sample set of the target debugging instruction. This process is a key link to ensure that the network can accurately identify and process private information.

[0111] First, the system uses the dense semantic mining branch to be debugged in the basic text privacy protection decision network to perform dense semantic mining on the evaluation report positive sample set of the target debugging instruction. This step aims to deeply analyze the semantic content in the evaluation report, extract key information, and form the first dense text semantic set. This set contains the core privacy-related semantic information in the evaluation report and is an important basis for subsequent discriminant training.

[0112] Next, the system uses the to-be-debugged discriminant branch in the basic text privacy protection decision network to perform discriminant training on the evaluation report positive sample set of the target debugging instruction based on the first dense text semantic set. In this process, the discriminant branch will learn how to accurately judge whether the evaluation report contains sensitive information based on the extracted dense text semantics, that is, to form the first discriminant training point of view. This point of view is a preliminary reflection of the discriminant branch's ability to identify privacy information.

[0113] Then, the system improves the configuration weights of the dense semantic mining branch to be debugged and the discriminative branch to be debugged based on the first discriminative training viewpoint. This is an iterative optimization process. By continuously adjusting the weights of the two branches, the network can be more accurate and efficient in identifying private information. The key to this step is to find a balance point so that the dense semantic mining branch can extract sufficiently rich semantic information, while the discriminative branch can accurately make judgments based on this information.

[0114] This process will continue until the first set debugging termination requirement is met. This requirement may include the discrimination accuracy reaching a certain threshold, the training loss being reduced to a certain level, etc., depending on the actual needs of the system. When the termination requirement is met, the system obtains the optimized original dense semantic mining branch and the discrimination branch.

[0115] Through this series of detailed debugging steps, the report text privacy protection system successfully built a text privacy protection decision network that can efficiently identify and process privacy information. This network not only has powerful semantic mining capabilities, but can also accurately judge sensitive information in evaluation reports, providing strong protection for corporate data security.

[0116] In this way, through a sophisticated debugging process, the text privacy protection decision network of the report text privacy protection system is more accurate and efficient in identifying and processing private information. This not only improves the overall performance of the system, but also provides a more solid barrier for the data security of the enterprise. At the same time, this optimization process can also be flexibly adjusted according to actual needs to adapt to the privacy protection needs in different scenarios.

[0117] It is worth mentioning that, according to the first discriminant training viewpoint, the configuration weights of the dense semantic mining branch to be debugged and the discriminant branch to be debugged are improved until the first set debugging termination requirement is met, and the original dense semantic mining branch and the discriminant branch are obtained, including:

[0118] Analyze the first discriminant training view, which reflects the accuracy and misjudgment of the discriminant branch when making privacy judgments based on the current dense text semantic set;

[0119] According to the first discriminant training viewpoint, a dynamic adjustment strategy is constructed, which can automatically and iteratively adjust the configuration weights of the dense semantic mining branch to be debugged and the discriminant branch to be debugged;

[0120] In each iteration, the weight of the dense semantic mining branch to be debugged is first adjusted to enhance its ability to extract privacy-related semantic features in the evaluation report; the specific adjustment methods include but are not limited to: increasing the attention to privacy features in misjudged samples and adjusting the parameters of the network layer to better capture these features;

[0121] Then, the weight of the discriminant branch to be debugged is adjusted to improve the accuracy of privacy discrimination based on the updated dense text semantic set; the adjustment method may include: optimizing the discrimination threshold, adjusting the parameters of the classification layer, or introducing a more complex discriminant model to improve the discrimination performance;

[0122] After each weight adjustment, the adjusted network is reused to process the evaluation report positive sample set and obtain new discriminative training points;

[0123] Compare the new discriminant training viewpoint with the previous viewpoint to evaluate whether the network performance has improved; if the performance has improved, accept the weight adjustment, otherwise roll back to the previous weight configuration;

[0124] Repeat "In each iteration, first adjust the weight of the dense semantic mining branch to be debugged to enhance its ability to extract privacy-related semantic features in the evaluation report; the specific adjustment methods include but are not limited to: increasing the attention to privacy features in misjudged examples, adjusting the parameters of the network layer to better capture these features; then adjust the weight of the discriminant branch to be debugged to improve its accuracy in privacy discrimination based on the updated dense text semantic set; the adjustment methods may include: optimizing the discrimination threshold, adjusting the parameters of the classification layer, or introducing a more complex discriminant model to improve the discrimination performance; after each weight adjustment, reuse the adjusted network to process the positive sample set of the evaluation report and obtain new discriminant training views" until the first set debugging termination requirements are met. These requirements may include but are not limited to: the discrimination accuracy reaches a preset threshold, the performance improvement for multiple consecutive iterations is less than a certain threshold, and the maximum number of iterations is reached;

[0125] When the debugging termination requirements are met, the currently optimized dense semantic mining branch to be debugged and the discriminant branch to be debugged are used as the final original dense semantic mining branch and the discriminant branch respectively.

[0126] In detail, in an embodiment of the report text privacy protection system, the detailed process of how to improve the configuration weights of the dense semantic mining branch to be debugged and the discriminative branch to be debugged according to the first discriminative training viewpoint until the first set debugging termination requirement is met, thereby obtaining the original dense semantic mining branch and the discriminative branch is as follows: First, the system deeply analyzes the first discriminative training viewpoint. This viewpoint is actually the feedback of the effect of the discriminative branch when performing privacy discrimination based on the current dense text semantic set, which reveals the accuracy and misjudgment of the discrimination. This is an important reference for subsequent optimization steps. Then, based on this viewpoint, the system constructs a dynamic adjustment strategy. The core of this strategy is that it can automatically and iteratively adjust the weight configuration of the two branches according to the feedback of the discriminative training viewpoint to seek the best performance. Entering the iterative optimization stage, the system first adjusts the weight of the dense semantic mining branch to be debugged. The purpose of the adjustment is to enhance the branch's ability to extract privacy-related semantic features in the evaluation report. For example, the system may increase the attention to privacy features in samples that were misjudged in previous discriminations, and better capture these features by adjusting the parameters of the network layer. Subsequently, the system adjusts the weight of the discriminative branch to be debugged. This step aims to improve the accuracy of privacy discrimination of the discriminant branch based on the updated dense text semantic set. To achieve this, the system may adopt a variety of adjustment methods, such as optimizing the discrimination threshold, adjusting the parameters of the classification layer, and even introducing a more complex discriminant model to improve the discrimination performance. After each weight adjustment, the system will re-use the adjusted network to process the evaluation report positive sample set and obtain new discriminant training points. This step is crucial because it provides an opportunity to verify the effect of the adjustment and provides a basis for subsequent adjustments. Next, the system compares the new discriminant training points with the previous points to evaluate whether the network performance has improved. If the performance has indeed improved, the system accepts this weight adjustment; otherwise, the system rolls back to the previous weight configuration and continues to look for a better solution. This process will be repeated until the first set debugging termination requirements are met. These requirements may include but are not limited to the discrimination accuracy reaching a preset threshold, the performance improvement of multiple consecutive iterations is less than a certain threshold, or the maximum number of iterations is reached. The setting of these conditions is designed to ensure that the optimization process is both effective and efficient. Finally, when the debugging termination requirements are met, the system determines the currently optimized dense semantic mining branch to be debugged and the discriminant branch to be debugged as the final original dense semantic mining branch and discriminant branch respectively. Through this series of sophisticated and complex debugging and optimization steps, the report text privacy protection system can build an efficient and accurate text privacy protection decision network. This not only improves the system's ability to identify and process privacy information, but also provides a more solid guarantee for the company's data security. At the same time, this dynamic adjustment and optimization strategy also provides a new solution and method for similar technical problems.

[0127] In another step, the original dense semantic mining branch in the original text privacy protection decision network is debugged through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set, including: performing dense semantic mining processing on the evaluation report positive sample set and the evaluation report derived sample set of each pending debugging instruction through the original dense semantic mining branch to obtain the second dense text semantic set of the evaluation report positive sample set and the third dense text semantic set of the evaluation report derived sample set; performing discriminant training on the past digital asset evaluation report set according to the second dense text semantic set and performing discriminant training on the evaluation report derived sample set according to the third dense text semantic set through the discriminant branch, respectively, to obtain the second discriminant training viewpoint of the evaluation report positive sample set of the pending debugging instruction and the third discriminant training viewpoint of the evaluation report derived sample set; improving the configuration weight of the original dense semantic mining branch according to the second discriminant training viewpoint and the third discriminant training viewpoint until the second set debugging termination requirement is met to obtain the target dense semantic mining branch.

[0128] Based on this embodiment, the report text privacy protection system will further debug the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set. This step is crucial because it is related to whether the system can accurately identify and process various privacy information in practical applications.

[0129] First, the system performs dense semantic mining on the evaluation report positive sample set and evaluation report derivative sample set of each pending debugging instruction through the original dense semantic mining branch. This process aims to deeply analyze the semantic content in these report sets and extract key information. After processing, the system obtains the second dense text semantic set of the evaluation report positive sample set and the third dense text semantic set of the evaluation report derivative sample set. These two sets represent the core semantic information in the positive samples and derivative samples, respectively.

[0130] Next, the system uses the discriminant branch to perform discriminant training on the past digital asset evaluation report set and the evaluation report derived sample set based on the second dense text semantic set and the third dense text semantic set, respectively. The purpose of this step is to train the discriminant branch so that it can accurately identify private information. Through training, the system obtains the second discriminant training viewpoint of the evaluation report positive sample set of pending debugging instructions, and the third discriminant training viewpoint of the evaluation report derived sample set. These two viewpoints reflect the discriminant performance of the discriminant branch on different sample sets.

[0131] Then, the system improves the configuration weights of the original dense semantic mining branch based on the second and third discriminant training viewpoints. This is an iterative optimization process, and the system continuously adjusts the weights to improve the dense semantic mining branch's ability to extract private information. During the optimization process, the system pays close attention to the feedback of the discriminant branch to ensure that the extracted semantic information can effectively support the discriminant work.

[0132] This process will continue until the second set debugging termination requirements are met. These requirements may include the discrimination accuracy reaching a certain predetermined threshold, the training loss decreasing to a certain level, or reaching the maximum number of iterations. When the termination conditions are met, the system obtains an optimized target dense semantic mining branch.

[0133] Through this series of detailed debugging steps, the report text privacy protection system successfully improved the dense semantic mining branch's ability to extract privacy information. This not only enhances the overall performance of the system, but also provides more reliable privacy protection for the company's digital asset evaluation reports. At the same time, this optimization method also has a certain degree of versatility and can provide a useful reference for other similar natural language processing tasks.

[0134] In this way, through the careful debugging and optimization process, the dense semantic mining branch in the report text privacy protection system is more accurate and efficient in extracting privacy information. This not only improves the practicality of the system, but also provides strong support for the data security and privacy protection of enterprises.

[0135] It is worth mentioning that the configuration weight of the original dense semantic mining branch is improved according to the second discriminant training viewpoint and the third discriminant training viewpoint until the second set debugging termination requirement is met to obtain the target dense semantic mining branch, including:

[0136] Comprehensively analyze the second and third discrimination training viewpoints to identify the weak links and misjudgments in the process of privacy information extraction and discrimination, so as to determine the semantic features that need to be optimized;

[0137] Construct a dynamic feedback-based weight adjustment mechanism that can fine-tune the network layer parameters in the original dense semantic mining branch according to the real-time feedback of the second and third discriminant training views;

[0138] In each iteration, firstly, in view of the deficiencies reflected in the second discriminant training viewpoint, the feature extraction weights corresponding to the positive sample set in the original dense semantic mining branch are adjusted to enhance the ability to extract privacy features that are misjudged or missed;

[0139] At the same time, in response to the problems reflected in the third discriminant training viewpoint, the weight of the part of the original dense semantic mining branch that processes the evaluation report derived sample set is adjusted to improve the recognition accuracy of the unique privacy features in the derived samples;

[0140] After each weight adjustment, the dense semantic mining branch is re-run to generate a new second dense text semantic set and a third dense text semantic set, and the updated second discriminant training viewpoint and the third discriminant training viewpoint are obtained through the discriminant branch;

[0141] Compare the discriminant training views before and after the update, evaluate the adjustment effect, and if the discriminant performance is improved, keep the weight adjustment, otherwise cancel the adjustment and try other optimization solutions;

[0142] Repeat "In each iteration, first adjust the feature extraction weights corresponding to the positive sample set of the evaluation report in the original dense semantic mining branch in response to the deficiencies reflected in the second discriminant training viewpoint, and strengthen the ability to extract privacy features that are misjudged or missed; at the same time, adjust the weights of the part of the original dense semantic mining branch that processes the derived sample set of the evaluation report in response to the problems reflected in the third discriminant training viewpoint, so as to improve the recognition accuracy of the unique privacy features in the derived samples; after each weight adjustment, re-run the dense semantic mining branch to generate a new second dense text semantic set and a third dense text semantic set, and obtain the updated second discriminant training viewpoint and the third discriminant training viewpoint through the discriminant branch" until the second set debugging termination requirements are met, which include but are not limited to: the discrimination accuracy reaches the preset high standard, the performance improvement for multiple consecutive iterations is less than the preset threshold, and the preset maximum number of iterations is reached;

[0143] When the debugging process meets the termination requirements, the currently adjusted dense semantic mining branch is saved as the target dense semantic mining branch for subsequent privacy protection decisions.

[0144] In detail, in the implementation process of the report text privacy protection system, the steps of improving the configuration weight of the original dense semantic mining branch according to the second discriminant training viewpoint and the third discriminant training viewpoint until the second set debugging termination requirement is met, thereby obtaining the target dense semantic mining branch, is a key link in system optimization and performance improvement. First, the system will comprehensively analyze the second discriminant training viewpoint and the third discriminant training viewpoint. These two viewpoints reflect the accuracy and misjudgment of the discriminant branch when processing the evaluation report positive sample set and the evaluation report derived sample set. Through analysis, the system can identify the weak links and misjudgments in the process of privacy information extraction and discrimination, and then determine which semantic features need to be further optimized. Next, the system will build a weight adjustment mechanism based on dynamic feedback. This mechanism can receive feedback from the second discriminant training viewpoint and the third discriminant training viewpoint in real time, and fine-tune the network layer parameters in the original dense semantic mining branch based on these feedbacks. This dynamic adjustment strategy helps the system gradually approach the optimal configuration in continuous iterations. In each iteration, the system will first adjust the deficiencies reflected in the second discriminant training viewpoint. For example, the system adjusts the feature extraction weights corresponding to the evaluation report positive sample set in the original dense semantic mining branch to enhance the ability to extract privacy features that are misjudged or missed. This helps reduce misjudgments in the subsequent discrimination process. At the same time, the system also makes adjustments to the problems reflected in the third discriminant training viewpoint. The system adjusts the weights of the part of the original dense semantic mining branch that processes the evaluation report derived sample set to improve the recognition accuracy of the unique privacy features in the derived samples. This helps enhance the system's generalization ability for different types of reports. After each weight adjustment, the system reruns the dense semantic mining branch to generate a new second dense text semantic set and a third dense text semantic set. Then, the updated second discriminant training viewpoint and the third discriminant training viewpoint are obtained through the discriminant branch. These updated views will provide important references for the next round of weight adjustments. The system compares the discriminant training viewpoints before and after the update to evaluate the adjustment effect. If the discrimination performance is improved, the system will retain this weight adjustment; otherwise, the system will cancel this adjustment and try other optimization solutions. This flexible adjustment strategy helps the system find the best configuration during exploration. This process is repeated until the second set debugging termination requirements are met. These requirements may include, but are not limited to: the discrimination accuracy reaches a preset high standard, the performance improvement for multiple consecutive iterations is less than a preset threshold, and the preset maximum number of iterations is reached. These conditions are set to ensure that the optimization process is both effective and efficient. Finally, when the debugging process meets the termination requirements, the system saves the currently adjusted dense semantic mining branch as the target dense semantic mining branch. This optimized branch will be used for subsequent privacy protection decisions, providing more reliable privacy protection for the company's digital asset evaluation report.Through this sophisticated and complex debugging and optimization steps, the report text privacy protection system can significantly improve the accuracy and efficiency of the dense semantic mining branch in extracting privacy information. This not only enhances the overall performance of the system, but also provides a more solid barrier for enterprise data security and privacy protection. At the same time, this optimization strategy based on dynamic feedback also provides a new solution and method for similar technical problems.

[0145] Under some optional design ideas, the original dense text semantics includes dense text mean semantics and semantic jump features; the privacy protection judgment point of view of the target digital asset valuation report is determined based on the semantic feature difference between the target dense text semantics and each original dense text semantics in the original dense text semantic set, including: for each original dense text semantic in the original dense text semantic set, determining the dense text semantic difference between the target dense text semantics and the dense text mean semantics in the original dense text semantics; determining the privacy protection judgment point of view of the target digital asset valuation report according to the normalized result of the dense text semantic difference, the mapping variables of the semantic jump features and the feature weighted result of the dense text semantic difference.

[0146] Based on this design idea, the report text privacy protection system will further refine the processing of dense text semantics. Here, the original dense text semantics not only includes conventional semantic information, but also specially considers the dense text mean semantics and semantic jump characteristics. The introduction of these two aspects enables the system to capture the privacy information in the text more comprehensively and accurately.

[0147] First, the system calculates the difference between the target dense text semantics and the dense text mean semantics of each original dense text semantics in the original dense text semantics set. This difference is called the dense text semantic difference. The purpose of this step is to quantify the degree of deviation between the target text and the original text in terms of overall semantics.

[0148] Next, the system will normalize the dense text semantic difference. Normalization is a commonly used data processing method that can limit data to a certain range and eliminate the impact of the data dimension on the analysis results. Here, the dense text semantic difference after normalization can more intuitively reflect the semantic similarity or difference between the target text and the original text.

[0149] In addition, the system also considers semantic jump features. Semantic jump features refer to sudden changes in semantics in text, which may hide important privacy information. The system captures this semantic jump through specific mapping variables so that it can be used in the subsequent discrimination process.

[0150] Finally, the system will comprehensively determine the privacy protection judgment point of the target digital asset evaluation report based on the normalized dense text semantic difference, the mapping variables of the semantic jump feature, and the feature weighting results of the dense text semantic difference. This judgment point is the system's comprehensive judgment on whether the target text has a privacy leakage risk, which will provide an important decision-making basis for subsequent privacy protection measures.

[0151] By introducing dense text mean semantics and semantic jump features, and combining normalization processing and feature weighting methods, the report text privacy protection system can more accurately identify the privacy information in the digital asset evaluation report, thereby providing more effective privacy protection for enterprises. This refined processing method not only improves the system's judgment accuracy, but also enhances the system's adaptability to different types of text, providing a more solid guarantee for the company's data security and privacy protection.

[0152] It is worth mentioning that according to the normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the feature weighted result of the dense text semantic difference, the privacy protection judgment point of view of the target digital asset evaluation report is determined, including: constructing a multi-dimensional discrimination model, which first receives the normalized result of the dense text semantic difference as input, and converts it into a numerical value representing the overall semantic difference through a preset algorithm; at the same time, the mapping variable of the semantic jump feature is introduced into the model as a dynamic adjustment factor to correct the discrimination bias caused by the semantic mutation in the text; then, the feature weighted result of the dense text semantic difference is used to fine-tune the model to ensure that the importance of key semantic features in the discrimination process is fully reflected; finally, the processing results of the above three aspects are comprehensively calculated, and a quantitative privacy protection judgment point of view is output according to the preset discrimination logic, which directly reflects the comprehensive evaluation results of the target digital asset evaluation report in terms of privacy leakage risk.

[0153] In detail, it is worth mentioning that the report text privacy protection system adopts a highly complex and sophisticated multi-dimensional discriminant model when determining the privacy protection discriminant view of the target digital asset evaluation report. This model not only considers the normalized results of dense text semantic differences, but also introduces the mapping variables of semantic jump features and the feature weighted results of dense text semantic differences, so as to more comprehensively and accurately assess the privacy risks of the report. For example, this multi-dimensional discriminant model first receives the normalized results of dense text semantic differences as input. This normalized result is obtained by calculating the difference between the target dense text semantics and the original dense text semantics and performing normalization processing, which reflects the degree of difference between the two in overall semantics. The model will use a preset algorithm to convert this normalized result into a specific value, which represents the difference in overall semantics. At the same time, the model will also consider the mapping variables of semantic jump features. Semantic jump features refer to sudden semantic changes in text, which may be caused by the insertion or replacement of certain sensitive information. In order to capture this semantic jump, the system will calculate a mapping variable that can dynamically reflect the stability of the semantics in the text. The model will introduce this mapping variable as a dynamic adjustment factor to correct the discrimination bias caused by the sudden change of semantics in the text. In addition, the model will use the feature weighting results of the dense text semantic difference to fine-tune the discrimination process. Feature weighting is a commonly used data processing method that can assign different weights to features according to their importance. Here, the system will assign higher weights to the key features in the dense text semantic difference based on historical data and actual needs to ensure that the importance of these features in the discrimination process is fully reflected. Finally, the model will perform a comprehensive calculation on the processing results of the above three aspects. This comprehensive calculation not only considers the overall semantic difference and semantic stability, but also the importance of key features. Through the preset discrimination logic, the model will output a quantitative privacy protection discrimination point of view. This discrimination point of view is a specific value or level, which directly reflects the comprehensive assessment results of the target digital asset assessment report in terms of privacy leakage risk. Through this multi-dimensional and refined discrimination model, the report text privacy protection system can more accurately identify privacy risks in digital asset evaluation reports, thereby providing enterprises with more targeted privacy protection measures. This discrimination method not only improves the accuracy and efficiency of privacy protection, but also provides a more solid guarantee for the data security and privacy protection of enterprises. At the same time, this innovative technical solution also demonstrates the broad application prospects of artificial intelligence technology in the field of privacy protection.

[0154] Under some preferred design ideas, the privacy protection judgment point of view of the target digital asset evaluation report is determined according to the normalized result of the dense text semantic difference, the mapping variables of the semantic jump feature and the feature weighted result of the dense text semantic difference, including: taking the normalized result of the dense text semantic difference, the mapping variables of the semantic jump feature and the minimum feature weighted result of the dense text semantic difference as the privacy protection judgment point of view of the target digital asset evaluation report; if the privacy protection judgment point of view of the target digital asset evaluation report is greater than the set judgment threshold of the original dense text semantics corresponding to the minimum feature weighted result, the target digital asset evaluation report is determined to be a sensitive evaluation report that requires privacy protection.

[0155] Based on this design idea, the report text privacy protection system will comprehensively consider multiple factors and use specific calculation methods to draw conclusions when determining the privacy protection judgment point of the target digital asset valuation report.

[0156] First, the system calculates the dense text semantic difference between the target dense text semantics and each original dense text semantics in the original dense text semantics set, and performs normalization. This normalization result reflects the similarity or difference between the target text and the original text in terms of overall semantics.

[0157] Next, the system considers semantic jump features. Semantic jump features refer to sudden changes in semantics in text, which may hide important private information. The system quantifies this semantic jump through specific mapping variables so that it can be used in the subsequent discrimination process.

[0158] In addition, the system also performs feature weighting on dense text semantic differences to highlight the impact of certain important features on the discrimination results. Here, the system pays special attention to the minimum feature weighting result of dense text semantic differences, because it may represent the minimum difference between the target text and the original text in key semantic features.

[0159] Then, the system combines the normalized results of dense text semantic differences, the mapped variables of semantic jump features, and the minimum feature weighted results of dense text semantic differences as the privacy protection judgment point of the target digital asset evaluation report. This judgment point is a comprehensive indicator used to evaluate whether the target text has the risk of privacy leakage.

[0160] Finally, the system will compare this privacy protection judgment opinion with the set judgment threshold. If the judgment opinion is greater than the set judgment threshold, the system will determine that the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection. This set judgment threshold is set based on historical data and actual needs to balance the sensitivity and accuracy of privacy protection.

[0161] By comprehensively considering multiple factors and adopting a specific calculation method, the report text privacy protection system can more accurately identify sensitive assessment reports that require privacy protection. This optimized discrimination method not only improves the system's discrimination accuracy, but also enhances the system's adaptability to different types of text, providing more reliable privacy protection for the company's digital asset assessment reports. At the same time, this discrimination method also helps reduce misjudgments and missed judgments, thereby improving the efficiency and effectiveness of privacy protection.

[0162] Please refer to Figure 2 The embodiment of the present application also provides a report text privacy protection system 100, including a processor 111, and a memory 112 and a bus 113 connected to the processor 111. The processor 111 and the memory 112 communicate with each other through the bus 113. The processor 111 is used to call the program instructions in the memory 112 to execute the above-mentioned report text privacy protection method for digital asset evaluation.

[0163] Based on the same inventive concept, a report text privacy protection system is also provided, comprising:

[0164] A data acquisition module is used to acquire a target digital asset evaluation report and an original dense text semantic set; the original dense text semantics in the original dense text semantic set is a feature relationship network of privacy semantics obtained by performing dense semantic mining on a positive sample set of evaluation reports corresponding to each debugging instruction in a debugging instruction set during the debugging process of a text privacy protection decision network;

[0165] A semantic mining module, used to perform dense semantic mining processing on the target digital asset evaluation report through a target dense semantic mining branch in the text privacy protection decision network to obtain target dense text semantics of the target digital asset evaluation report;

[0166] A privacy protection module is used to determine the privacy protection judgment point of view of the target digital asset evaluation report based on the difference in semantic features between the target dense text semantics and each original dense text semantics in the original dense text semantic set; the privacy protection judgment point of view characterizes whether the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection.

[0167] Optionally, the debugging steps of the text privacy protection decision network include:

[0168] Obtaining a set of past digital asset evaluation reports corresponding to each debugging instruction in the debugging instruction set; the set of past digital asset evaluation reports includes a set of positive evaluation report samples and a set of derived evaluation report samples, wherein the derived evaluation report samples in the derived evaluation report sample set are obtained by deriving the corresponding positive evaluation report samples;

[0169] Determine the target debugging instruction in the debugging instruction set, debug the basic text privacy protection decision network through the evaluation report positive sample set of the target debugging instruction, and obtain the original text privacy protection decision network and the first dense text semantic set of the target debugging instruction in the debugging process;

[0170] Determining a feature relationship network of the first dense text semantic set as the original dense text semantics in the original dense text semantic set;

[0171] Freeze the discriminant branch in the original text privacy protection decision network, debug the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set, and obtain the target dense semantic mining branch and the second dense text semantic set of each pending debugging instruction in the debugging process; the second dense text semantic set is a dense text semantic set obtained by dense semantic mining of the positive sample set of the evaluation report of the pending debugging instruction, and the pending debugging instruction set is a plurality of debugging instructions in the debugging instruction set except the target debugging instruction;

[0172] Determine a feature relationship network of each of the second dense text semantic sets as the original dense text semantics in the original dense text semantic set, and adjust the original dense text semantic set;

[0173] Wherein, the discrimination branch and the target dense semantic mining branch constitute the text privacy protection decision network;

[0174] Among them, the debugging of the basic text privacy protection decision network through the positive sample set of the evaluation report of the target debugging instruction includes: performing dense semantic mining processing on the positive sample set of the evaluation report of the target debugging instruction through the dense semantic mining branch to be debugged in the basic text privacy protection decision network to obtain the first dense text semantic set; performing discriminant training on the positive sample set of the evaluation report of the target debugging instruction according to the first dense text semantic set through the discriminant branch to be debugged in the basic text privacy protection decision network to obtain the first discriminant training viewpoint of the positive sample set of the evaluation report of the target debugging instruction; improving the configuration weights of the dense semantic mining branch to be debugged and the discriminant branch to be debugged according to the first discriminant training viewpoint until the first set debugging termination requirement is met, and the original dense semantic mining branch and the discriminant branch are obtained;

[0175] Among them, the debugging of the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set includes: performing dense semantic mining processing on the evaluation report positive sample set and the evaluation report derived sample set of each pending debugging instruction through the original dense semantic mining branch to obtain the second dense text semantic set of the evaluation report positive sample set and the third dense text semantic set of the evaluation report derived sample set; performing discriminant training on the past digital asset evaluation report set according to the second dense text semantic set and performing discriminant training on the evaluation report derived sample set according to the third dense text semantic set through the discriminant branch, to obtain the second discriminant training viewpoint of the evaluation report positive sample set of the pending debugging instruction and the third discriminant training viewpoint of the evaluation report derived sample set; improving the configuration weight of the original dense semantic mining branch according to the second discriminant training viewpoint and the third discriminant training viewpoint until the second set debugging termination requirement is met to obtain the target dense semantic mining branch.

[0176] Optionally, the original dense text semantics includes dense text mean semantics and semantic jump features; then determining the privacy protection judgment viewpoint of the target digital asset evaluation report based on the semantic feature difference between the target dense text semantics and each original dense text semantics in the original dense text semantics set includes:

[0177] For each original dense text semantics in the original dense text semantics set, determining a dense text semantics difference between the target dense text semantics and the dense text mean semantics in the original dense text semantics;

[0178] Determining the privacy protection judgment viewpoint of the target digital asset evaluation report according to the normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the feature weighted result of the dense text semantic difference;

[0179] Among them, the privacy protection judgment point of view of the target digital asset evaluation report is determined according to the normalized result of the dense text semantic difference, the mapping variables of the semantic jump feature and the feature weighted result of the dense text semantic difference, including: taking the normalized result of the dense text semantic difference, the mapping variables of the semantic jump feature and the minimum feature weighted result of the dense text semantic difference as the privacy protection judgment point of view of the target digital asset evaluation report; if the privacy protection judgment point of view of the target digital asset evaluation report is greater than the set judgment threshold of the original dense text semantics corresponding to the minimum feature weighted result, it is determined that the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection.

[0180] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity or system including the element.

[0181] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0182] 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 changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A report text privacy protection method for digital asset evaluation, characterized in that: Applied to a report text privacy protection system, the method at least comprises: Obtain a target digital asset evaluation report and an original dense text semantic set; the original dense text semantics in the original dense text semantic set is a feature relationship network of privacy semantics obtained by performing dense semantic mining on a positive sample set of evaluation reports corresponding to each debugging instruction in a debugging instruction set during the debugging process of a text privacy protection decision network; Performing dense semantic mining processing on the target digital asset evaluation report through the target dense semantic mining branch in the text privacy protection decision network to obtain the target dense text semantics of the target digital asset evaluation report; Determine the privacy protection judgment point of view of the target digital asset evaluation report according to the difference in semantic features between the target dense text semantics and each original dense text semantics in the original dense text semantics set; the privacy protection judgment point of view indicates whether the target digital asset evaluation report is a sensitive evaluation report requiring privacy protection; The debugging steps of the text privacy protection decision network include: Obtaining a set of past digital asset evaluation reports corresponding to each debugging instruction in the debugging instruction set; the set of past digital asset evaluation reports includes a set of positive evaluation report samples and a set of derived evaluation report samples, wherein the derived evaluation report samples in the derived evaluation report sample set are obtained by deriving the corresponding positive evaluation report samples; Determine the target debugging instruction in the debugging instruction set, debug the basic text privacy protection decision network through the evaluation report positive sample set of the target debugging instruction, and obtain the original text privacy protection decision network and the first dense text semantic set of the target debugging instruction in the debugging process; Determining a feature relationship network of the first dense text semantic set as the original dense text semantics in the original dense text semantic set; Freeze the discriminant branch in the original text privacy protection decision network, debug the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set, and obtain the target dense semantic mining branch and the second dense text semantic set of each pending debugging instruction in the debugging process; the second dense text semantic set is a dense text semantic set obtained by dense semantic mining of the positive sample set of the evaluation report of the pending debugging instruction, and the pending debugging instruction set is a plurality of debugging instructions in the debugging instruction set except the target debugging instruction; Determine a feature relationship network of each of the second dense text semantic sets as the original dense text semantics in the original dense text semantic set, and adjust the original dense text semantic set; Among them, the discrimination branch and the target dense semantic mining branch constitute the text privacy protection decision network.

2. The method according to claim 1, characterized in that: The debugging of the basic text privacy protection decision network through the evaluation report positive sample set of the target debugging instruction includes: Performing dense semantic mining processing on the evaluation report positive sample set of the target debugging instruction through the to-be-debugged dense semantic mining branch in the basic text privacy protection decision network to obtain the first dense text semantic set; Performing discriminative training on the evaluation report positive sample set of the target debugging instruction according to the first dense text semantic set through the to-be-debugged discriminative branch in the basic text privacy protection decision network, thereby obtaining a first discriminative training viewpoint on the evaluation report positive sample set of the target debugging instruction; The configuration weights of the dense semantic mining branch to be debugged and the discriminant branch to be debugged are improved according to the first discriminant training viewpoint until the first set debugging termination requirement is met, thereby obtaining the original dense semantic mining branch and the discriminant branch.

3. The method according to claim 1, characterized in that The debugging of the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set includes: Performing dense semantic mining processing on the evaluation report positive sample set and the evaluation report derived sample set of each of the debugging instructions to be processed through the original dense semantic mining branch to obtain a second dense text semantic set of the evaluation report positive sample set and a third dense text semantic set of the evaluation report derived sample set; The discriminant branch performs discriminant training on the past digital asset evaluation report set according to the second dense text semantic set and performs discriminant training on the evaluation report derived sample set according to the third dense text semantic set, thereby obtaining a second discriminant training viewpoint of the evaluation report positive sample set of the pending debugging instruction and a third discriminant training viewpoint of the evaluation report derived sample set; The configuration weight of the original dense semantic mining branch is improved according to the second discriminant training viewpoint and the third discriminant training viewpoint until the second set debugging termination requirement is met to obtain the target dense semantic mining branch.

4. The method according to any one of claims 1 to 3, characterized in that: The original dense text semantics includes dense text mean semantics and semantic jump features; The determining of the privacy protection judgment viewpoint of the target digital asset evaluation report based on the difference in semantic features between the target dense text semantics and each original dense text semantics in the original dense text semantics set includes: For each original dense text semantics in the original dense text semantics set, determining a dense text semantics difference between the target dense text semantics and the dense text mean semantics in the original dense text semantics; Determining the privacy protection judgment viewpoint of the target digital asset evaluation report according to the normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the feature weighted result of the dense text semantic difference; The step of determining the privacy protection judgment viewpoint of the target digital asset evaluation report according to the normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the feature weighted result of the dense text semantic difference includes: The normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the minimum feature weighted result of the dense text semantic difference are used as the privacy protection judgment viewpoint of the target digital asset evaluation report; If the privacy protection discrimination opinion of the target digital asset evaluation report is greater than the set discrimination threshold of the original dense text semantics corresponding to the minimum feature weighted result, the target digital asset evaluation report is determined to be a sensitive evaluation report that requires privacy protection.

5. A report text privacy protection system, characterized in that: include: A data acquisition module is used to acquire a target digital asset evaluation report and an original dense text semantic set; the original dense text semantics in the original dense text semantic set is a feature relationship network of privacy semantics obtained by performing dense semantic mining on a positive sample set of evaluation reports corresponding to each debugging instruction in a debugging instruction set during the debugging process of a text privacy protection decision network; A semantic mining module, used to perform dense semantic mining processing on the target digital asset evaluation report through a target dense semantic mining branch in the text privacy protection decision network to obtain target dense text semantics of the target digital asset evaluation report; A privacy protection module, used to determine the privacy protection judgment point of view of the target digital asset evaluation report according to the difference in semantic features between the target dense text semantics and each original dense text semantics in the original dense text semantics set; the privacy protection judgment point of view indicates whether the target digital asset evaluation report is a sensitive evaluation report that needs privacy protection; The debugging steps of the text privacy protection decision network include: Obtaining a set of past digital asset evaluation reports corresponding to each debugging instruction in the debugging instruction set; the set of past digital asset evaluation reports includes a set of positive evaluation report samples and a set of derived evaluation report samples, wherein the derived evaluation report samples in the derived evaluation report sample set are obtained by deriving the corresponding positive evaluation report samples; Determine the target debugging instruction in the debugging instruction set, debug the basic text privacy protection decision network through the evaluation report positive sample set of the target debugging instruction, and obtain the original text privacy protection decision network and the first dense text semantic set of the target debugging instruction in the debugging process; Determining a feature relationship network of the first dense text semantic set as the original dense text semantics in the original dense text semantic set; Freeze the discriminant branch in the original text privacy protection decision network, debug the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set, and obtain the target dense semantic mining branch and the second dense text semantic set of each pending debugging instruction in the debugging process; the second dense text semantic set is a dense text semantic set obtained by dense semantic mining of the positive sample set of the evaluation report of the pending debugging instruction, and the pending debugging instruction set is a plurality of debugging instructions in the debugging instruction set except the target debugging instruction; Determine a feature relationship network of each of the second dense text semantic sets as the original dense text semantics in the original dense text semantic set, and adjust the original dense text semantic set; Wherein, the discrimination branch and the target dense semantic mining branch constitute the text privacy protection decision network; Among them, the debugging of the basic text privacy protection decision network through the positive sample set of the evaluation report of the target debugging instruction includes: performing dense semantic mining processing on the positive sample set of the evaluation report of the target debugging instruction through the dense semantic mining branch to be debugged in the basic text privacy protection decision network to obtain the first dense text semantic set; performing discriminant training on the positive sample set of the evaluation report of the target debugging instruction according to the first dense text semantic set through the discriminant branch to be debugged in the basic text privacy protection decision network to obtain the first discriminant training viewpoint of the positive sample set of the evaluation report of the target debugging instruction; improving the configuration weights of the dense semantic mining branch to be debugged and the discriminant branch to be debugged according to the first discriminant training viewpoint until the first set debugging termination requirement is met, and the original dense semantic mining branch and the discriminant branch are obtained; Among them, the debugging of the original dense semantic mining branch in the original text privacy protection decision network through the past digital asset evaluation report set of each pending debugging instruction in the pending debugging instruction set includes: performing dense semantic mining processing on the evaluation report positive sample set and the evaluation report derived sample set of each pending debugging instruction through the original dense semantic mining branch to obtain the second dense text semantic set of the evaluation report positive sample set and the third dense text semantic set of the evaluation report derived sample set; performing discriminant training on the past digital asset evaluation report set according to the second dense text semantic set and performing discriminant training on the evaluation report derived sample set according to the third dense text semantic set through the discriminant branch, to obtain the second discriminant training viewpoint of the evaluation report positive sample set of the pending debugging instruction and the third discriminant training viewpoint of the evaluation report derived sample set; improving the configuration weight of the original dense semantic mining branch according to the second discriminant training viewpoint and the third discriminant training viewpoint until the second set debugging termination requirement is met to obtain the target dense semantic mining branch.

6. The report text privacy protection system according to claim 5, characterized in that: The original dense text semantics includes dense text mean semantics and semantic jump features; then determining the privacy protection judgment viewpoint of the target digital asset evaluation report based on the semantic feature difference between the target dense text semantics and each original dense text semantic in the original dense text semantic set includes: For each original dense text semantics in the original dense text semantics set, determining a dense text semantics difference between the target dense text semantics and the dense text mean semantics in the original dense text semantics; Determining the privacy protection judgment viewpoint of the target digital asset evaluation report according to the normalized result of the dense text semantic difference, the mapping variable of the semantic jump feature and the feature weighted result of the dense text semantic difference; Among them, the privacy protection judgment point of view of the target digital asset evaluation report is determined according to the normalized result of the dense text semantic difference, the mapping variables of the semantic jump feature and the feature weighted result of the dense text semantic difference, including: taking the normalized result of the dense text semantic difference, the mapping variables of the semantic jump feature and the minimum feature weighted result of the dense text semantic difference as the privacy protection judgment point of view of the target digital asset evaluation report; if the privacy protection judgment point of view of the target digital asset evaluation report is greater than the set judgment threshold of the original dense text semantics corresponding to the minimum feature weighted result, it is determined that the target digital asset evaluation report is a sensitive evaluation report that requires privacy protection.

7. A report text privacy protection system, characterized in that: It comprises a processor, a memory and a bus connected to the processor; the processor and the memory communicate with each other via the bus; The processor is used to call the computer program in the memory to execute the report text privacy protection method for digital asset evaluation as described in any one of claims 1-4.

8. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, the report text privacy protection method for digital asset evaluation described in any one of claims 1-4 is implemented.

Images