Change scheme pre-auditing method and system based on large operation and maintenance model

Abstract

The invention relates to the technical field of operation and maintenance management intellectualization, in particular to a change scheme pre-auditing method and system based on an operation and maintenance large model, and the method comprises the following steps: document analysis, chapter identification and extraction; the method has the beneficial effects that the pre-auditing efficiency of the change scheme is improved through an intelligent means, the auditing period is shortened, and the change implementation process is accelerated; the completeness and standardization of the change scheme are ensured, and the implementation risk caused by the incomplete or non-standard scheme is reduced; automatic auditing is combined with manual auditing, human errors are effectively reduced, and auditing quality is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent operation and maintenance management, and in particular to a method and system for pre-examination of change plans based on a large operation and maintenance model. Background Art

[0002] In the information system operation and maintenance of modern enterprises, change operations are frequent and complex, and the quality of the change plan is directly related to the stability of the system and the continuity of business.

[0003] Traditional change plan pre-review relies primarily on manual experience, resulting in low efficiency, inconsistent review standards, and the tendency to miss key details. With the development of big data and artificial intelligence technologies, big operation and maintenance models have emerged. These models, with their powerful data processing and pattern recognition capabilities, offer new insights and methods for change plan pre-review.

[0004] Therefore, it is necessary to propose a change plan pre-examination method and system based on the operation and maintenance big model, utilize the advantages of the operation and maintenance big model, break through the technical limitations of the traditional audit model, solve the problems existing in the traditional change plan pre-examination, and establish a set of efficient, accurate and standardized change plan pre-examination methods to ensure the safety and reliability of change operations and improve the overall level of operation and maintenance work. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for pre-examination of change plans based on an operation and maintenance big model, so as to solve the problems raised in the above background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a method for pre-examination of change plans based on an operation and maintenance large model, comprising the following steps:

[0007] Document preprocessing: Preprocess the change plan document to remove irrelevant information such as headers, footers, and blank characters. Unify the document format, unify all headings to the same format, and unify all paragraphs to the same indentation and line spacing.

[0008] Document parsing: For change plan documents in different formats, use the corresponding parsing library to extract the document content into plain text, while retaining the title, paragraph, and list structure information in the document;

[0009] Chapter identification and extraction: Utilize the large operation and maintenance model to learn the chapter structure features in documents. Train the model with a large amount of annotated data so that the model can automatically identify and extract the chapter structure of unannotated documents. The extracted chapters are manually checked and automatically verified to ensure their accuracy and completeness.

[0010] Preferably, the process of using the operation and maintenance large model to learn chapter structure features in a document and automatically identify and extract the chapter structure includes:

[0011] Model training: Use a large number of documents with manually labeled chapter structures as labeled data to train the operation and maintenance large model, enabling the model to learn the characteristics of chapters in the documents, such as the lexical characteristics of titles, the semantic characteristics of paragraphs, and the structural characteristics of lists;

[0012] Model optimization: During the model training process, adopt stochastic gradient descent and Adam optimization algorithms, use dropout and weight decay regularization techniques to prevent the model from overfitting, and improve the model performance by adjusting hyperparameters such as the learning rate, batch size, and number of hidden layer units;

[0013] Chapter prediction: Use the trained model to predict the chapter structure of unlabeled documents, realizing the automatic recognition and extraction of chapters.

[0014] Preferably, it also includes a change level judgment step:

[0015] Model learning: The operation and maintenance large model learns the specification requirements, risk level classification criteria, and approval process information corresponding to different types of change operations by learning a large number of historical change cases and change specification documents;

[0016] Level judgment: Based on the scope of influence, urgency, and complexity, train the change specification data through a supervised learning algorithm to establish a change level judgment model. The input of the model is the key feature information of the change plan, and the output is the change level and the corresponding confidence level;

[0017] Comprehensive evaluation: Combine the specific content and context information of the change plan to comprehensively evaluate the change level output by the model to ensure the accuracy of the change level judgment.

[0018] Preferably, the plan review step includes:

[0019] Process and standard establishment: Establish clear review processes and standards, formulate detailed review points and scoring criteria, decompose the review points into multiple scoring dimensions, such as content integrity, professional term standardization, and solution rationality, and design corresponding prompt words for each dimension to guide the large model to evaluate;

[0020] Model learning: The operation and maintenance large model learns the rules and standards of change management by analyzing a large number of historical change data, specification documents, and actual cases, extracts key features from the data, and uses supervised learning or semi-supervised learning methods to learn how to identify and judge the standardization and accuracy of change plans on the labeled data;

[0021] Content review: The content of the operation and maintenance large model review plan mainly includes four steps: content understanding, logical verification, fact checking, and change impact analysis. Through natural language processing technology, it understands the content of the change plan, verifies the rationality of logical relationships, checks the data and facts cited in the plan, and analyzes the specific impacts of the change on the system, business, and users.

[0022] Preferably, in the plan review step, although the operation and maintenance large model is used to automatically review the plan for in-depth content analysis, discover potential problems and risks, and reduce omissions and errors that may occur in manual review, the participation of manual review is still required. For complex situations that are difficult for the model to judge, professionals intervene for analysis. The results of manual review are used to further optimize the model, improve its accuracy and reliability, and provide support for the final change decision.

[0023] A system for a pre-review method of a change plan based on an operation and maintenance large model, comprising:

[0024] Document preprocessing module: used to preprocess the input change plan document, remove irrelevant information in the document, such as headers, footers, and blank characters, and unify the document format, including unifying all headings into the same format and unifying all paragraphs into the same indentation and line spacing;

[0025] Document parsing module: For change plan documents in different formats, call the corresponding parsing library to extract the document content as plain text, while retaining the heading, paragraph, and list structure information in the document;

[0026] Chapter recognition and extraction module: Using the trained operation and maintenance large model, based on a large amount of labeled data to learn the chapter structure features in the document, perform chapter recognition and extraction on the preprocessed and parsed document, and output the extracted chapter content.

[0027] Preferably, the chapter recognition and extraction module includes:

[0028] Model training unit: Use a large number of documents with manually labeled chapter structures as labeled data to train the operation and maintenance large model, so that the model learns the vocabulary features, semantic features, and structure features of the chapters in the document;

[0029] Model optimization unit: During the model training process, adopt stochastic gradient descent, Adam optimization algorithm, combined with dropout, weight decay regularization technology, and adjust hyperparameters such as learning rate, batch size, and number of hidden layer units to optimize the model;

[0030] Chapter prediction unit: Use the trained and optimized model to predict the chapter structure of the unlabeled change plan document, and achieve automatic recognition and extraction of chapters;

[0031] Chapter Verification Unit: Manually check and automatically verify the extracted chapters. The manual check is carried out by professional operation and maintenance personnel to check the accuracy and integrity of each chapter one by one. The automatic verification checks the standardization of chapter titles and the integrity of content by writing verification scripts.

[0032] Preferably, it further includes a change level judgment module, which includes:

[0033] Model Learning Unit: The operation and maintenance large model learns a large number of historical change cases and change specification documents to master the specification requirements, risk level classification criteria, and approval process information corresponding to different change operations;

[0034] Level Judgment Unit: Based on the scope of influence, urgency, and complexity, use supervised learning algorithms to train the change specification data to establish a change level judgment model. The input of the model is the key feature information of the change plan, and the output is the change level and the corresponding confidence level;

[0035] Comprehensive Evaluation Unit: Combine the specific content and context information of the change plan to comprehensively evaluate the change level output by the model. When the change plan involves major modifications to the core business system, adjust the change level according to the business importance, and further review the integrity and accuracy of the plan.

[0036] Preferably, it further includes a plan review module, which includes:

[0037] Audit Standard Establishment Unit: Establish clear audit processes and standards, formulate detailed audit key points and scoring criteria, decompose the audit key points into multiple scoring dimensions such as content integrity, professional term standardization, and solution rationality, and design corresponding prompt words for each dimension;

[0038] Model Learning Unit: The operation and maintenance large model learns the rules and standards of change management by analyzing a large amount of historical change data, specification documents, and actual cases, extracts key features such as change types, scope of influence, implementation steps, and risk levels, and uses supervised learning or semi-supervised learning methods to learn how to identify and judge the standardization and accuracy of change plans on labeled data;

[0039] Content Review Unit: It includes a content understanding sub-unit, which understands the technical details and implementation steps of the change plan through natural language processing technology; a logical verification sub-unit, which verifies the rationality of the logical relationship in the change plan; a fact checking sub-unit, which checks the accuracy of the data and facts cited in the plan; and a change impact analysis sub-unit, which analyzes the specific impact of the change on the system, business, and users.

[0040] Preferably, the plan review module further includes:

[0041] Manual review interface: Provide a way for manual review to be involved. For complex situations that are difficult for the model to judge, professionals will intervene for analysis;

[0042] Model optimization unit: Utilize the results of manual review to further optimize the operation and maintenance large model, improving the accuracy and reliability of the model;

[0043] Decision support unit: Provide support for the final change decision to ensure the correctness of the change implementation. At the same time, record and feedback the review results and decision-making basis for subsequent continuous improvement and optimization of the system.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] The method and system for pre-reviewing change plans based on the operation and maintenance large model proposed by the present invention improve the efficiency of pre-reviewing change plans through intelligent means, shorten the review cycle, and accelerate the process of change implementation; ensure the integrity and standardization of change plans, reducing implementation risks caused by incomplete or non-standard plans; the combination of automatic review and manual review effectively reduces human errors and improves the review quality. Brief description of the drawings

[0046] Figure 1 It is a schematic diagram for section identification and extraction of the present invention;

[0047] Figure 2 It is a schematic diagram for change level judgment of the present invention;

[0048] Figure 3 It is a schematic diagram for content review of the operation and maintenance large model of the present invention. Detailed implementation manners

[0049] In order to clearly and completely describe the objectives, technical solutions, and advantages of the present invention, the following further details the embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are some but not all embodiments of the present invention, and are only used to explain the embodiments of the present invention, not to limit the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0050] Embodiment 1, please refer to Figures 1 to 3 , the present invention provides a technical solution: A method for pre-reviewing change plans based on the operation and maintenance large model. First, extract the necessary sections and verify the content integrity; secondly, judge whether the change level is accurate by learning the change specifications; finally, review the standardization and accuracy of the plan content.

[0051] Chapter identification and extraction are key steps to ensure the integrity and standardization of the change plan document. Before performing chapter identification and extraction, it is first necessary to preprocess the change plan document. This includes removing irrelevant information in the document, such as headers, footers, blank characters, etc., to ensure the purity and standardization of the document content. In addition, it is also necessary to unify the document format, such as unifying all headings into the same format, and unifying all paragraphs into the same indentation and line spacing, etc. These preprocessing steps help to improve the accuracy and efficiency of subsequent chapter identification and extraction.

[0052] Change plan documents usually exist in multiple formats, such as PDF, Word, Markdown, etc. In order to accurately identify and extract chapters, it is necessary to parse documents in different formats. For example, for PDF documents, a PDF parsing library (such as PyPDF2, pdfminer, etc.) can be used to extract the document content as plain text; for Word documents, a Word parsing library (such as python-docx) can be used to extract the document content as plain text; for Markdown documents, a Markdown parsing library (such as markdown2) can be used to extract the document content as plain text. During the parsing process, it is also necessary to pay attention to retaining the structural information in the document, such as headings, paragraphs, lists, etc., for subsequent chapter identification and extraction.

[0053] By training an operation and maintenance large model to automatically learn the chapter structure features in the document, the identification and extraction of chapters can be achieved. When training the model, a large amount of labeled data, that is, documents with manually labeled chapter structures, is required. Through these labeled data, the model can learn the features of chapters in the document, such as the lexical features of headings, the semantic features of paragraphs, the structural features of lists, etc. The trained model can be used to predict the chapter structure of unlabeled documents, realizing the automatic identification and extraction of chapters. In order to improve the accuracy and efficiency of chapter identification, it is necessary to train and optimize the deep learning model. During the training process, some commonly used optimization algorithms, such as Stochastic Gradient Descent (SGD), Adam, etc., can be used. At the same time, some regularization techniques, such as dropout, weight decay, etc., can also be used to prevent the model from overfitting. In addition, the performance of the model can be further improved by adjusting the hyperparameters of the model, such as the learning rate, batch size, number of hidden layer units, etc.

[0054] After the model training is completed, the trained model can be used to identify and extract chapters from unlabeled documents. The extracted chapters need to be verified to ensure their accuracy and integrity. The verification methods can include manual inspection and automatic verification.

[0055] Manual inspection involves professional operation and maintenance personnel checking each extracted chapter one by one to ensure it conforms to the actual content and structure of the document. Automatic verification is to write some verification scripts to automatically check the extracted chapters, such as checking whether the chapter titles conform to the specifications and whether the chapter content is complete.

[0056] In operation and maintenance management, accurately judging the change level is a key link to ensure the effectiveness of change management. The change level determines the approval process, resource allocation, and implementation priority of the change. By learning change specifications through the operation and maintenance large model, accurate judgment of the change level can be achieved, thereby improving the efficiency and reliability of change management.

[0057] Through learning a large number of historical change cases and change specification documents, the operation and maintenance large model can master information such as the specification requirements, risk level classification criteria, and approval process corresponding to different types of change operations. These specifications usually include aspects such as the nature, scope, impact on the system, and involved business processes of the change. For example, changes may be divided into three levels: low risk, medium risk, and high risk, each with clear judgment criteria and processing procedures.

[0058] The judgment of the change level is usually based on the impact scope, urgency, and complexity. The operation and maintenance large model trains the change specification data through supervised learning algorithms to establish a change level judgment model. The input of the model is the key feature information of the change plan (such as change type, impact scope, operation complexity, etc.), and the output is the change level (such as low risk, medium risk, high risk) and the corresponding confidence level. Through learning a large amount of labeled data, the model can accurately judge the change level and provide detailed judgment basis.

[0059] To verify the accuracy of the judgment result, the operation and maintenance large model will comprehensively evaluate the change level output by the model in combination with the specific content and context information of the change plan. For example, if the change plan involves major modifications to the core business system, even if the model judges it as a medium risk level, the operation and maintenance large model will, based on its understanding of the business importance, recommend raising the change level to high risk and further review the integrity and accuracy of the plan.

[0060] When using the operation and maintenance large model for plan review, first, a clear review process and standards need to be established. This includes formulating detailed review points and scoring criteria to ensure that each plan is reviewed according to the same standard. For example, the review points can be decomposed into multiple scoring dimensions, such as content integrity, standardization of professional terms, and rationality of the solution. For each dimension, corresponding prompt words are designed to guide the large model for evaluation.

[0061] The operation and maintenance large model learns and understands the rules and standards of change management by analyzing a large amount of historical change data, specification documents, and actual cases. The large model extracts key features from the collected data, such as change types, impact scopes, implementation steps, risk levels, etc., and uses supervised learning or semi-supervised learning methods to learn how to identify and judge the standardization and accuracy of change plans on labeled data.

[0062] The content review of the operation and maintenance large model for the plan mainly includes four steps. First, content understanding. The operation and maintenance large model understands the content of the change plan through natural language processing technology, including technical details, implementation steps, etc. Second, logical verification. The model verifies whether the logical relationships in the change plan are reasonable, such as whether the risk assessment and control measures correspond. Third, fact checking. The model checks whether the data and facts cited in the plan are accurate and compares them with the technical database or knowledge base. Fourth, change impact analysis. The model analyzes the specific impact of the change on the system, business, and users to ensure that the description in the plan is accurate.

[0063] Using the operation and maintenance large model to automatically review the plan can conduct in-depth content analysis, discover potential problems and risks, and at the same time reduce the omissions and errors that may occur in manual review. Although the operation and maintenance large model has significant advantages in reviewing the standardization and accuracy of plan content, the participation of manual review is still required. For complex situations that are difficult for the model to judge, professional personnel need to intervene for analysis. The results of manual review can be used to further optimize the model, improve its accuracy and reliability, and can provide support for the final change decision to ensure the correctness of change implementation.

[0064] This method significantly improves the efficiency and quality of the pre-review of change plans through intelligent and automated means, not only ensuring the integrity and standardization of change plans, but also improving the accuracy of change level judgment and reducing operation and maintenance risks. In practical applications, this method can bring significant benefits to the enterprise's operation and maintenance work, improve the overall level of operation and maintenance work, and provide strong support for the stable operation of the enterprise's information system.

[0065] Example 2. On the basis of Example 1, a system for the pre-review method of change plans based on the operation and maintenance large model is proposed, including:

[0066] Document preprocessing module: used to preprocess the input change plan document, remove irrelevant information in the document, such as headers, footers, blank characters, and unify the document format, including unifying all titles into the same format and unifying all paragraphs into the same indentation and line spacing;

[0067] Document parsing module: For change plan documents in different formats, call the corresponding parsing library to extract the document content as plain text, while retaining the title, paragraph, and list structure information in the document;

[0068] Chapter Identification and Extraction Module: Using the trained operation and maintenance large model, based on a large amount of labeled data to learn the chapter structure features in the document, identify and extract chapters from the preprocessed and parsed document, and output the extracted chapter content. It includes: Model Training Unit: Use a large number of documents with manually labeled chapter structures as labeled data to train the operation and maintenance large model, enabling the model to learn the lexical features, semantic features, and structural features of chapters in the document; Model Optimization Unit: During the model training process, adopt stochastic gradient descent and Adam optimization algorithms, combined with dropout and weight decay regularization techniques, as well as adjust hyperparameters such as learning rate, batch size, and number of hidden layer units to optimize the model; Chapter Prediction Unit: Use the trained and optimized model to predict the chapter structure of the unlabeled change plan document, realizing automatic identification and extraction of chapters; Chapter Verification Unit: Manually check and automatically verify the extracted chapters. Manual check is carried out by professional operation and maintenance personnel to check the accuracy and integrity of each chapter one by one, and automatic verification checks the standardization of chapter titles and the integrity of content by writing verification scripts.

[0069] It also includes a change level judgment module, which includes: Model Learning Unit: The operation and maintenance large model learns the specification requirements, risk level classification standards, and approval process information corresponding to different change operations by learning a large number of historical change cases and change specification documents; Level Judgment Unit: Based on the scope of influence, urgency, and complexity, use supervised learning algorithms to train the change specification data to establish a change level judgment model. The input of the model is the key feature information of the change plan, and the output is the change level and the corresponding confidence level; Comprehensive Evaluation Unit: Combine the specific content and context information of the change plan to comprehensively evaluate the change level output by the model. When the change plan involves major modifications to the core business system, adjust the change level according to the business importance, and further review the integrity and accuracy of the plan.

[0070] It also includes a solution review module, which includes: a review standard establishment unit: establishing clear review processes and standards, formulating detailed review points and scoring criteria, decomposing the review points into multiple scoring dimensions such as content integrity, standardization of professional terms, and rationality of solution, and designing corresponding prompt words for each dimension; a model learning unit: the operation and maintenance large model learns the rules and standards of change management by analyzing a large amount of historical change data, specification documents and actual cases, extracts key features such as change types, impact scopes, implementation steps, and risk levels, and uses supervised learning or semi-supervised learning methods to learn how to identify and judge the standardization and accuracy of change solutions on labeled data; a content review unit: including a content understanding sub-unit, understanding the technical details and implementation steps of the change solution through natural language processing technology; a logical verification sub-unit, verifying the rationality of the logical relationship in the change solution; a fact checking sub-unit, checking the accuracy of the data and facts cited in the solution; a change impact analysis sub-unit, analyzing the specific impact of the change on the system, business, and users.

[0071] The solution review module also includes: a manual review interface: providing a way for manual review to be involved. For complex situations that are difficult for the model to judge, professionals intervene for analysis; a model optimization unit: using the results of manual review to further optimize the operation and maintenance large model, improving the accuracy and reliability of the model; a decision support unit: providing support for the final change decision to ensure the correctness of change implementation, and at the same time recording and feedbacking the review results and decision-making basis for subsequent continuous improvement and optimization of the system.

[0072] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A pre - review method for change solutions based on an operation and maintenance large - model, characterized in that: It includes the following steps: Document preprocessing: Preprocess the change plan document, remove irrelevant information in the document such as headers, footers, and blank characters, and unify the document format. Unify all headings into the same format, and unify all paragraphs into the same indentation and line spacing; Document parsing: For change plan documents in different formats, use the corresponding parsing library to extract the document content as plain text, and retain the heading, paragraph, and list structure information in the document; Chapter identification and extraction: Utilize the operation and maintenance large model to learn the chapter structure features in the document, train the model with a large amount of labeled data, enable the model to automatically identify and extract the chapter structure of unlabeled documents, and conduct manual inspection and automatic verification on the extracted chapters to ensure their accuracy and integrity.

2. The pre - review method for change plans based on an operation and maintenance large - model according to claim 1, wherein: The process of using the operation and maintenance large model to learn the chapter structure features in the document and automatically identify and extract the chapter structure includes: Model training: Use a large number of documents with manually labeled chapter structures as labeled data to train the operation and maintenance large model, enabling the model to learn the features of chapters in the document, such as the lexical features of headings, the semantic features of paragraphs, and the structural features of lists; Model optimization: During the model training process, adopt stochastic gradient descent and Adam optimization algorithms, use dropout and weight decay regularization techniques to prevent the model from overfitting, and improve the model performance by adjusting hyperparameters such as the learning rate, batch size, and number of hidden layer units; Chapter prediction: Use the trained model to predict the chapter structure of unlabeled documents to achieve automatic identification and extraction of chapters.

3. The pre - review method for change solutions based on the operation and maintenance large - model according to claim 2, wherein: It also includes the step of change level judgment: Model learning: The operation and maintenance large model learns the specification requirements, risk level classification criteria, and approval process information corresponding to different types of change operations by learning a large number of historical change cases and change specification documents; Level judgment: Based on the scope of influence, urgency, and complexity, train the change specification data through a supervised learning algorithm to establish a change level judgment model. The input of the model is the key feature information of the change plan, and the output is the change level and the corresponding confidence level; Comprehensive evaluation: Combine the specific content and context information of the change plan to comprehensively evaluate the change level output by the model to ensure the accuracy of the change level judgment.

4. The pre - review method for change plans based on the operation and maintenance large - model according to claim 3, characterized in that: The steps of plan review include: Process and standard establishment: Establish clear review processes and standards, formulate detailed review key points and scoring criteria, decompose the review key points into multiple scoring dimensions, such as content integrity, professional term standardization, and solution rationality, and design corresponding prompt words for each dimension to guide the large model to conduct evaluations; Model learning: The operation and maintenance large model learns the rules and standards of change management by analyzing a large amount of historical change data, specification documents, and actual cases, extracts key features from the data, and uses supervised learning or semi-supervised learning methods to learn how to identify and judge the standardization and accuracy of change plans on labeled data. Content Review: The content review plan for the operation and maintenance large model mainly includes four steps: content understanding, logical verification, fact checking, and change impact analysis. Through natural language processing technology, it understands the content of the change plan, verifies the rationality of logical relationships, checks the data and facts cited in the plan, and analyzes the specific impacts of the change on the system, business, and users.

5. The pre-review method for change solutions based on an operation and maintenance large model according to claim 4, characterized in that: In the plan review step, although the operation and maintenance large model is used to automatically review the plan for in-depth content analysis, discover potential problems and risks, and reduce the omissions and errors that may occur in manual review, the participation of manual review is still required. For complex situations that are difficult for the model to judge, professionals are involved in the analysis. The results of manual review are used to further optimize the model, improve its accuracy and reliability, and provide support for the final change decision.

6. A system for the pre - review method of change solutions based on the operation and maintenance large - model according to claim 1, characterized in that: Including: Document Preprocessing Module: It is used to preprocess the input change plan document, remove irrelevant information in the document, such as headers, footers, and blank characters, and unify the document format, including unifying all headings into the same format and unifying all paragraphs into the same indentation and line spacing; Document Parsing Module: For change plan documents in different formats, call the corresponding parsing library to extract the document content as plain text, while retaining the heading, paragraph, and list structure information in the document; Chapter Identification and Extraction Module: Using the trained operation and maintenance large model, based on a large amount of labeled data to learn the chapter structure features in the document, identify and extract chapters from the preprocessed and parsed document, and output the extracted chapter content.

7. A system according to claim 6, characterized in that: The Chapter Identification and Extraction Module includes: Model Training Unit: Use a large number of documents with manually labeled chapter structures as labeled data to train the operation and maintenance large model, so that the model learns the vocabulary features, semantic features, and structure features of chapters in the document; Model Optimization Unit: During the model training process, adopt stochastic gradient descent and Adam optimization algorithms, combined with dropout and weight decay regularization techniques, and adjust hyperparameters such as learning rate, batch size, and the number of hidden layer units to optimize the model; Chapter Prediction Unit: Use the trained and optimized model to predict the chapter structure of the unlabeled change plan document to achieve automatic identification and extraction of chapters; Chapter Verification Unit: Manually check and automatically verify the extracted chapters. Manual checking is carried out by professional operation and maintenance personnel to check the accuracy and integrity of each chapter one by one, and automatic verification checks the standardization of chapter headings and content integrity by writing verification scripts.

8. A system according to claim 7, characterized in that: It also includes a change level judgment module, which includes: Model Learning Unit: The operation and maintenance large model learns a large number of historical change cases and change specification documents to master the specification requirements, risk level classification standards, and approval process information corresponding to different change operations; Level Judgment Unit: Based on the scope of influence, urgency, and complexity, use supervised learning algorithms to train the change specification data to establish a change level judgment model. The input of the model is the key feature information of the change plan, and the output is the change level and the corresponding confidence level; Comprehensive Evaluation Unit: Combine the specific content of the change plan and context information to comprehensively evaluate the change level output by the model. When the change plan involves major modifications to the core business system, adjust the change level according to the business importance, and further review the integrity and accuracy of the plan.

9. A system according to claim 8, characterized in that: It also includes a Plan Review Module, which includes: Audit Standard Establishment Unit: Establish clear audit processes and standards, formulate detailed audit key points and scoring criteria, decompose the audit key points into multiple scoring dimensions such as content integrity, standardization of professional terms, and rationality of solution, and design corresponding prompt words for each dimension; Model Learning Unit: The operation and maintenance large model learns the rules and standards of change management by analyzing a large amount of historical change data, specification documents, and actual cases, extracts key features such as change types, impact scope, implementation steps, and risk levels, and uses supervised learning or semi-supervised learning methods to learn how to identify and judge the standardization and accuracy of change plans on labeled data; Content Review Unit: Includes a Content Understanding Sub-unit, which understands the technical details and implementation steps of the change plan through natural language processing technology; a Logic Verification Sub-unit, which verifies the rationality of the logical relationships in the change plan; a Fact Verification Sub-unit, which checks the accuracy of the data and facts cited in the plan; and a Change Impact Analysis Sub-unit, which analyzes the specific impact of the change on the system, business, and users.

10. A system according to claim 9, wherein: The Plan Review Module also includes: Manual Review Interface: Provide a way for manual review to participate. For complex situations that are difficult for the model to judge, professional personnel intervene for analysis; Model Optimization Unit: Utilize the results of manual review to further optimize the operation and maintenance large model, improving the accuracy and reliability of the model; Decision Support Unit: Provide support for the final change decision to ensure the correctness of change implementation, and at the same time record and feedback the review results and decision-making basis for subsequent continuous improvement and optimization of the system.

Images