Information system operation and maintenance work ticket auxiliary auditing method based on large model

Abstract

The invention relates to the technical field of work ticket auxiliary auditing, and discloses an information system operation and maintenance work ticket auxiliary auditing method based on a large model, and the method comprises the steps: obtaining the content of a work ticket, carrying out the text cleaning of the content, and constructing a content accuracy auditing prompt word; the cue word is transmitted into a large model, and a content accuracy auditing opinion is obtained; constructing a work scheme rationality auditing prompt word, and obtaining a work scheme rationality auditing opinion and a potential risk point list; constructing a safety measure completeness auditing prompt word, and obtaining a safety measure completeness auditing opinion; obtaining a content compliance auditing opinion, and generating a final work ticket auxiliary auditing opinion; according to the method, the auditing efficiency, quality and safety are comprehensively improved, and a powerful guarantee is provided for safe and stable operation of an enterprise information system. And meanwhile, through knowledge precipitation and a continuous optimization mechanism, continuous improvement of the auditing capability is realized, and core competitiveness of enterprise operation and maintenance management is formed.

Description

Technical Field

[0001] The present invention relates to the technical field of auxiliary auditing of work tickets, and more specifically, to an auxiliary auditing method for information system operation and maintenance work tickets based on a large model. Background Art

[0002] Information systems are the core of an enterprise's production and operations. To ensure their normal operation, developing and implementing information system operation and maintenance work tickets and process management systems is crucial for enterprise safety. The work ticket mechanism is a management system that regulates the scheduling, review, and recording of information system operation and maintenance work. Accurate completion, strict review, and implementation are crucial for improving operation and maintenance efficiency, mitigating operational risks, and ensuring system security.

[0003] A work ticket primarily includes the work ticket number, applicant, application time, planned work hours, staff, work location, work plan, risk analysis, safety measures, and the person in charge of the work. The work ticket is completed by the person in charge of the work and must be reviewed and approved by an auditor with relevant work experience and expertise before it can be executed.

[0004] The traditional manual review method has the following problems:

[0005] Low audit efficiency: Manual item-by-item auditing is time-consuming and labor-intensive;

[0006] Inconsistent audit standards: Differences in knowledge and experience among auditors lead to inconsistent audit results;

[0007] Easy to omit: Long audit time can easily lead to loss of concentration and omission of important risk points;

[0008] Limited expertise: Even experienced professionals may have areas of knowledge gaps.

[0009] Large models (such as ChatGPT, Llama, Qwen, and DeepSeek) possess strong language understanding, rich knowledge, excellent reasoning, and robust generation capabilities, making them ideally suited to assisting with the review of information system operation and maintenance work tickets. Applying large models to work ticket review can overcome the limitations of traditional manual review and improve review efficiency and quality. Summary of the Invention

[0010] The present invention proposes an information system operation and maintenance work ticket auxiliary audit method based on a large model, which can effectively utilize the language understanding, knowledge integration and logical reasoning capabilities of the large model to provide auditors with comprehensive audit opinions as a reference, thereby improving the quality and efficiency of the audit work.

[0011] The present invention provides an information system operation and maintenance work ticket auxiliary review method based on a large model, comprising the following steps:

[0012] Obtain the work ticket content and perform text cleaning on it to form a standard work ticket content text;

[0013] Construct a content accuracy review prompt, which is composed of the work ticket content and the content accuracy review instruction. This prompt is passed to the big model to obtain the content accuracy review opinion.

[0014] Based on the work ticket content and the review opinions on the accuracy of the content, a work plan rationality review prompt is constructed. This prompt is composed of the work plan content in the work ticket and the work plan rationality review instructions. This prompt is passed to the big model to obtain the work plan rationality review opinions and a list of potential risk points.

[0015] Based on the work plan rationality review opinions and the list of potential risk points, a safety measures completeness review prompt is constructed. This prompt is composed of the risk point analysis and safety measures content in the work ticket, along with the safety measures completeness review instructions. This prompt is then passed to the big model to obtain the safety measures completeness review opinions.

[0016] Based on the work ticket content, content accuracy review opinions, work plan rationality review opinions, and safety measures completeness review opinions, a content compliance review prompt is constructed. This prompt is composed of the work ticket content and the content compliance review instructions. This prompt is passed to the big model to obtain the content compliance review opinion.

[0017] The final work ticket auxiliary review opinions are generated by comprehensively considering the review opinions on content accuracy, work plan rationality, safety measures completeness and content compliance.

[0018] Furthermore, when the work ticket is a paper document, the step of obtaining the content of the work ticket includes: taking a photo or scanning the paper work ticket, performing text recognition on the obtained image, and processing the recognition result into a work ticket content text in a standard format.

[0019] Furthermore, when the work ticket is an electronic document, the step of obtaining the work ticket content includes: parsing the electronic document content, and processing the parsing result into a work ticket content text in a standard format.

[0020] Furthermore, when the work ticket is located in the information management system, the step of obtaining the work ticket content includes: extracting the work ticket report content from the information management system, and processing the extraction result into a work ticket content text in a standard format.

[0021] Furthermore, content accuracy review instructions are used to guide the large model to evaluate whether the work ticket contains key content such as staff, work supervisor, work content, risk point analysis, safety measures, etc.

[0022] Furthermore, the work plan rationality review instructions are used to guide the big model to evaluate the work plan from aspects such as the standardization of professional terminology and the rationality of the solution, and to analyze the risk points of the work plan.

[0023] Furthermore, when constructing the prompt words for reviewing the rationality of the work plan, the following steps are also included:

[0024] Obtain enterprise information system knowledge related to the work plan by searching the enterprise information system knowledge base;

[0025] The acquired enterprise information system knowledge is used as known information and combined with the work plan and the work plan rationality review instructions to form prompt words.

[0026] Furthermore, the safety measures completeness review instructions are used to guide the risk point analysis in the large model assessment work ticket and the rationality and completeness of the corresponding safety measures.

[0027] Furthermore, when constructing content compliance review prompt words, the following steps are also included:

[0028] By searching the enterprise operation and maintenance specification knowledge base, obtain enterprise operation and maintenance regulations and specifications related to the work ticket content;

[0029] The acquired enterprise operation and maintenance regulations and specifications are used as known information and spliced ​​with the work ticket content and content compliance review instructions to form prompt words.

[0030] Furthermore, the method further comprises the following steps:

[0031] Choose different types of large models for different review stages:

[0032] During the work plan rationality review phase, select a large model that has been fine-tuned with expertise in the operations and maintenance field;

[0033] During the security measures completeness review phase, select large models that have been fine-tuned with domain expertise;

[0034] During the content accuracy review and content compliance review stages, select the general large model.

[0035] The beneficial effects of the present invention are:

[0036] Significantly improve audit efficiency: Using this method for auxiliary audits can shorten audit time; significantly reduce the workload of auditors, allowing them to focus on higher-value decisions; and can process multiple work tickets at the same time, improving audit throughput; reduce human negligence and omissions, and ensure the comprehensiveness and consistency of audits; combine professional knowledge bases and fine-tuning models to provide in-depth professional audit opinions, and multi-dimensional audits (accuracy, rationality, security, and compliance) to ensure the overall improvement of work ticket quality; improve the coverage of safety measures, ensure that each risk point has corresponding preventive measures, establish a risk warning mechanism, and proactively identify potential high-risk operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0038] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that these embodiments are discussed solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. Furthermore, features described in some examples may be combined in other examples.

[0039] At least one embodiment of the present invention discloses a method for assisting in reviewing information system operation and maintenance work tickets based on a large model, such as Figure 1 As shown, the following steps are included:

[0040] Step 1: Obtain the work ticket content and perform text cleaning on the content to form a standard format work ticket content text;

[0041] In this embodiment, the work ticket content acquisition step is used to acquire the work ticket content to be reviewed and process it to form a standard work ticket content text. Depending on the different forms of the work ticket, different acquisition and processing methods can be used:

[0042] When the work ticket is a paper document:

[0043] Use a high-precision scanning device or smartphone to take a photo or scan the paper work ticket;

[0044] Use deep learning-based OCR (Optical Character Recognition) technology, such as Tesseract OCR, Google Vision API, or Baidu OCR, to perform text recognition on the acquired images;

[0045] Post-process the OCR recognition results, including correcting possible recognition errors and adjusting text format;

[0046] Arrange the processed text content according to the predefined structured format to form a standard format of work ticket content text;

[0047] When processing OCR recognized text, the following text correction algorithms can be used for post-processing:

[0048] Text similarity calculation function:

[0049]

[0050] Where T1 and T2 are the two texts to be compared, EditDistance is the edit distance function, len is the text length function, and max is the maximum value function. When the similarity exceeds the preset threshold θ (usually 0.8), the text is considered to have OCR recognition errors and needs to be corrected.

[0051] The structural transformation of the work ticket content can be expressed as a mapping function:

[0052] f:D raw →D struct ;

[0053] Among them, D raw The original work ticket content, D struct This is the structured content of the work ticket.

[0054] When the work ticket is an electronic document (such as Word, PDF, etc.):

[0055] For Word documents, use document parsing tools to parse the document content and extract text and structure information;

[0056] For PDF documents, use PDF parsing tools to parse the content;

[0057] For Excel tables, use table data processing tools to extract and parse table data;

[0058] Based on the structural characteristics of the document, identify and extract key information of the work ticket, such as title, applicant, work plan, etc.

[0059] Organize the extracted information into work ticket content text in a standard format;

[0060] When a work ticket is in the information management system:

[0061] Directly obtain work ticket data through the API interface or database connection provided by the system;

[0062] Extract work ticket information from the database using database query or object-relational mapping techniques;

[0063] According to the system data model, the acquired data is converted into the work ticket content text in a standard format;

[0064] For different information management systems, specific adapters can be developed to unify data acquisition methods;

[0065] Regardless of the original form of the work ticket, the core of obtaining the work ticket content is to convert the original work ticket content in different forms into a unified standard format.

[0066] The standard work ticket content text is usually in a structured format and contains the following main fields:

[0067] Ticket number: unique identification number of the work ticket;

[0068] Applicant: Name of the person applying for the work ticket;

[0069] Application time: the time when the work ticket application is submitted;

[0070] Planned working time: including planned start time and planned end time;

[0071] Staff: List of staff involved in the work;

[0072] Work location: the specific location where the work is performed;

[0073] Work plan: detailed description of the work plan;

[0074] Risk point analysis: list of identified risk points;

[0075] Safety measures: a list of safety measures taken against risk points;

[0076] Work leader: the person responsible for this work;

[0077] Issuer: The person who reviews and issues the work ticket;

[0078] The work ticket content acquisition step also includes data quality checks to ensure that key information is not lost during the text conversion process. When implementing this, the following strategies can be adopted:

[0079] Set up mandatory field checks to ensure that key information (such as work plans, risk analysis, etc.) is not missing;

[0080] Implement format validation to ensure that the date, time, etc. are in the correct format;

[0081] Conduct consistency checks to ensure consistency of information such as staff members and work leaders;

[0082] The output of this step is the work ticket content text in a standard format, which provides a unified input format for the subsequent review step.

[0083] Step 2: Construct a content accuracy review prompt. This prompt is composed of the work ticket content and the content accuracy review instruction. This prompt is passed to the big model to obtain the content accuracy review opinion.

[0084] The specific implementation steps are as follows:

[0085] Build a content accuracy review prompt template:

[0086] The prompt word template contains the following main parts:

[0087] Role definition: The reviewer is defined as an experienced information system operation and maintenance work ticket review expert;

[0088] Task Description: Required to check the accuracy and completeness of the work ticket content;

[0089] Work Ticket Content: A placeholder for the work ticket content;

[0090] Evaluation points: List the aspects that need to be evaluated, including the completeness of basic information, the completeness of work content descriptions, the completeness of safety information, the completeness of responsible person information, etc.

[0091] Output requirements: Requires detailed pointing out of incomplete or inaccurate parts and providing suggestions for improvement;

[0092] Fill the work ticket content into the prompt word template to generate a complete prompt word:

[0093] Replace the work ticket content placeholder in the template and insert the work ticket content in the standard format;

[0094] According to the work ticket type (such as software upgrade, hardware maintenance, daily inspection, etc.), the prompt word template can be appropriately adjusted to add review points for specific types of work tickets;

[0095] Pass the complete prompt word into the universal large model:

[0096] Available general large models include but are not limited to GPT-4, Llama3, Claude3, Qwen2, etc.;

[0097] Set appropriate inference parameters, such as a lower temperature parameter (to maintain output certainty) and a moderate diversity control parameter;

[0098] Interact with large models through API calls or SDKs;

[0099] Parsing the review comments returned by the large model:

[0100] Use text parsing methods to extract key information from large model outputs;

[0101] Categorize and organize audit comments into a structured format, including the following information:

[0102] Completeness score: a quantitative assessment of the completeness of the work ticket content;

[0103] Problem list: discovered problems, each problem includes the problem type, involved fields, problem description, severity, improvement suggestions, etc.

[0104] Overall assessment: evaluation of the overall integrity of the work order;

[0105] Completeness score calculation function:

[0106]

[0107] Among them, C score is the total score of the work ticket completeness, n is the total number of inspection items, w i is the weight of the i-th inspection item, c i The completion status of the i-th check item (1 for completed, 0 for incomplete). You can set higher weights for important fields, such as a work plan with a weight of 2.0 and common fields with a weight of 1.0.

[0108] Problem severity function:

[0109] S severity =α·I impact +β·F field_importance ;

[0110] Among them, S severity Indicates the severity score of the problem, I impact Indicates the degree of impact of the problem on the execution of the work ticket, F field_importance Indicates the importance of the field, α and β are weight coefficients, usually α + β = 1. Impact degree I impact and field importance F field_importance A rating scale of 1-5 is usually used.

[0111] The focus of content accuracy audits is to thoroughly check the completeness of work tickets, ensuring that all necessary information is included. This lays the foundation for subsequent rationality and safety audits. By leveraging the powerful understanding capabilities of the universal big model, we can quickly identify missing and inaccurate content in work tickets, improving audit efficiency and accuracy.

[0112] For complex or specialized work tickets, the prompts can be further optimized and specific checkpoints can be added to ensure comprehensiveness and professionalism. For example, for work tickets involving database operations, the prompts can emphasize whether a data backup plan is included; for work tickets involving network equipment adjustments, the checklist can emphasize whether network topology change instructions are included.

[0113] Step 3: Based on the work ticket content and the review opinion on the accuracy of the content, a work plan rationality review prompt is constructed. This prompt is composed of the work plan content in the work ticket and the work plan rationality review instructions. This prompt is passed to the big model to obtain the work plan rationality review opinion and a list of potential risk points.

[0114] The specific implementation steps are as follows:

[0115] Knowledge base retrieval and preparation:

[0116] Build an enterprise information system knowledge base index and use a vector database to store the embedded vectors of system documents;

[0117] Extract keywords and concepts related to the work plan from the work ticket and construct a search query;

[0118] Use semantic retrieval methods to obtain information related to work plans from the enterprise information system knowledge base;

[0119] Semantic similarity calculation function:

[0120]

[0121] Among them, q is the query text vector, d i is the vector representation of the i-th document in the knowledge base, V q is the vector embedding representation of the query text, is the vector embedding representation of the i-th document in the knowledge base, and cos is the cosine similarity function. The search results are sorted in descending order of similarity, and the top K results are selected, where K is the preset threshold for the number of search results.

[0122] Constructing a work plan rationality review prompt word: The prompt word structure includes the following main parts:

[0123] Role definition: Defined as a senior information system operation and maintenance expert;

[0124] Assessment Target: Includes work ticket content, content accuracy review opinions, and relevant system knowledge;

[0125] Evaluation dimensions: List the aspects that need to be evaluated, including the standardization of professional terminology, completeness of operational procedures, technical feasibility, rationality of resource requirements, identification of risk points, etc.

[0126] Output requirements: Detailed analysis and recommendations are required, and all possible risk points are listed;

[0127] Fill in the prompt words and pass them into the professional large model:

[0128] Fill the work ticket content, content accuracy review opinions and relevant information retrieved from the knowledge base into the prompt word template;

[0129] Choose a large model that has been fine-tuned with IT operations domain expertise;

[0130] Set appropriate inference parameters to ensure the professionalism and depth of the output results;

[0131] Review comments returned by the analytical professional model:

[0132] Comprehensive scoring function for the rationality of the work plan:

[0133]

[0134] Among them, R score is the comprehensive score of the rationality of the work plan, m is the total number of evaluation dimensions, w j is the weight of the jth evaluation dimension, r j The score of the jth evaluation dimension (usually a score of 1-5). Evaluation dimensions may include the standardization of professional terminology, completeness of operational procedures, technical feasibility, rationality of resource requirements, and identification of risk points.

[0135] Calculation of risk point impact:

[0136] RiskImpact(r i )=P(r i )·S(r i );

[0137] Among them, RiskImpact(r i ) is the risk point r i The impact score, P(r i ) is the risk point r i The probability of occurrence (between 0 and 1), S(r i ) is the risk point r i The severity of the impact (1-5). The higher the risk impact score, the more priority the risk point needs to be addressed.

[0138] Reviewing the rationality of a work plan is one of the most technically demanding aspects of the entire review process, requiring a combination of enterprise-specific system knowledge and industry best practices. Using a large model fine-tuned for specialized domains can provide a more accurate and in-depth technical assessment.

[0139] The review criteria can be further refined for different types of work plans. For example:

[0140] For system upgrades: focus on evaluating version compatibility, upgrade path rationality, and rollback plan completeness;

[0141] For network configuration work: focus on evaluating the impact of network topology changes, the correctness of configuration commands, and the rationality of redundancy design;

[0142] For database maintenance work: focus on evaluating data consistency assurance measures, performance impact assessment, and data security protection measures;

[0143] By introducing the enterprise information system knowledge base, the big model can conduct more targeted assessments based on the enterprise's specific system environment and technology stack, improving the accuracy and practicality of the audit.

[0144] Step 4: Based on the work plan rationality review opinion and the list of potential risk points, a safety measures completeness review prompt is constructed. This prompt is composed of the risk point analysis and safety measures content in the work ticket, along with the safety measures completeness review instructions. This prompt is then passed to the big model to obtain the safety measures completeness review opinion.

[0145] The specific implementation steps are as follows:

[0146] Integrate risk point information:

[0147] Collect the risk items listed in the original risk point analysis of the work ticket;

[0148] Integrate potential risk points identified during the rationality review phase of the work plan;

[0149] De-duplicate and merge risk points to form a complete risk point list;

[0150] According to the risk impact and probability of occurrence, risk points are graded (such as high, medium, and low);

[0151] Constructing security measures integrity audit prompt words: The prompt word structure includes the following main parts:

[0152] Role definition: Defined as an operations and maintenance security expert focused on IT system security;

[0153] Assessment objects: including work order content, risk point analysis of the original work order, safety measures of the original work order, and newly identified risk points;

[0154] Evaluation dimensions: including completeness of risk point analysis, coverage of safety measures, effectiveness of safety measures, operability of safety measures, etc.

[0155] Output requirements: Point out the risk points that are not effectively covered and propose specific safety measures;

[0156] Fill in the prompt words and pass them into the professional large model:

[0157] Fill the work ticket content, original risk point analysis, original safety measures and newly identified risk points into the prompt word template;

[0158] Select a large model fine-tuned with security domain knowledge to ensure professional understanding of IT operations security risks;

[0159] When using a fine-tuning model, lower temperature parameters can be used to ensure the accuracy and consistency of the output;

[0160] The security measures completeness review opinions returned by the parsing model:

[0161] Extract assessment results on the completeness of risk point analysis, coverage, effectiveness, and operability of safety measures

[0162] Identify risk points that are not effectively covered and provide corresponding safety measures;

[0163] Organize the review comments into a structured format that includes the following information:

[0164] Risk Analysis Completeness Score: A quantitative assessment of the comprehensiveness of the original risk analysis;

[0165] Safety measures coverage: the degree to which safety measures cover risk points;

[0166] Safety measures effectiveness score: the degree to which safety measures can effectively reduce risks;

[0167] Safety measure operability score: the degree to which the safety measure is specific and operable;

[0168] List of uncovered risk points: including risk description, severity, recommended measures, etc.;

[0169] Improvement suggestions: Overall improvement suggestions for safety measures

[0170] Risk point coverage calculation function:

[0171]

[0172] Among them, N covered is the number of risk points covered by existing security measures, N total The total number of all identified risk points. CoverageRate represents the coverage rate of risk points, ranging from 0 to 1. The closer the value is to 1, the more comprehensive the coverage of identified risk points by security measures.

[0173] Safety measures effectiveness scoring function:

[0174]

[0175] Among them, EffScore (r i ,m i ) is the risk point r i Corresponding safety measures m i The effectiveness score (1-5), N total is the total number of all identified risk points, E effectiveness Indicates the overall effectiveness score of the security measures, ranging from 1 to 5. A higher value indicates better overall effectiveness of the security measures.

[0176] Comprehensive score of security measures completeness:

[0177] S score =γ1·CoverageRate+γ2·E effectiveness +γ3·O operability ;

[0178] Among them, O operability Score the operability of the safety measures (1-5), score is the comprehensive score of the completeness of security measures, γ1, γ2 and γ3 are weight coefficients, representing the importance of coverage, effectiveness and operability in the comprehensive score, and γ1+γ2+γ3=1. Comprehensive score S score The value range depends on the normalization of each item and is usually designed to be between 1-5 points or 0-1.

[0179] Reviewing the completeness of security measures is a critical step in ensuring safe operations. By systematically assessing the alignment of risk points with security measures, operational security risks can be effectively mitigated. Using a large model fine-tuned with security domain knowledge can provide more professional security assessments.

[0180] Specific safety checklists can be introduced for different types of operation and maintenance work:

[0181] For operations involving important data: focus on checking data backup, data recovery testing, data isolation measures, etc.

[0182] For operations involving network configuration: focus on checking firewall rule verification, access control measures, traffic monitoring methods, etc.

[0183] For operations involving permission changes: focus on checking compliance with the principle of least privilege, enabling audit logs, and setting permission validity periods;

[0184] In addition, security measures should also consider emergency plans for different scenarios, including but not limited to:

[0185] Rollback plan for failed operations;

[0186] Emergency response process for system anomalies;

[0187] Recovery measures for data corruption;

[0188] alternatives to service disruptions;

[0189] Through comprehensive and in-depth safety measures assessment, ensure that the safety protection system in the work ticket can effectively deal with possible risks.

[0190] Step 5: Based on the work ticket content, content accuracy review opinions, work plan rationality review opinions, and safety measures completeness review opinions, a content compliance review prompt is constructed. This prompt is composed of the work ticket content and the content compliance review instructions. This prompt is passed to the big model to obtain the content compliance review opinion.

[0191] The specific implementation steps are as follows:

[0192] Knowledge base retrieval and preparation:

[0193] Build an enterprise operation and maintenance specification knowledge base, including the following contents:

[0194] Enterprise IT management system and process specifications;

[0195] Industry regulations and standards (such as ISO27001, ITIL, COBIT, etc.);

[0196] Enterprise-specific security policies and compliance requirements;

[0197] Historical audit cases and best practices;

[0198] Use semantic retrieval technology to retrieve regulations and specifications related to the work ticket content from the knowledge base;

[0199] Based on the type and content of the work ticket, select the most relevant compliance requirements for review;

[0200] Constructing content compliance audit prompt words: The prompt word structure includes the following main parts:

[0201] Role definition: Defined as an expert focused on IT operations and compliance;

[0202] Evaluation objects: including the content of the work order, summary of previous review opinions, relevant enterprise regulations and industry standards;

[0203] Assessment dimensions: including process compliance, role and authority compliance, technical standards compliance, security policy compliance, time window compliance, etc.

[0204] Output requirements: Point out any non-compliance issues and provide specific compliance recommendations;

[0205] Fill in the prompt words and pass them into the general large model:

[0206] Fill in the work ticket content, summary of previous review opinions and relevant regulations and standards into the prompt word template;

[0207] Use a general-purpose model for compliance assessment, as compliance audits are primarily based on clear rules and standards;

[0208] If the company has specific compliance checklists or assessment criteria, these can be provided to the model as additional information;

[0209] Content compliance review opinions returned by the parsing model:

[0210] Extract the assessment results of each compliance dimension;

[0211] Identify non-compliance items and corresponding compliance recommendations;

[0212] Organize the review comments into a structured format that includes the following information:

[0213] Compliance assessment: assessment results of each compliance dimension and overall compliance status;

[0214] List of non-compliance issues: including problem areas, problem descriptions, references, severity, and improvement suggestions;

[0215] Compliance Improvement Suggestions: Overall improvement suggestions for work ticket compliance. Compliance score calculation function:

[0216]

[0217] Where k is the total number of compliance inspection items, CompItem i is the compliance level of the i-th compliance item (between 0 and 1), W i is the weight of the compliance item, and ComplianceScore is the final compliance score result (between 0 and 1, the closer to 1, the higher the compliance). i satisfy Ensure standardization of scoring.

[0218] Content compliance review is a key element in work ticket review, ensuring that operations and maintenance work adheres to internal enterprise regulations and external regulatory requirements. By incorporating enterprise-specific specifications and standards, the large model can provide targeted compliance assessments.

[0219] Compliance requirements can vary significantly for different types of businesses and industries:

[0220] Financial institutions: They need to pay special attention to compliance requirements in areas such as data security, access control, and audit tracking;

[0221] Medical institutions: They need to focus on compliance requirements in areas such as patient data privacy protection and system availability;

[0222] Government agencies: They must strictly comply with relevant laws and regulations and security level protection requirements;

[0223] In addition, compliance audits should also focus on cross-departmental collaboration and compliance with the scope of impact of changes:

[0224] Whether relevant departments and users have been notified as required;

[0225] Whether necessary change approvals have been obtained;

[0226] Whether an emergency plan that complies with regulations has been formulated;

[0227] Through comprehensive compliance audits, we ensure that work tickets comply with various internal and external regulations and standards of the enterprise, thereby reducing compliance risks.

[0228] Step 6: Based on the review opinions on content accuracy, work plan rationality, safety measures completeness, and content compliance, generate the final auxiliary review opinions for the work ticket;

[0229] The specific implementation steps are as follows:

[0230] Integration and analysis of audit results:

[0231] Collect structured results across four review phases;

[0232] Problems found at each stage are classified according to their severity, usually into:

[0233] Critical: Must be resolved; otherwise, it may lead to system failure, data loss, or security incidents.

[0234] High-risk issues (High): It is strongly recommended to resolve them, as they may affect system stability or security;

[0235] Medium-risk issues (Medium): Recommend solutions, which may affect operational efficiency or cause potential problems;

[0236] Low-risk issues (Low): can be resolved as appropriate, mainly optimization or improvement suggestions;

[0237] Analyze the correlation between problems and identify common problems and root causes;

[0238] Summarize the scores of each stage and the overall evaluation results;

[0239] Construct review comments to generate prompt words:

[0240] The prompt word structure consists of the following main parts:

[0241] Role definition: Defined as a senior IT operation and maintenance management expert;

[0242] Input content: including basic information of the work ticket and structured results of the four review stages;

[0243] Output requirements: Generate audit opinions including overall assessment of work tickets, list of key issues, improvement suggestions, audit conclusions, etc.;

[0244] Quality requirements: Ensure that the audit opinions are professional, objective, and constructive, and can provide valuable reference for work ticket reviewers;

[0245] Fill in the prompt words and pass them into the advanced general model:

[0246] Fill the structured results of each review stage into the prompt word template;

[0247] Use the advanced general model to generate final review opinions to ensure the quality and integrity of the output content;

[0248] Set appropriate output format requirements to ensure that the generated audit opinions have a clear structure and are easy to read and use;

[0249] Post-processing and formatting:

[0250] Optimize the format of the audit opinions generated by the large model to ensure clear structure and highlighted key points;

[0251] Audit opinions can be converted into specific formats (such as PDF reports, Word documents or HTML pages, etc.) according to enterprise requirements;

[0252] Add appropriate visual elements, such as problem severity distribution charts, review score radar charts, etc.

[0253] Ensure that key information is highlighted to help reviewers quickly grasp the main issues of the work ticket;

[0254] Personalization:

[0255] Based on the preferences and needs of different auditors, audit opinion versions with different granularity can be provided;

[0256] Provide detailed version: contains comprehensive problem analysis and suggestions, suitable for in-depth review;

[0257] Provide a summary version: contains only key issues and conclusions, suitable for quick review;

[0258] Provide customized audit views for different roles (such as technical auditors, management auditors);

[0259] The final work ticket auxiliary review opinion usually contains the following main parts:

[0260] Basic information: work ticket number, applicant, work content, planned execution time, etc.;

[0261] Overall evaluation: an overall assessment of the quality of the work ticket, including strengths and major problems;

[0262] Critical Issues List: List all issues found by severity, with concise descriptions and recommendations;

[0263] Improvement suggestion details: Provide more detailed improvement suggestions for each issue;

[0264] Audit conclusion: Give a clear audit conclusion (pass / conditionally pass / fail) and follow-up treatment suggestions;

[0265] Through this structured, hierarchical audit feedback, auditors can quickly understand the main issues and risks of the work ticket, thereby making more accurate audit decisions. At the same time, the work ticket filler can also receive clear improvement guidance to improve the quality of the work ticket.

[0266] Comprehensive scoring function for problem severity:

[0267] ProblemSeverity(p)=max(S accuracy (p),S rationality (p),S safety (p),S compliance (p));

[0268] Among them, S accuracy (p), S rationality (p), S safety (p) and S compliance (p) represents the severity score of the problem p in the four dimensions of accuracy, rationality, security, and compliance. The max function takes the maximum value of these four scores as the final problem severity score, ensuring that the severity of the problem is determined by its most serious dimension. Scores are usually integer values ​​1-5, where 1 represents the lowest severity and 5 represents the highest severity. The overall scoring function of the work ticket is:

[0269] TotalScore=α1·C score +α2·R score +α3·Sscore +α4·ComplianceScore;

[0270] Among them, C score Represents the content accuracy score, R score Indicates the rationality score of the work plan, S score represents the security measure completeness score, and ComplianceScore represents the content compliance score. α1, α2, α3, and α4 are weight coefficients, and α1+α2+α3+α4=1.

[0271] Through this structured, hierarchical audit feedback, auditors can quickly understand the main issues and risks of the work ticket, thereby making more accurate audit decisions. At the same time, the work ticket filler can also receive clear improvement guidance to improve the quality of the work ticket.

[0272] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make more forms of equivalent embodiments based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. A method for assisting in reviewing information system operation and maintenance work tickets based on a large model, characterized in that: The following steps are involved: Obtain the work ticket content and perform text cleaning on it to form a standard work ticket content text; Construct a content accuracy review prompt, which is composed of the work ticket content and the content accuracy review instruction. This prompt is passed to the big model to obtain the content accuracy review opinion. Based on the work ticket content and the review opinions on the accuracy of the content, a work plan rationality review prompt is constructed. This prompt is composed of the work plan content in the work ticket and the work plan rationality review instructions. This prompt is passed to the big model to obtain the work plan rationality review opinions and a list of potential risk points. Based on the work plan rationality review opinions and the list of potential risk points, a safety measures completeness review prompt is constructed. This prompt is composed of the risk point analysis and safety measures content in the work ticket, along with the safety measures completeness review instructions. This prompt is then passed to the big model to obtain the safety measures completeness review opinions. Based on the work ticket content, content accuracy review opinions, work plan rationality review opinions, and safety measures completeness review opinions, a content compliance review prompt is constructed. This prompt is composed of the work ticket content and the content compliance review instructions. This prompt is passed to the big model to obtain the content compliance review opinion. The final work ticket auxiliary review opinions are generated by comprehensively considering the review opinions on content accuracy, work plan rationality, safety measures completeness and content compliance.

2. The information system operation and maintenance work ticket auxiliary review method based on a large model according to claim 1 is characterized in that: When the work ticket is a paper document, the steps of obtaining the content of the work ticket include: taking a photo or scanning the paper work ticket, performing text recognition on the obtained image, and processing the recognition result into a standard format of the work ticket content text.

3. The information system operation and maintenance work ticket auxiliary review method based on a large model according to claim 1 is characterized in that: When the work ticket is an electronic document, the step of obtaining the work ticket content includes: parsing the electronic document content, and processing the parsing result into a work ticket content text in a standard format.

4. The information system operation and maintenance work ticket auxiliary review method based on a large model according to claim 1 is characterized in that: When the work ticket is in the information management system, the step of obtaining the work ticket content includes: extracting the work ticket report content from the information management system, and processing the extraction result into a work ticket content text in a standard format.

5. The information system operation and maintenance work ticket auxiliary review method based on a large model according to claim 1 is characterized in that: The content accuracy review instruction is used to guide the large model to evaluate whether the work ticket contains key content such as staff, work supervisor, work content, risk point analysis, and safety measures.

6. The information system operation and maintenance work ticket auxiliary review method based on a large model according to claim 1 is characterized in that: The work plan rationality review instructions are used to guide the large model to evaluate the work plan from the aspects of professional terminology standardization and solution rationality and to analyze the risk points of the work plan.

7. The information system operation and maintenance work ticket auxiliary review method based on a large model according to claim 1 is characterized in that: When constructing the work plan rationality review prompt words, the following steps are also included: Obtain enterprise information system knowledge related to the work plan by searching the enterprise information system knowledge base; The acquired enterprise information system knowledge is used as known information and combined with the work plan and the work plan rationality review instructions to form prompt words.

8. The information system operation and maintenance work ticket auxiliary review method based on a large model according to claim 1 is characterized in that: The safety measures completeness review instruction is used to guide the risk point analysis in the large model assessment work ticket and the rationality and completeness of the corresponding safety measures.

9. The information system operation and maintenance work ticket auxiliary review method based on a large model according to claim 1 is characterized in that: When constructing content compliance audit prompt words, the following steps are also included: By searching the enterprise operation and maintenance specification knowledge base, obtain enterprise operation and maintenance regulations and specifications related to the work ticket content; The acquired enterprise operation and maintenance regulations and specifications are used as known information and spliced ​​with the work ticket content and content compliance review instructions to form prompt words.

10. The information system operation and maintenance work ticket auxiliary review method based on a large model according to claim 1 is characterized in that: The method further comprises the following steps: Choose different types of large models for different review stages: During the work plan rationality review phase, select a large model that has been fine-tuned with expertise in the operations and maintenance field; During the security measures completeness review phase, select large models that have been fine-tuned with domain expertise; During the content accuracy review and content compliance review stages, select the general large model.

Images