Work order auditing management method and system based on large language model, and medium

Abstract

The invention discloses a work order auditing management method and system based on a large language model and a medium. The method comprises the steps of receiving input and correspondingly matching a work order auditing rule; based on the matched work order auditing rule, analyzing and processing the input through a large language model NLP, and converting the input into a natural language rule; carrying out synonym and antonym conversion on input white words, and carrying out equal rule conversion on converted words based on a corpus word library; performing rule text and code conversion on the converted natural language rule; and carding the processed rules into SQL statements which can be identified by the system according to the fields and the judger rule base. Updating the rule knowledge base; and writing the processed SQL statements which can be identified by the program into a rule knowledge base. According to the method, non-technical personnel can adjust and modify the work order information verification rule by using a natural language by using a large language model technology, so that the intelligent level of a work order system is enhanced, the development cost of the system is reduced, the auditing efficiency of a work order process system is improved, and the method has relatively good practicability.

Description

Technical Field

[0001] The present invention belongs to the technical field of work order review management, and specifically relates to a work order review management method, system and medium based on a large language model. Background Art

[0002] The work order process is to an enterprise what blood circulation is to the human body. It is the core of organizational operations. It ensures that tasks and information can be efficiently delivered to the appropriate departments and employees, and provides the necessary resources for each work link. The lack of a clear work order process will lead to confusion in work tasks, reduced efficiency, and waste of resources. Therefore, a complete work order process system is crucial to the daily operation of an enterprise.

[0003] As the scale of the enterprise expands and the business scenarios evolve, the work order system needs to be constantly updated and upgraded. System developers usually need to perform the following four types of development work: The first is to add new features and develop work order processes that did not exist before.

[0004] The second method is to modify the existing work order process to adapt to changes in business scenarios and adjust the verification rules of work order information.

[0005] The third method is to develop external interfaces to share work order data and receive external data.

[0006] The fourth type is to conduct data statistics and analysis, and perform correlation analysis on different work orders.

[0007] For the second type of development work, the present invention provides a work order review management solution based on a large language model, which modifies the existing work order process to adapt to changes in business scenarios and adjust work order information verification rules. Summary of the invention

[0008] The purpose of the present invention is to provide a work order review management method, system and medium based on a large language model, aiming to utilize the large language model to adjust and modify the work order review rules with natural language, thereby enhancing the intelligence level of the work order system, reducing the development cost of the system, and improving the review efficiency of the work order process system.

[0009] The present invention is mainly achieved through the following technical solutions: A work order review management method based on a large language model includes the following steps: Step S101: receiving input and matching work order review rules, wherein the work order review rules include rules for manual review, joint review of related work orders, and self-review of work orders; Step S102: Based on the matched work order review rules, analyze and process the input through the large language model NLP, and convert it into natural language rules; perform synonym and antonym conversions on the vernacular in the input, and perform rule conversions such as equal rules on the converted words based on the corpus word library; Step S103: Perform rule text and code conversion on the converted natural language rules. According to the field and judgment symbol rule library, sort out the rules processed in Step S102 into SQL statements recognizable by the system; Step S104: Update the rule knowledge base, write the SQL statements recognizable by the program obtained from the processing into the rule knowledge base, so as to realize that when the work order is transferred, by querying the SQL in the rule knowledge base and supplementing the parameter values required in the SQL, the normal operation of the SQL is realized.

[0010] To better implement the present invention, further, in the step S103, through the TEXT2API technology and the TEXT2SQL technology respectively, convert the natural language rules into API calls and SQL statements, so as to seamlessly connect the text input and the system operation.

[0011] To better implement the present invention, further, in the step S101, the inputs corresponding to the rules for manual review include: "Work orders sent by XX must be reviewed by me"; "Work orders of the contract type must be manually reviewed by me offline"; "All work orders during the XX time period require my manual review".

[0012] To better implement the present invention, further, in the step S102, for the rule of manual review, "Work orders sent by XX must be reviewed by me", through the processing of the large language model NLP, it is split into: type = update, sender = XX, reviewer = current logged-in user; "Work orders of the contract type must be manually reviewed by me offline", through the processing of the large language model NLP, it is split into: type = update, work order type = contract work order, reviewer = current logged-in user; "All work orders during the XX time period require my manual review", through the processing of the large language model NLP, it is split into: type = update, work order creation time > XX time, work order creation time < XX time, reviewer = current logged-in user.

[0013] To better implement the present invention, further, in the step S101, the inputs corresponding to the rules for joint review of associated work orders include: "The payment link of the reimbursement process can be completed only after the review of the accrual process is passed" and "The total project amount set in the project budget process cannot be higher than the business opportunity budget of this project in the business opportunity process".

[0014] To better implement the present invention, further, in step S102, for the rule of joint review of associated work orders, "the payment link of the reimbursement process can be completed only when the accrual process review is passed", through the processing of the large language model NLP, it is split into: type = query, pre - process = accrual process, pre - process status = review passed, current process = reimbursement process, process link = payment link, review action = click to complete; "In the project budget process, the total project amount set cannot be higher than the business opportunity budget of this project in the business opportunity process", through the processing of the large language model NLP, it is split into: type = query, pre - process = business opportunity process, field = business opportunity budget, current process = project budget process, process field = total project amount,

cannot be higher than

&lt;=

[0015] To better implement the present invention, further, in step S101, the inputs corresponding to the rule of self - review of work orders include: "The accommodation expenses in the reimbursement process cannot be greater than XX yuan per day. If it is greater than XX yuan, a reason needs to be written. The format of the reason is: Due to XX reason, there is no room that meets the accommodation standard, so the room with the lowest room rate is selected, and each room exceeds the budget by XX yuan per day", "The invoice number entered needs to be the same as the invoice number in the uploaded attachment", "When the budget is less than XX yuan, it does not need to be reviewed by me", "When the total amount is greater than XX yuan, give me a prompt when I click to agree, reminding me to check carefully".

[0016] To better implement the present invention, further, in step S102, for the rule of self - review of work orders, "The accommodation expenses in the reimbursement process cannot be greater than XX yuan per day. If it is greater than XX yuan, a reason needs to be written. The format of the reason is: Due to XX reason, there is no room that meets the accommodation standard, so the room with the lowest room rate is selected, and each room exceeds the budget by XX yuan per day", through the processing of the large language model NLP, it is split into: type = query, work order type = reimbursement process,

cannot be greater than

&lt;=

greater than

&gt;

the same as

=

less than

&lt;

[0017] The present invention is mainly achieved through the following technical solutions: A work order review management system based on a large language model, based on the above-mentioned work order review management method based on a large language model, characterized in that it includes an input and receiving module, a natural language processing module, a TEXT2API module and a TEXT2SQL module, a code conversion module, a database management module, and a knowledge base management module; The natural language processing module is used to process user input, perform language understanding, semantic analysis, synonym and antonym conversion, and rule conversion based on a pre-made corpus word library; The TEXT2API module is used to convert natural language rules into API calls; The TEXT2SQL module is used to further convert the rules into SQL statements; The code conversion module is used to sort out and convert rules based on the field and predicate rule library; The database management module is used to store, classify and retrieve SQL rules; The knowledge base management module is used to maintain and update the rule knowledge base.

[0018] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned work order review management method based on a large language model.

[0019] The beneficial effects of the present invention are as follows: (1) The present invention utilizes large language model technology to enable non-technical personnel to use natural language to adjust and modify the work order information verification rules, thereby enhancing the intelligence level of the work order system, reducing system development costs, and improving the audit efficiency of the work order process system.

[0020] (2) The system can realize the automatic conversion from natural language to SQL statements, and use large language models such as GPT to understand and process natural language instructions input by users. Through TEXT2API and TEXT2SQL technologies, the system converts natural language rules into API calls and database query languages, thereby seamlessly connecting text input with system operations. In addition, the system intelligently generates work order review rules, dynamically manages the rule base, supports multiple types of work order review processes, and has a built-in prefabricated corpus vocabulary for rule conversion to ensure the accuracy and consistency of the rules. The present invention realizes the intelligence of the work order review process, reduces manual intervention, improves the review efficiency and accuracy, and brings innovation and breakthroughs to the field of work order management.

[0021] (3) The present invention improves efficiency: By using a large language model and natural language processing (NLP) technology, non-developers can directly use natural language to define work order review rules, thereby reducing dependence on professional developers and speeding up the creation and modification of work order review rules. The present invention reduces costs: Automated work order review rule creation reduces the cost of system development and maintenance. Enterprises no longer need to frequently invest development resources to adjust the work order system, thereby saving manpower and financial resources. The present invention enhances user experience: users can express work order review rules in natural language, which is closer to daily communication and improves the user's interactive experience with the system.

[0022] (4) The present invention improves accuracy: it can intelligently extract key information from natural language and generate correct SQL statements, reducing human errors and improving the accuracy of work order review. The present invention improves the security and reliability of the system: automated rule generation and review reduces human intervention, reduces security risks caused by human errors, and enhances the security and reliability of the system. The present invention is flexible and scalable: as business needs change, work order review rules can be quickly adjusted, and the system has good flexibility and scalability and can adapt to changing business scenarios.

[0023] (5) The present invention reduces resource waste: through the automated and intelligent work order review process, it reduces resource waste caused by chaotic and disordered processes, ensuring efficient use of resources. The present invention improves decision-making quality: it can automatically review work orders based on the comparison of work order content and rules, helping decision makers to quickly obtain accurate information and improve decision-making quality. The present invention enhances the intelligence level of the system: by integrating advanced AI technology, the intelligence level of the work order management system is improved, providing enterprises with a more intelligent and automated solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The present invention is a flowchart of a work order review management method based on a large language model. DETAILED DESCRIPTION

[0025] Embodiment 1: A work order review management system based on a large language model. Based on the large model TEXT2API, the work order content, work order fields and review rules are converted into rule verification conditions that can be used by the system in vernacular. The subsequent system can compare the work order content with the rules to realize automatic review of the work order. Specifically including: Input and receiving module: performs natural language input processing and can receive and display the user's work order configuration rule input through user interface (UI) design.

[0026] Natural language processing module: performs language understanding, semantic analysis, synonym and antonym conversion, and rule conversion based on pre-made corpus word library.

[0027] TEXT2API module: converts natural language rules into API calls; TEXT2SQL module: further converts the rules into database query language (SQL) statements; Code conversion module: used to sort and convert rules based on the field and predicate rule base.

[0028] Database management module: used to store, classify and retrieve SQL rules; Knowledge base management module: used to maintain and update the rule knowledge base.

[0029] The system can realize the automatic conversion of natural language to SQL statements, and use large language models such as GPT to understand and process natural language instructions input by users. Through TEXT2API and TEXT2SQL technology, the system converts natural language rules into API calls and database query languages, thereby seamlessly connecting text input and system operations. In addition, the system intelligently generates work order review rules, dynamically manages the rule base, supports multi-type work order review processes, and has a built-in prefabricated corpus vocabulary for rule conversion to ensure the accuracy and consistency of the rules. The present invention realizes the intelligence of the work order review process, reduces manual intervention, improves the review efficiency and accuracy, and brings innovation and breakthroughs to the field of work order management.

[0030] Embodiment 2: A work order review management method based on a large language model, such as Figure 1 As shown, the following steps are included: S101: User inputs work order configuration rules In step S101, the system may receive a user's work order request through a GTP dialogue receiving module.

[0031] The work order review rules are divided into three types: manual review, joint review of related work orders, and work order self-review.

[0032] in: 1. Manual review: For example, input: "The work order sent by XXX must be reviewed by me", "Contract-type work orders must be reviewed offline by me", "All work orders in a certain time period require manual review by me", etc.

[0033] 2. Joint review of related work orders: For example, input: "The payment process can only be completed in the reimbursement process after the accrual process is reviewed and approved", "The total amount of the project in the project budget process cannot be higher than the business opportunity budget of this project in the business opportunity process".

[0034] 3. Self-examination of work orders: For example, input: "The accommodation fee in the reimbursement process cannot be greater than 300 yuan per day. If it is greater than 300 yuan, a reason must be written. The format of the reason is as follows: Due to XX reason, there are no rooms that meet the accommodation standards, so the room with the lowest price is selected. Each room exceeds the budget of XX yuan per day", "The entered invoice number must be consistent with the invoice number in the uploaded attachment", "When the budget is less than XX yuan, it does not need to be reviewed by me", "When the total amount is greater than XX yuan, when I click to agree, I will be prompted to check it carefully."

[0035] S102: Large Language Model NLP Processing In step S102, the system uses a large language model to convert the user's input vernacular into synonyms and antonyms, which are used by the system to convert into recognizable SQL statements for data preparation. The conversion words are converted according to the rules of the pre-made corpus word library.

[0036] For example: 1. Manual review: 1.1. "The work order sent by someone must be reviewed by me" is processed by the system through NLP and split into: type = update, issuer = someone, reviewer = current logged in user; 1.2. "Contract work orders must be reviewed offline by me" is processed by the system through NLP and split into: type = update, work order type = contract work order, reviewer = current logged in user; 1.3. "All work orders in a certain time period require manual review by me" The system is processed through NLP and split into: type = update, work order creation time > a certain time, work order creation time < a certain time, reviewer = current logged in user.

[0037] 2. Joint review of related work orders: 2.1. The system processes through NLP the statement that "only when the accrual process audit is passed can the reimbursement process complete the payment link", and breaks it down into: type = query, previous process = accrual process, previous process status = audit passed, current process = reimbursement process, process link = payment link, audit action = click to complete.

[0038] 2.2. The system processes through NLP the statement that "in the project budget process, the total project amount set cannot be higher than the business opportunity budget of this project in the business opportunity process", and breaks it down into: type = query, previous process = business opportunity process, field = business opportunity budget, current process = project budget process, process field = total project amount, [cannot be higher than] is converted to [<=], process field <= business opportunity budget.

[0039] 3. Work order self - review category: 3.1. The system processes through NLP the statement that "the accommodation expense in the reimbursement process cannot be greater than 300 yuan per day. If it is greater than 300 yuan, a reason needs to be written. The format of the reason is as follows: Due to XX reasons, there are no rooms meeting the accommodation standard, so the room with the lowest price is selected, and the budget of each room exceeds XX yuan per day", and breaks it down into: type = query, work order type = reimbursement process, [cannot be greater than] is converted to [<=], [greater than] is converted to [>], accommodation expense field > 300, remarks field like 'Due to XX reasons, there are no rooms meeting the accommodation standard, so the room with the lowest price is selected, and the budget of each room exceeds XX yuan per day'.

[0040] 3.2. The system processes through NLP the statement that "the invoice number entered needs to be the same as the invoice number in the uploaded attachment", and breaks it down into: type = query, fields to be compared = recognized invoice number text in attachment, [the same as] is converted to [=], invoice number field = fields to be compared.

[0041] 3.3. The system processes through NLP the statement that "when the budget is less than XX yuan, it does not need to be audited by me", and breaks it down into: type = query, [less than] is converted to [<], budget field < XX yuan, auditor!= current logged - in user.

[0042] 3.4. The system processes through NLP the statement that "when the total amount is greater than XX yuan, when I click to agree, give me a prompt, reminding me that I need to check carefully", and breaks it down into: type = query, [greater than] is converted to [>], total amount field > XX yuan, audit status = submitted, prompter = current user, prompt content = The total amount is greater than XX yuan, and it needs to be checked carefully during the audit.

[0043] S103: Rule text - code conversion In step S103, based on the results of NLP processing, the system sorts out the rules processed by NLP into SQL statements recognizable by the system through the field and judgment symbol rule library.

[0044] For example: 1. Manual review: 1.1 Type = Update, Issuer = Someone, Reviewer = Current Login User Convert the code according to the fields and SQL rule base: If type = Update, use the update set statement. The escaped SQL this time is: update work order table set Reviewer = Current Login User where Issuer = Someone.

[0045] 1.2 Type = Update, Work Order Type = Contract Work Order, Reviewer = Current Login User Convert the code according to the fields and SQL rule base: If type = Update, use the update set statement. The escaped SQL is: update work order table set reviewer = current login user where work order type = contract work order.

[0046] 1.3 Type = Update, Work Order Creation Time > Certain Time, Work Order Creation Time < Certain Time, Reviewer = Current Login User Convert the code according to the fields and SQL rule library: If type = Update, use the update set statement. The escaped SQL is: update work order table set reviewer = current login user where work order creation time > certain time and work order creation time < certain time.

[0047] 2. Joint review of related work orders: 2.1 Type = Query, Pre-process = Provisioning process, Pre-process status = Approved, Current process = Reimbursement process, Process link = Payment link, Approval action = Click to complete. Convert the code according to the field and SQL rule library: If type = Query, use the select from statement. The escaped SQL is: SELECT * FROM reimbursement process table WHERE EXISTS( SELECT 1 FROM accrual flow chart WHERE accrual process table. Pre-process ID = reimbursement process table. Pre-process ID AND accrual process table. Pre-process status = 'approved' ) AND reimbursement process table. Process link = 'payment link' AND Reimbursement Process Table. Review Action = 'Click to Complete'.

[0048] 2.2 Type = Query, Pre-process = Opportunity process, Field = Opportunity budget, Current process = Project budget process, Process field = Project total amount, [Cannot be higher than] converted to [<=], Process field <= Opportunity budget Convert the code based on the field and SQL rule library: If type = Query, use the select from statement. The escaped SQL is: SELECT * FROM Project Budget Flow Chart WHERE Project Budget Flow Chart. Total Project Amount>( SELECT Business Opportunity Budget FROM Business Opportunity Flow Chart WHERE Business Opportunity Flow Table.Project ID = Project Budget Flow Table.Project ID ) AND Project Budget Flow Chart.Project ID=Business Opportunity Flow Chart.Project ID.

[0049] 3. Work order self-review: 3.1 Type = Query, Work Order Type = Reimbursement Process, [Cannot be greater than] converted to [<=], [Greater than] converted to [>], Accommodation Fee Field>300, Remarks Field like 'Due to XX reasons, there are no rooms that meet the accommodation standards. Select the room with the lowest room price. Each room exceeds the budget by XX yuan per day'. Convert the code according to the field and SQL rule library: If type = query, use the select from statement. The escaped SQL is: SELECT * FROM Reimbursement Process Table WHERE Reimbursement Process Table. Accommodation Fee>300 AND(Reimbursement Process Table. Note NOT LIKE 'Due to XX reason, there are no rooms that meet the accommodation standards, so we choose the room with the lowest room rate. Each room exceeds the budget by XX yuan per day' OR Reimbursement Process Table. Note IS NULL).

[0050] 3.2 Type = query, field to be compared = identify the text of the attachment invoice number, [Consistent] converted to [=], invoice number field = field to be compared. Convert the code according to the field and SQL rule library: If type = query, use the select from statement. The escaped SQL is: SELECT * FROM Invoice Information Table WHERE Invoice Information Table. Invoice Number<>( SELECTAttachment Invoice Number FROM Attachment Table WHERE Attachment Table.Attachment ID = Invoice Information Table.Attachment ID )。

[0051] 3.3 Type = Query, convert

less than

&lt;

[0052] 3.4 Type = Query, convert

greater than

&gt;

[0053] S104: Rule Knowledge Base Management In step S103, the system classifies the SQL rules into types such as Query and Update, and writes the SQL recognizable by the program into the rule knowledge base. When the system process application performs work order transfer, it retrieves the SQL in the rule knowledge base and accurately supplements the parameter values required in the SQL to achieve the normal operation of the SQL. The work order system can perform the next relevant operations according to the result returned by the SQL execution.

[0054] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A work order review management method based on a large language model, characterized in that: It includes the following steps: Step S101: Receive the input and match the work order review rules correspondingly. The work order review rules include rules for manual review, associated work order joint review, and work order self-review; Step S102: Based on the matched work order review rules, analyze and process the input through the large language model NLP to convert it into natural language rules; perform synonym and antonym conversions on the vernacular in the input, and perform rule conversions such as those based on the corpus word library on the converted words; Step S103: Perform rule text and code conversion on the converted natural language rules. According to the field and judgment symbol rule library, sort out the rules processed in Step S102 into SQL statements recognizable by the system; Step S104: Update the rule knowledge base, write the SQL statements recognizable by the program obtained from the processing into the rule knowledge base, so as to realize that when the work order is transferred, by querying the SQL in the rule knowledge base and supplementing the parameter values required in the SQL, the normal operation of the SQL is realized.

2. A work order review management method based on a large language model according to claim 1, characterized in that: In Step S103, through the TEXT2API technology and the TEXT2SQL technology respectively, convert the natural language rules into API calls and SQL statements, so as to seamlessly connect the text input and the system operation.

3. A work order review management method based on a large language model according to claim 1 or 2, characterized in that: In Step S101, the inputs corresponding to the rules for manual review include: "Work orders sent by XX must be reviewed by me"; "Work orders of the contract type must be reviewed offline by me"; "All work orders during the XX time period require my manual review".

4. A work order review management method based on a large language model according to claim 3, characterized in that: In Step S102, for the rule of manual review, "Work orders sent by XX must be reviewed by me", through the processing of the large language model NLP, it is split into: type = update, sender = XX, reviewer = current logged-in user; "Work orders of the contract type must be reviewed offline by me", through the processing of the large language model NLP, it is split into: type = update, work order type = contract work order, reviewer = current logged-in user; "All work orders during the XX time period require my manual review", through the processing of the large language model NLP, it is split into: type = update, work order creation time > XX time, work order creation time < XX time, reviewer = current logged-in user.

5. A work order review management method based on a large language model according to claim 1 or 2, characterized in that: In Step S101, the inputs corresponding to the rules for associated work order joint review include: "The payment link of the reimbursement process can be completed only when the accrual process review is passed" and "The total project amount set in the project budget process cannot be higher than the business opportunity budget of this project in the business opportunity process".

6. The work order review management method based on a large language model according to claim 5 is characterized in that: In Step S102, for the rule of associated work order joint review, "The payment link of the reimbursement process can be completed only when the accrual process review is passed", through the processing of the large language model NLP, it is split into: type = query, preprocess = accrual process, preprocess status = review passed, current process = reimbursement process, process link = payment link, review action = click to complete; "The total project amount set in the project budget process cannot be higher than the business opportunity budget of this project in the business opportunity process". After being processed by the large language model NLP, it is split into: type = query, pre - process = business opportunity process, field = business opportunity budget, current process = project budget process, process field = total project amount, 【cannot be higher than】 is converted to 【<=】, process field <= business opportunity budget.

7. A work order review management method based on a large language model according to claim 1 or 2, characterized in that: In the said step S101, the inputs corresponding to the rules of the work order self - review include: "The accommodation expense in the reimbursement process cannot be greater than XX yuan per day. If it is greater than XX yuan, a reason needs to be written. The format of the reason is: Due to XX reason, there is no room that meets the accommodation standard, so the room with the lowest price is selected, and each room exceeds the budget by XX yuan per day", "The invoice number entered needs to be the same as the invoice number in the uploaded attachment", "When the budget is less than XX yuan, it does not need to be reviewed by me", "When the total amount is greater than XX yuan, when I click to approve, give me a prompt reminding me to check carefully".

8. A work order review management method based on a large language model according to claim 7, characterized in that: In the said step S102, for the rules of the work order self - review, "The accommodation expense in the reimbursement process cannot be greater than XX yuan per day. If it is greater than XX yuan, a reason needs to be written. The format of the reason is: Due to XX reason, there is no room that meets the accommodation standard, so the room with the lowest price is selected, and each room exceeds the budget by XX yuan per day". After being processed by the large language model NLP, it is split into: type = query, work order type = reimbursement process, 【cannot be greater than】 is converted to 【<=】, 【greater than】 is converted to 【>】, accommodation expense field > 300, note field like 'Due to XX reason, there is no room that meets the accommodation standard, so the room with the lowest price is selected, and each room exceeds the budget by XX yuan per day'; "The invoice number entered needs to be the same as the invoice number in the uploaded attachment". After being processed by the large language model NLP, it is split into: type = query, field to be compared = recognized invoice number text in the attachment, 【the same as】 is converted to 【=】, invoice number field = field to be compared; "When the budget is less than XX yuan, it does not need to be reviewed by me". After being processed by the large language model NLP, it is split into: type = query, 【less than】 is converted to 【<】, budget field < XX yuan, reviewer!= current logged - in user; "When the total amount is greater than XX yuan, when I click to approve, give me a prompt reminding me to check carefully". After being processed by the large language model NLP, it is split into: type = query, 【greater than】 is converted to 【>】, total amount field > XX yuan, review status = submitted, prompter = current user, prompt content = The total amount is greater than XX yuan, and it is necessary to check carefully during review.

9. A work order review management system based on a large language model, based on the work order review management method based on a large language model according to any one of claims 1 to 8, characterized in that: It includes an input and reception module, a natural language processing module, a TEXT2API module, a TEXT2SQL module, a code conversion module, a database management module, and a knowledge base management module; The said natural language processing module is used to process the user's input, perform language understanding, semantic analysis, conversion of synonyms and antonyms, and rule conversion based on a pre - fabricated corpus thesaurus; The said TEXT2API module is used to convert natural language rules into API calls; The TEXT2SQL module is used to further convert the rules into SQL statements; The code conversion module is used to sort out and convert rules based on the field and predicate rule library; The database management module is used to store, classify and retrieve SQL rules; The knowledge base management module is used to maintain and update the rule knowledge base.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the work order review management method based on a large language model described in any one of claims 1 to 8 is implemented.

Images