Contract content pre-auditing method and device based on large language model, equipment and medium

Abstract

The invention discloses a contract content pre-auditing method and device based on a large language model, equipment and a medium. The method comprises the following steps: extracting contract content of a contract to be pre-reviewed; extracting an auditing parameter in at least one work order corresponding to the contract to be pre-audited, and determining an auditing rule based on the auditing parameter; combining the contract content, the auditing parameter, the auditing rule and a preset cue word template to obtain a pre-auditing cue word; and inputting the pre-auditing prompt word into the large language model to obtain a pre-auditing result. Through the above mode, contract pre-auditing can be realized by using the large language model, the pre-auditing result is obtained, auditing personnel only need to verify the pre-auditing result, the auditing personnel can be assisted to quickly approve the contract, the working efficiency of the auditing personnel is improved, auditing errors caused by experience and subjective judgment of the auditing personnel are avoided, and the success rate of contract auditing is improved. And the auditing accuracy is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device and medium for pre-screening contract content based on a large language model. Background Technology

[0002] With the rapid development of company and business operations, the number and content of contracts have increased rapidly. During the work order approval process, companies typically need to conduct rule-based reviews and checks on uploaded contracts. Currently, these reviews are usually performed manually by reviewers. However, these reviews are cumbersome and redundant, and the repetitive daily work reduces the efficiency of reviewers. Furthermore, the review results are easily influenced by the reviewers' experience and subjective judgment, making it difficult to guarantee the accuracy and comprehensiveness of the review. Summary of the Invention

[0003] This application mainly provides a method, apparatus, equipment and medium for pre-examination of contract content based on a large language model, in order to solve the problems of low efficiency and accuracy caused by the cumbersome manual contract review process in the prior art.

[0004] To address the aforementioned technical problems, this application adopts the following technical solution: a method for pre-screening contract content based on a large language model. This method includes: Extract the contract content of the contract to be pre-examined; Extract the audit parameters from at least one work order corresponding to the contract to be pre-audited, and determine the audit rules based on the audit parameters; The contract content, the review parameters, the review rules, and the preset prompt word template are combined to obtain the pre-review prompt words; The pre-screening prompts are input into the large language model to obtain the pre-screening results.

[0005] In one optional embodiment of this application, extracting the contract content of the contract to be pre-examined includes: Based on a pre-defined file extraction library, annotations and modifications are removed from the contract to be pre-reviewed, and the modified contract content is extracted.

[0006] In one optional embodiment of this application, the step of extracting audit parameters from at least one work order corresponding to the contract to be pre-audited, and determining audit rules based on the audit parameters, includes: In response to the work order corresponding to the contract to be pre-reviewed, the review parameters in the work order are extracted, and the review rules are determined based on the review parameters; In response to the multiple work orders corresponding to the contract to be pre-reviewed, the review parameters in each work order are extracted asynchronously, and the corresponding review rules are determined based on each review parameter.

[0007] In one optional embodiment of this application, the step of combining the contract content, the review parameters, the review rules, and the preset prompt word template to obtain the pre-review prompt words includes: In response to a work order corresponding to the contract to be pre-reviewed, the contract content, the review parameters and the review rules are converted into a marked document, and the marked document is combined with the preset prompt word template to obtain the pre-review prompt word; In response to the multiple work orders corresponding to the contract to be pre-reviewed, the contract content and the corresponding review parameters and review rules are converted into multiple sets of marked documents, and the multiple sets of marked documents are combined with the preset prompt word template to obtain multiple pre-review prompt words.

[0008] In an optional embodiment of this application, after combining multiple sets of marked documents with preset prompt word templates to obtain multiple pre-screening prompt words, the method further includes: Multiple pre-screening prompts are asynchronously input into a large language model to obtain multiple pre-screening results.

[0009] In an optional embodiment of this application, after inputting the pre-screening prompts into a large language model and obtaining the pre-screening result, the method further includes: The structured preliminary review results are converted into a table format and fed back to the reviewers, and the preliminary review results are recorded in the review log of the corresponding work order.

[0010] To address the aforementioned technical problems, another technical solution adopted in this application is: providing a contract content pre-screening device based on a large language model, comprising: The content extraction module is used to extract the contract content of contracts to be pre-examined; The work order extraction module is used to extract the audit parameters from at least one work order corresponding to the contract to be pre-approved, and to determine the audit rules based on the audit parameters; The prompt word combination module is used to combine the contract content, the review parameters, the review rules and the preset prompt word template to obtain the pre-review prompt words; The pre-screening module is used to input the pre-screening prompts into the large language model to obtain the pre-screening results.

[0011] To solve the above-mentioned technical problems, another technical solution adopted in this application is: to provide a computer device, including a memory, a processor and a computer program stored in the memory, characterized in that the processor executes the computer program to implement the steps of the above-mentioned contract content pre-examination method based on a large language model.

[0012] To solve the above-mentioned technical problems, another technical solution adopted in this application is: to provide a storage medium on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements the steps of the above-mentioned contract content pre-examination method based on a large language model.

[0013] To solve the above-mentioned technical problems, another technical solution adopted in this application is: to provide a computer program product, including a computer program, characterized in that, when the computer program is executed by a processor, it implements the steps of the above-mentioned contract content pre-examination method based on a large language model.

[0014] The beneficial effects of this application are as follows: Unlike existing technologies, this application discloses a method, apparatus, device, and medium for contract content pre-review based on a large language model. This method extracts the contract content of the contract to be pre-reviewed and combines it with review parameters and rules extracted from the corresponding work order. Pre-review prompts are generated using a preset prompt template and input into the large language model. The large language model is then used to perform contract pre-review, obtaining the pre-review result. Reviewers only need to verify the pre-review result, which assists reviewers in quickly approving contracts, improves their work efficiency, avoids review errors caused by reviewers' experience and subjective judgment, and improves review accuracy. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 This is a flowchart illustrating an embodiment of the contract content pre-examination method based on a large language model provided in this application; Figure 2 This is a flowchart illustrating the generation of pre-review prompt words in an embodiment of the contract content pre-review method based on a large language model provided in this application. Figure 3 This is a schematic diagram of an embodiment of the contract content pre-screening device based on a large language model provided in this application; Figure 4 This is a schematic diagram of the structure of an embodiment of the storage medium provided in this application; Figure 5 This is a schematic diagram of the structure of an embodiment of the computer device provided in this application; Figure 6 This is a schematic diagram of the structure of an embodiment of the computer program product provided in this application. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0017] The terms "first," "second," and "third" used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0018] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0019] This application provides a method for pre-screening contract content based on a large language model. (See reference...) Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the contract content pre-screening method based on a large language model provided in this application. The contract content pre-screening method based on a large language model includes: S10: Extract the contract content of the contract to be pre-examined.

[0020] The contract document serves as the starting point and basis for the work order approval process. It is a standardized input file submitted by the work order initiator to initiate a specific request or task related to the contract. As an attachment submitted by the work order initiator, the contract document is usually a "draft" or "draft for review." To ensure standardization and efficiency of the process, the submission format of the contract document is usually specified as .docx (to facilitate subsequent revisions, annotations, and version control by subsequent processors) or .pdf (usually used for submitting a finalized version or a version provided by the other party, aiming to maintain a fixed format and prevent unintentional modification during circulation).

[0021] Large Language Models (LLMs) are deep learning models trained on massive amounts of text data, capable of understanding and generating human language. Currently, mainstream LLM models include the GPT series (such as GPT-3.5 and GPT-4), the Llama series (such as Llama 2 and Llama 3), the DeepSeek series, and the Qwen series.

[0022] This application introduces a large language model to assist reviewers in achieving rapid review. When approving work orders, reviewers can select contracts requiring pre-review (i.e., contracts awaiting pre-review) through a visual interface to enter the automated pre-review process. The large language model can be flexibly selected according to actual needs, such as the deepseek-r1 model or the Qwen3 model; no specific limitation is made here.

[0023] In this application, the content of the contract to be pre-examined is extracted, including: Based on a pre-defined document extraction library, annotations and modifications are removed from the contracts awaiting pre-review, and the modified contract content is extracted.

[0024] In this application, since the contract document may undergo multiple revisions during the approval process, the revised contract document will contain numerous annotations and modifications. These annotations and modifications are redundant information that does not need to be considered in the preliminary review of the contract. Therefore, when extracting the contract content, it is necessary to remove the annotations and modifications in the contract to be reviewed and only extract the revised contract content for preliminary review.

[0025] Based on a pre-defined file extraction library, the contract content of the contract to be pre-examined is extracted, ensuring that only the modified contract content is extracted during the content extraction process. The pre-defined file extraction library can be any library capable of extracting the main content and metadata (such as comments and revisions) separately, such as the python mammoth library, python-docx library, and python-docx2txt library. The library can be flexibly adjusted according to needs, and no specific limitation is made here. This application uses the python mammoth library as an example for illustration.

[0026] The contract content extracted using the Python mammoth library is an HTML-formatted string. HTML format can accurately distinguish semantic structures such as headings, paragraphs, lists, and tables, and retains basic formatting such as bold, italics, underlines, and hyperlinks. It is also an ideal intermediate format for conversion to other formats (such as PDF, EPUB, and Markdown), providing strong flexibility and compatibility.

[0027] S20: Extract audit parameters from at least one work order corresponding to the contract to be pre-audited, and determine audit rules based on the audit parameters.

[0028] In this application, audit parameters are extracted from at least one work order corresponding to the contract to be pre-audited, and audit rules are determined based on the audit parameters, including: In response to a work order corresponding to a contract pending pre-review, the review parameters in the work order are extracted, and the review rules are determined based on the review parameters; In response to multiple work orders corresponding to contracts awaiting pre-review, the review parameters of each work order are extracted asynchronously, and the corresponding review rules are determined based on each review parameter.

[0029] A contract typically corresponds to one or more work orders. For example, an annual IT operations and maintenance contract may create hundreds of related work orders of different types, such as "fault reporting," "routine inspection," and "system upgrade," which are stored in a work order database, based on the contract terms. The types of parameters that need to be verified differ for different types of work orders. Therefore, when a contract to be pre-approved corresponds to multiple work orders, the verification parameters for each work order need to be extracted separately.

[0030] When a contract awaiting pre-review corresponds to a work order, the audit parameters that need to be verified in that work order are extracted from the work order database. For example, the audit parameters that need to be verified in a certain work order include: amount, date, term, tax rate, customer, partner, customer payment method, company involved, and contract document content.

[0031] For the extracted audit parameters, corresponding audit rules are formulated, which define the content that needs to be audited by the large language model. The audit rules can be dynamically adjusted based on the audit parameters or audit requirements. For example, if the audit parameters include the amount {amount}, then audit rules such as "verify whether the amount parameter is consistent with the total cost amount in the contract content" can be formulated for the amount audit.

[0032] The review requirements vary depending on the type of contract (such as commercial contracts or legal contracts). The review parameters and corresponding review rules can be dynamically adjusted according to the review requirements of the contract type.

[0033] In some embodiments, an audit rule database can be constructed, in which audit rules corresponding to common audit parameters are stored in advance. After the audit parameters are extracted, the corresponding audit rules are extracted from the audit rule database based on the audit parameters.

[0034] In one specific embodiment, the audit parameters extracted from a work order include: amount, date, term, tax rate, customer, partner, customer payment method, company involved, and contract document content.

[0035] The audit rules corresponding to the audit parameters extracted from this work order include: 1. Verify that the amount parameters match the total cost amount in the contract; 2. Check if the contract contains the amount in words, and whether the amount in words matches the amount in figures; 3. Verify that the tax rate parameters are consistent with the tax rates in the contract; 4. Verify that the term (number of months) parameter is consistent with the validity period agreed in the contract (including gifts); 5. Verify whether the contract specifies the exact payment schedule; 6. Verify that the customer's payment method parameters are consistent with the agreement in the contract. The judgment criteria are as follows: prepayment / postpayment; 7. Check whether the contract includes a refund clause and whether the refund period is more than 30 days. 8. Check if the contract stipulates any additional complimentary services; 9. Verify whether the contract involves the purchase of two or more products, and if so, which products are involved; 10. Verify whether the contract requires our company to pay a performance bond or pledge, and what the amount is.

[0036] When a contract awaiting pre-review corresponds to multiple work orders, the review parameters in each work order are extracted asynchronously. That is, the review parameters in each work order are extracted separately, and the corresponding review rules for each work order are determined based on the review parameters extracted from each work order. In other words, one work order corresponds to a set of review parameters and a set of review rules. The parameter extraction between work orders is asynchronous and does not interfere with each other.

[0037] By asynchronously extracting multiple work orders from a contract, subsequent multi-threaded review is facilitated, thus improving overall review efficiency.

[0038] S30: Combine the contract content, review parameters, review rules and preset prompt word templates to obtain pre-review prompt words.

[0039] In this application, contract content, review parameters, review rules, and preset prompt word templates are combined to obtain pre-review prompt words, including: In response to a work order corresponding to a contract awaiting pre-review, the contract content, review parameters, and review rules are converted into a marked document, and the marked document is combined with a preset prompt word template to obtain the pre-review prompt words; In response to multiple work orders corresponding to contracts awaiting pre-review, the contract content, along with the corresponding review parameters and rules, is converted into multiple sets of marked documents. These marked documents are then combined with preset prompt templates to generate multiple pre-review prompts.

[0040] In this application, the extracted contract content, review parameters, and review rules can be converted into markdown document format to facilitate their use as prompts for the large language model. This format conversion can be achieved using the Python Beautiful Soup library. The converted markdown documents are then populated into a preset prompt template to obtain pre-review prompts that can be input into the large language model. The preset prompt template can be configured to set the large language model as a contract analysis assistant, and pre-review constraints can be flexibly set according to requirements (such as tax rate calculation formulas, methods for distinguishing between the client's name and the invoice name, methods for comparing validity periods, etc.). The preset prompt template also needs to reserve space for filling in the contract content, review parameters, and review rules.

[0041] In one embodiment, a preset prompt word template may be as follows: You are a contract analysis assistant who is always loyal to the client's input, verifying the information based on the client's input: 1. All extracted information must be based on the verification parameters and the actual content in the contract documents; 2. Key information extracted from the contract document should be accurate and ensure that the information actually exists in the contract document; 3. When relevant information cannot be extracted from the document, this should be clearly stated; 4. Maintain objectivity and impartiality during the process, and do not introduce personal bias; 5. Pay attention to the distinction between the name of Party A and the name of Party A on the invoice. The name of Party A refers to the company name of Party A in the contract, while the name of Party A on the invoice refers to the invoice header of Party A in the contract. 6. Strictly compare amounts based on numerical values, ignoring format differences; 7. If the contract does not specify a tax rate, calculate the tax rate yourself. The formula for calculating the tax rate is: Tax rate = Amount including tax / Amount excluding tax; 8. If there are multiple amounts in capital letters in the contract, all of them need to be checked against the amounts in lowercase. For example, if the amount in lowercase is 150,000 and the amount in capital letters is One Hundred and Fifty Thousand Yuan Only (not One Hundred and Fifty Thousand Yuan Only), then the check result is correct. 9. If the contract does not involve performance bonds or pledge deposits, the verification result is incorrect; 10. The Chinese capital numerals are as follows: One (1), Two (2), Three (3), Four (4), Five (5), Six (6), Seven (7), Eight (8), Nine (9), Ten (10), Hundred (100), Thousand (1000), Ten Thousand (10000), Hundred Million (100000000).

[0042] Please check the content in the contract document according to the following items, and force the verification result to be output in Chinese for each item.

[0043] Audit rules: Fill in the extracted audit rules.

[0044] Verification parameters: Fill in the extracted audit parameters, and fill in the audit content in the parameter item contract document content {context}.

[0045] When the contract to be pre-audited corresponds to a single work order, after converting the contract content, audit parameters, and audit rules into a markup document format, combine them with the preset prompt template to generate a complete pre-audit prompt; when the contract to be pre-audited corresponds to multiple work orders, convert the contract content, the corresponding audit parameters, and audit rules into a markup document format. When combining with the preset prompt template, combine them in the form of one contract content, one set of audit parameters, and one set of audit rules as one set of markup documents with the preset prompt template to obtain a complete pre-audit prompt. Asynchronously combine multiple sets of markup documents with the preset prompt template respectively to obtain multiple pre-audit prompts, and each pre-audit prompt does not interfere with each other.

[0046] In this application, the preset prompt template can be flexibly reused by replacing the three core contents of contract content, audit parameters, and audit rules in each audit, as well as adding or modifying constraint limit words, and is applicable to the audit scenarios of various contract types.

[0047] S40: Input the pre-audit prompt into the large language model to obtain the pre-audit result.

[0048] When the contract to be pre-audited corresponds to a single work order, directly send the obtained pre-audit prompt to the API service of the large language model in the form of a dialogue request. After analysis and processing, the large language model will return a structured pre-audit result. Among them, the structured pre-audit result is a pre-audit result in JSON format. For example, {{"verification result": [{{"item": "1", "audit content": "xxxxx", "audit result": "correct", "audit details": "xxxxxx"}}, {{"item": "2", "audit content": "xxxxxx", "audit result": "incorrect", "audit details": "xxxxxx"}}, ...]}}.

[0049] In this application, after combining multiple sets of marked documents with preset prompt word templates to obtain multiple pre-screening prompt words, it also includes: Multiple pre-screening prompts are asynchronously input into a large language model to obtain multiple pre-screening results.

[0050] When a contract awaiting pre-review corresponds to multiple work orders, for the multiple sets of pre-review prompts obtained, each pre-review prompt is input into the large language model in the form of an asynchronous dialogue request. For each pre-review prompt, the large language model will return the corresponding structured pre-review result after analysis and processing.

[0051] In this application, after inputting the pre-examination prompts into a large language model and obtaining the pre-examination results, the following are also included: The structured pre-review results are converted into a table format and fed back to the reviewers, and the pre-review results are recorded in the review log of the corresponding work order.

[0052] After analyzing and processing the pre-review prompts, the large language model can generate not only pre-review results but also review suggestions. After returning the structured pre-review results and review suggestions, the model can convert them into a table format and present them to the reviewers. This allows the reviewers to view the pre-review status and make timely modifications and adjustments to the contract based on the review suggestions. The model can also automatically record the pre-review results in the review log of the corresponding work order for easy tracking and tracing.

[0053] In one embodiment, see Figure 2 , Figure 2 This is a flowchart illustrating the generation of pre-review prompts in an embodiment of the contract content pre-review method based on a large language model provided in this application. For a contract to be pre-reviewed, the contract content is first extracted into HTML format using a preset file extraction library (such as the Python Mammoth library). Then, the contract content is converted into markup document format using the Beautiful Soup library. Simultaneously, for the work order corresponding to the contract to be pre-reviewed, the review parameters in the work order are extracted, the corresponding review rules are determined, and the review parameters and review rules are converted into markup document format. The contract content, review parameters, and review rules, all of which are in markup document format, are combined together using a preset prompt template to generate pre-review prompts. The pre-review prompts can then be input into the large language model via a dialogue request, where the large language model performs understanding and analysis and returns the pre-review result.

[0054] By extracting contract content, review parameters, and review rules from contracts and work orders, pre-review prompts are generated and input into a large language model. The large language model is then used to assist in rapid pre-review, which effectively improves the work efficiency of reviewers and avoids review errors caused by reviewers' experience and subjective judgment, thereby improving the accuracy of the review.

[0055] This application provides a contract content pre-screening device based on a large language model, see reference. Figure 3 , Figure 3 This is a schematic diagram of an embodiment of the contract content pre-screening device based on a large language model provided in this application. The contract content pre-screening device based on a large language model includes: Content extraction module 10 is used to extract the contract content of the contract to be pre-examined; The work order extraction module 20 is used to extract the audit parameters from at least one work order corresponding to the contract to be pre-approved, and to determine the audit rules based on the audit parameters; The prompt word combination module 30 is used to combine contract content, review parameters, review rules and preset prompt word templates to obtain pre-review prompt words; The pre-screening module 40 is used to input pre-screening prompts into the large language model to obtain pre-screening results.

[0056] The above-mentioned content extraction module 10, work order extraction module 20, prompt word combination module 30 and pre-review module 40 interact to realize the process of contract content pre-review. You can refer to the specific description of steps S10 to S40 above. The repeated parts will not be repeated here.

[0057] See Figure 4 , Figure 4 This is a schematic diagram of an embodiment of the storage medium provided in this application.

[0058] The storage medium 400 stores program data 410, which, when executed by the processor, implements, as follows: Figure 1 The steps of the contract content pre-screening method based on a large language model are described.

[0059] The program data 410 is stored in a storage medium 400 and includes several instructions for causing a network device (which may be a router, personal computer, server, or other network device) or processor to execute all or part of the steps of the methods described in the various embodiments of this application.

[0060] Optionally, the storage medium 400 can be any medium that can store program data, such as a USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), disk, or optical disc.

[0061] See Figure 5 , Figure 5 This is a schematic diagram of the structure of an embodiment of the computer device provided in this application.

[0062] The device 500 includes a processor 520 and a memory 510 connected to each other. The memory 510 stores a computer program. When the processor 520 executes the computer program, it implements the steps of the contract content pre-examination method based on the large language model described above.

[0063] See Figure 6 , Figure 6 This is a schematic diagram of the structure of an embodiment of the computer program product provided in this application.

[0064] The computer program product 600 stores program data 610, which, when executed by the processor, implements, as follows: Figure 1 The steps of the contract content pre-screening method based on a large language model are described.

[0065] Unlike existing technologies, this application discloses a method, apparatus, device, and medium for contract content pre-review based on a large language model. This method extracts the contract content of the contract to be pre-reviewed and combines it with review parameters and rules extracted from the corresponding work order. Pre-review prompts are generated using a preset prompt template and input into a large language model. The large language model is then used to perform contract pre-review, yielding the pre-review result. Reviewers only need to verify the pre-review result, which assists reviewers in quickly approving contracts, improves their work efficiency, avoids review errors caused by reviewers' experience and subjective judgment, and enhances review accuracy.

[0066] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the storage medium embodiments and computer device embodiments are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0067] This application can be used in a wide range of general-purpose or special-purpose computing system environments or configurations. For example: personal computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, network PCs, minicomputers, distributed computing environments including any of the above systems or devices, etc.

[0068] In the several embodiments provided in this application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative; multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed.

[0069] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0070] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0071] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for pre-screening contract content based on a large language model, characterized in that, include: Extract the contract content of the contract to be pre-examined; Extract the audit parameters from at least one work order corresponding to the contract to be pre-audited, and determine the audit rules based on the audit parameters; The contract content, the review parameters, the review rules, and the preset prompt word template are combined to obtain the pre-review prompt words; The pre-screening prompts are input into the large language model to obtain the pre-screening results.

2. The contract content pre-screening method based on a large language model according to claim 1, characterized in that, The extraction of contract content from contracts awaiting pre-review includes: Based on a pre-defined file extraction library, annotations and modifications are removed from the contract to be pre-reviewed, and the modified contract content is extracted.

3. The contract content pre-screening method based on a large language model according to claim 1, characterized in that, The step of extracting audit parameters from at least one work order corresponding to the contract to be pre-approved, and determining audit rules based on the audit parameters, includes: In response to the work order corresponding to the contract to be pre-reviewed, the review parameters in the work order are extracted, and the review rules are determined based on the review parameters; In response to the multiple work orders corresponding to the contract to be pre-reviewed, the review parameters in each work order are extracted asynchronously, and the corresponding review rules are determined based on each review parameter.

4. The contract content pre-screening method based on a large language model according to claim 3, characterized in that, The step of combining the contract content, the review parameters, the review rules, and the preset prompt word template to obtain the pre-review prompt words includes: In response to a work order corresponding to the contract to be pre-reviewed, the contract content, the review parameters and the review rules are converted into a marked document, and the marked document is combined with the preset prompt word template to obtain the pre-review prompt word; In response to the multiple work orders corresponding to the contract to be pre-reviewed, the contract content and the corresponding review parameters and review rules are converted into multiple sets of marked documents, and the multiple sets of marked documents are combined with the preset prompt word template to obtain multiple pre-review prompt words.

5. The contract content pre-screening method based on a large language model according to claim 4, characterized in that, After combining multiple sets of the marked documents with preset prompt word templates to obtain multiple pre-screening prompt words, the method further includes: Multiple pre-screening prompts are asynchronously input into a large language model to obtain multiple pre-screening results.

6. The contract content pre-screening method based on a large language model according to claim 1, characterized in that, After inputting the pre-screening prompts into the large language model and obtaining the pre-screening results, the process further includes: The structured preliminary review results are converted into a table format and fed back to the reviewers, and the preliminary review results are recorded in the review log of the corresponding work order.

7. A contract content pre-screening device based on a large language model, characterized in that, include: The content extraction module is used to extract the contract content of contracts to be pre-examined; The work order extraction module is used to extract the audit parameters from at least one work order corresponding to the contract to be pre-approved, and to determine the audit rules based on the audit parameters; The prompt word combination module is used to combine the contract content, the review parameters, the review rules and the preset prompt word template to obtain the pre-review prompt words; The pre-screening module is used to input the pre-screening prompts into the large language model to obtain the pre-screening results.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the contract content pre-examination method based on a large language model as described in any one of claims 1-6.

9. A storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the contract content pre-examination method based on a large language model as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the contract content pre-examination method based on a large language model as described in claims 1-6.

Images