Contract review method and device based on large model, medium and equipment

Abstract

The application provides a contract review method and device based on a large model, a medium, and equipment. The method comprises the following steps: obtaining a to-be-reviewed contract document; converting the format of the to-be-reviewed contract document into a format that can be recognized by an information extraction large model; inputting the to-be-reviewed contract document after format conversion into the information extraction large model to obtain contract element information in the to-be-reviewed contract document; providing a contract preview function and an element positioning function on a page to display the to-be-reviewed contract document, and positioning the contract element information on the displayed to-be-reviewed contract document; and after a user modifies the contract element information on the page, making corresponding modifications to the to-be-reviewed contract document. The application improves the contract review efficiency, reduces the labor cost, reduces the omissions or errors caused by human factors, and improves the accuracy and reliability of the contract review.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of contract review, in particular to a contract review method and device based on a large model, a medium and equipment. BACKGROUND

[0002] In the current business environment, contracts as the basis of legal and commercial transactions, the accuracy and completeness of its content is crucial to protect the interests of all parties. The traditional contract review method relies on manual reading and analysis, which is not only time-consuming and laborious, but also prone to omissions or errors due to human factors. With the rapid development of big data and artificial intelligence technology, using automated means to extract key information from contracts has become a hot topic in the industry. SUMMARY

[0003] In view of at least one of the above technical problems, the embodiments of the present application provide a contract review method and device based on a large model, a medium and equipment.

[0004] According to a first aspect, the embodiments of the present application provide a contract review method based on a large model, which comprises:

[0005] obtaining a contract document to be reviewed;

[0006] converting the format of the contract document to be reviewed into a format that can be recognized by an information extraction large model;

[0007] inputting the contract document to be reviewed after format conversion into the information extraction large model to obtain contract element information in the contract document to be reviewed; wherein the information extraction large model is used to extract contract element information from the contract document to be reviewed;

[0008] providing a contract preview function and an element positioning function on the page to display the contract document to be reviewed, and positioning the contract element information on the displayed contract document to be reviewed;

[0009] when the user modifies the contract element information on the page, making corresponding modifications to the contract document to be reviewed.

[0010] In one embodiment, the format that can be recognized by the information extraction large model is a docx format, and correspondingly, the conversion of the format of the contract document to be reviewed into a format that can be recognized by the information extraction large model comprises at least one of the following:

[0011] if the format of the contract document to be reviewed is a doc format, the format of the contract document to be reviewed is converted into a docx format;

[0012] If the format of the to-be-reviewed contract document is a pdf format, an optical character recognition technology is used to convert the format of the to-be-reviewed contract document into a docx format.

[0013] In one embodiment, the information extraction large model is further configured to analyze whether the extracted contract element information has semantic risks, and output an analysis result; wherein the semantic risks include statement ambiguity and / or expression ambiguity.

[0014] In one embodiment, the information extraction large model is further configured to, if the to-be-reviewed contract document has a table, analyze whether the total amount and / or the total quantity in the table are correct, and output an analysis result.

[0015] In one embodiment, the method further comprises:

[0016] providing an element extraction feedback function to enable a user to label and feedback on contract element information that is incorrectly extracted by the information extraction large model;

[0017] After receiving the labeled data fed back by the user, the information extraction large model is fine-tuned.

[0018] In one embodiment, the contract element information includes at least one of contract party information, a signing amount, a service scope, and a payment condition.

[0019] In one embodiment, the training process of the information extraction large model comprises:

[0020] a plurality of training samples are formed, and the training samples are divided into a training set, a validation set, and a test set; wherein each training sample is a contract document and contract element information in the contract document;

[0021] a pre-trained model is selected;

[0022] an adjustment strategy is selected, the adjustment strategy including partial adjustment and full adjustment;

[0023] hyperparameters used in the training process are set;

[0024] parameters of the pre-trained model are initialized according to the adjustment strategy;

[0025] the pre-trained model is trained using the training set and the adjustment strategy, the model is periodically evaluated using the validation set during the training process, and adjustments are continued according to the evaluation results to obtain a converged information extraction large model;

[0026] the information extraction large model is performance-evaluated using the test set to obtain performance indicators of the information extraction large model.

[0027] According to a second aspect, the application provides a contract review device based on a large model, comprising:

[0028] a contract acquisition module, configured to acquire a to-be-reviewed contract document;

[0029] a format conversion module, configured to convert a format of the to-be-reviewed contract document into a format recognizable by an information extraction large model;

[0030] a model extraction module, configured to input the to-be-reviewed contract document after format conversion into the information extraction large model to obtain contract element information in the to-be-reviewed contract document; the information extraction large model is configured to extract the contract element information from the to-be-reviewed contract document;

[0031] a man-machine interaction module, configured to provide a contract preview function and an element positioning function on a page to display the to-be-reviewed contract document and position the contract element information on the displayed to-be-reviewed contract document;

[0032] a contract modification module, configured to modify the to-be-reviewed contract document correspondingly after a user modifies the contract element information on the page.

[0033] In an embodiment, the format recognizable by the information extraction large model is a docx format, and correspondingly, the format conversion module is configured to perform at least one of the following:

[0034] if the format of the to-be-reviewed contract document is a doc format, the format of the to-be-reviewed contract document is converted into a docx format;

[0035] if the format of the to-be-reviewed contract document is a pdf format, an optical character recognition technology is used to convert the format of the to-be-reviewed contract document into a docx format.

[0036] In an embodiment, the information extraction large model is further configured to analyze whether the extracted contract element information has semantic risks, and output an analysis result; the semantic risks include statement ambiguity and / or expression ambiguity.

[0037] In an embodiment, the information extraction large model is further configured to analyze whether a total amount and / or a total quantity in a table in the to-be-reviewed contract document are correct, and output an analysis result, if the to-be-reviewed contract document has the table.

[0038] In an embodiment, the device can further comprise:

[0039] The labeling feedback module is configured to provide a contract element extraction feedback function, so that a user labels and feeds back contract element information that is incorrectly extracted by the information extraction large model. After receiving the labeled data fed back by the user, the information extraction large model is adjusted.

[0040] In one embodiment, the contract element information includes at least one of contract party information, a signing amount, a service range, and a payment condition.

[0041] In one embodiment, the apparatus can further include:

[0042] The model training module is configured to perform the following steps: forming a plurality of training samples and dividing the training samples into a training set, a validation set, and a test set; each training sample is a contract document and contract element information in the contract document; selecting a pre-trained model; selecting an adjustment strategy, the adjustment strategy including partial adjustment and full adjustment; setting hyperparameters used in the training process; initializing parameters of the pre-trained model according to the adjustment strategy; training the pre-trained model using the training set and the adjustment strategy, periodically evaluating the model using the validation set during the training process, and continuing to adjust according to the evaluation results to obtain a converged information extraction large model; and evaluating the performance of the information extraction large model using the test set to obtain performance indicators of the information extraction large model.

[0043] According to a third aspect, an embodiment of the present application provides a computer readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method of the first aspect.

[0044] According to a fourth aspect, an embodiment of the present application provides a computing device including a memory and a processor, the memory having executable code stored therein, and the processor executing the executable code to implement the method of the first aspect.

[0045] The contract review method and device based on a large model provided by the embodiment of the present application, the format of the to-be-reviewed contract document is converted into a format that can be recognized by an information extraction large model, then the to-be-reviewed contract document after format conversion is input into the information extraction large model, contract element information in the to-be-reviewed contract document is obtained, the to-be-reviewed contract document is displayed through a contract preview function and an element positioning function provided on a page, and the contract element information is positioned on the displayed to-be-reviewed contract document, so that the user can review the contract element information, and if the user finds that the contract element information has errors, the errors can be modified on the page, and after modification, the to-be-reviewed contract is also completed corresponding modification. It can be seen that the embodiment of the present application realizes automatic extraction of contract element information based on a large model, compared with the traditional manual review method, the embodiment of the present application can quickly identify and extract key information in the contract through automatic means, greatly shortens the time period of contract review, that is, improves the contract review efficiency, reduces the labor cost, reduces the omissions or errors caused by human factors, and improves the accuracy and reliability of contract review. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 FIG. 1 is a flowchart of a contract review method based on a large model in an embodiment of the present application;

[0047] Figure 2 FIG. 2 is a structural block diagram of a contract review device based on a large model in an embodiment of the present application. DETAILED DESCRIPTION

[0048] In a first aspect, the embodiment of the present application provides a contract review method based on a large model, referring to FIG. 1, Figure 1 The method comprises the following steps S110-S150:

[0049] S110, obtaining a to-be-reviewed contract document;

[0050] The method provided by the embodiment of the present application can be executed by a reasoning server.

[0051] S120, converting the format of the to-be-reviewed contract document into a format that can be recognized by an information extraction large model;

[0052] In one embodiment, the format that can be recognized by the information extraction large model is a docx format, and correspondingly, the step S120 of converting the format of the to-be-reviewed contract document into a format that can be recognized by the information extraction large model can comprise at least one of the following:

[0053] (1) if the format of the to-be-reviewed contract document is a doc format, the format of the to-be-reviewed contract document is converted into a docx format;

[0054] (2) If the format of the to-be-reviewed contract document is a pdf format, an optical character recognition technology is used to convert the format of the to-be-reviewed contract document into a docx format.

[0055] It can be seen that if the format of the to-be-reviewed contract document is a doc format, the contract format conversion program can be used to convert it into a docx format. If the format of the to-be-reviewed contract document is a pdf format, an optical character recognition technology is used to convert it into a docx format. This facilitates accurate identification by the information extraction large model.

[0056] S130, inputting the format-converted to-be-reviewed contract document into the information extraction large model to obtain contract element information in the to-be-reviewed contract document; wherein the information extraction large model is used to extract contract element information from the to-be-reviewed contract document;

[0057] It can be seen that the function of the information extraction large model is to extract contract element information from the to-be-reviewed contract document.

[0058] The contract element information can include at least one of contract party information, a signing amount, a service range, and a payment condition.

[0059] In one embodiment, the information extraction large model can also be used to analyze whether the extracted contract element information has semantic risks, and output an analysis result; wherein the semantic risks include statement ambiguity and / or expression ambiguity.

[0060] It can be seen that in addition to outputting contract element information, the information extraction large model also analyzes whether the contract element information has semantic risks, for example, judges whether the contract requirement information has semantic ambiguity and expression ambiguity risks, and outputs the analysis result after obtaining the analysis result. In this way, the user can know whether the contract element information has semantic risks, and the semantic risks can be reviewed.

[0061] In one embodiment, the information extraction large model can also be used to analyze whether the total amount and / or the total quantity in the table are correct if the to-be-reviewed contract document has a table, and output an analysis result.

[0062] It can be seen that in addition to outputting contract element information, the information extraction large model can also check whether there is a table in the to-be-reviewed contract document. If there is a table, it analyzes whether the total amount, total quantity, and other data in the table are correct, and outputs an analysis result. In this way, it can be known whether the total value in the table has an error, and the total value can be reviewed.

[0063] S140, providing a contract preview function and an element positioning function on the page to display the to-be-reviewed contract document and position the contract element information on the displayed to-be-reviewed contract document;

[0064] S150, after the user modifies the contract element information on the page, modifying the to-be-reviewed contract document accordingly.

[0065] It can be seen that the contract preview function can display the to-be-reviewed contract document, so that the user can preview the to-be-reviewed contract document. The element positioning function can position the extracted contract element information on the to-be-reviewed contract document, for example, highlight the contract element information on the to-be-reviewed contract document by highlighting color, so that the user can clearly see the contract element information on the to-be-reviewed contract document, and the user can review the contract element information. If errors are found during review, the contract element information can be modified directly on the to-be-reviewed contract document. After modification, the to-be-reviewed contract document is modified accordingly.

[0066] It can be seen that the system has the functions of previewing the contract and modifying the original text.

[0067] In one embodiment, the method can further include:

[0068] providing an element extraction feedback function to enable the user to label and feed back the contract element information extracted by the information extraction large model incorrectly; after receiving the labeled data fed back by the user, optimizing the information extraction large model.

[0069] It can be seen that through the element extraction feedback function, the user can label and feed back the contract element information extracted by the information extraction large model incorrectly. For example, the extracted contract element information is not the real key information, that is, the information extraction is unreasonable. After receiving the feedback, the information extraction large model is optimized, thereby improving the extraction accuracy of the information extraction large model.

[0070] In actual scenarios, the system also has a contract management function, and the user can perform operations such as uploading, deleting, and downloading a contract through the contract management function.

[0071] In one embodiment, the training process of the information extraction large model includes the following 7 steps:

[0072] 1. Form a plurality of training samples, and divide the training samples into a training set, a validation set, and a test set; wherein each training sample is a contract document and contract element information in the contract document;

[0073] 2. Select a pre-training model;

[0074] 3. Select an adjustment strategy, which includes partial adjustment and full adjustment;

[0075] 4. Set hyperparameters used in the training process;

[0076] 5. Initialize the parameters of the pre-trained model according to the adjustment strategy;

[0077] 6. Train the pre-trained model using the training set and the adjustment strategy. During the training process, use the validation set to periodically evaluate the model, and continue to adjust according to the evaluation results to obtain a converged information extraction large model;

[0078] 7. Evaluate the performance of the information extraction large model using the test set to obtain the performance indicators of the information extraction large model.

[0079] As can be seen, the training process of the information extraction large model can include the following steps:

[0080] (1) First, prepare a contract information set including multiple contract documents, and then determine the contract element information of each contract according to the contract text, such as the information of the parties (lessee, authorized party, contractor, and principal), amount, tax, account, etc. Label the contract element information of each contract on the contract to form a training sample, and thus obtain multiple training samples, and then divide the multiple training samples into a training set, a validation set, and a test set.

[0081] (2) Then, select a suitable pre-trained model according to the characteristics of the contract and the extraction target.

[0082] (3) Next, select an appropriate fine-tuning strategy according to the task requirements and available resources. Consider whether to perform full adjustment or partial adjustment, or even the level and range of adjustment.

[0083] (4) Set the hyperparameters used in the training process, such as learning rate, batch size, and training rounds. The selection of these hyperparameters has an important impact on the performance and convergence speed of fine-tuning.

[0084] (5) Initialize the parameters of the pre-trained model. For full adjustment, all model parameters are randomly initialized; for partial adjustment, only the parameters of the top layer or a small number of layers are randomly initialized.

[0085] (6) Train the pre-trained model using the prepared training set and adjustment strategy. During the training process, use the validation set to periodically evaluate the model, and adjust the hyperparameters according to the evaluation results.

[0086] (7) After training is completed, evaluate the final model using the test set to obtain the final performance indicators.

[0087] After the training is completed, the obtained information extraction large model is deployed on an inference server for performing the extraction of contract element information in a contract document.

[0088] In recent years, deep learning models, especially large-scale pre-training models such as BERT and GPT series, have made remarkable achievements in natural language processing. These models can understand and generate complex text content, providing new possibilities for the automated extraction of contract information. By training these models, they can learn to identify key elements in contract text, such as contract party information, signing date, service scope, payment terms, and breach of contract liability, thereby significantly improving the efficiency and accuracy of contract review. However, contract texts often have high levels of professionalism, complexity, and diversity, containing a large number of legal terms and specific industry expressions, which pose higher requirements for the generalization ability and understanding ability of the model. In addition, the key information in the contract is often hidden in long paragraphs of text, and the structure and format of different contracts differ greatly. How to effectively and accurately extract key information from these unstructured data has become a major challenge for current large model-based contract key information extraction methods. The information extraction large model obtained by the training process described above can efficiently and accurately extract key information from various contract texts, which is of great significance for improving contract review efficiency and reducing legal risks.

[0089] The embodiment of the present application proposes a large model-based contract key information extraction method, which takes a language large model as the core and integrates advanced artificial intelligence technology, aiming to provide users with more intelligent, convenient, and personalized information acquisition and interaction experience, so that information extraction does not have to rely on manual extraction, greatly improving the review efficiency. Due to the integration of OCR text recognition, language large model, and other technologies, the large model is used for deep understanding and analysis of contracts, realizing the extraction, sorting, and review analysis of contract key information. The embodiment of the present application supports batch processing of contracts, with an average of 20 seconds per contract, compared with 20 contracts per week per person in the previous manual method, and the information processing speed has been greatly improved. Therefore, the method provided by the embodiment of the present application has the characteristics of convenience, speed, labor saving, and efficiency, and can batch process contract documents to quickly extract contract element information.

[0090] The embodiment of the present application has the following beneficial effects:

[0091] (1) The large model-based contract key information extraction method proposed in the embodiment of the present application, compared with the traditional manual review method, can quickly identify and extract key information in the contract through automated means, greatly shortening the time period of contract review. This not only reduces labor costs, but also enables enterprises to respond more quickly to market changes and seize business opportunities.

[0092] (2) With the powerful ability of deep learning models, especially large-scale pre-training models, the embodiments of the present application can accurately understand the complexity and diversity of contract texts, effectively identify and extract key information hidden in long paragraphs. This reduces the omissions or errors caused by human factors and improves the accuracy and reliability of contract review.

[0093] (3) By accurately extracting key information from contracts, the embodiments of the present application can help enterprises discover potential legal risk points, such as unclear payment terms and vague breach of contract responsibilities. This helps enterprises take measures in advance to prevent legal risks and protect their own interests.

[0094] (4) The method provided by the embodiments of the present application has high flexibility and scalability, and can be customized and developed according to the specific needs of different enterprises. For example, for contract texts in a specific industry, a special model can be trained to extract key information related to the industry, thereby better meeting the actual needs of enterprises.

[0095] (5) The embodiments of the present application mark an important step in the intelligent and automated direction of contract review. With the continuous progress of technology and the expansion of application scenarios, the method provided by the embodiments of the present application is expected to become one of the mainstream means in the field of contract review in the future, providing strong support for the compliance operation and risk management of enterprises.

[0096] In summary, by using the contract key information extraction method based on large models of the present application, the key information and semantic risks of batch contracts are analyzed and displayed, thereby greatly liberating professional human resources, reducing the professional requirements of human resources, greatly shortening the time of contract key information extraction and semantic risk analysis, and being able to timely and accurately provide guidance for enterprises to determine the research and development direction. Moreover, it has significant advantages in improving the efficiency, accuracy, and legal risk prevention and control capabilities of review, while supporting customization needs and promoting the intelligent development of the contract review field.

[0097] In a second aspect, the embodiments of the present application provide a contract review device based on a large model, which is shown in Figure 2 The device 100 comprises:

[0098] A contract acquisition module 110 is configured to acquire a contract document to be reviewed.

[0099] A format conversion module 120 is configured to convert the format of the contract document to be reviewed into a format that can be recognized by an information extraction large model.

[0100] The model extraction module 130 is configured to input the format-converted to-be-reviewed contract document into the information extraction large model to obtain contract element information in the to-be-reviewed contract document.

[0101] The human-computer interaction module 140 is configured to provide a contract preview function and an element positioning function on a page to display the to-be-reviewed contract document and position the contract element information on the displayed to-be-reviewed contract document.

[0102] The contract modification module 150 is configured to modify the to-be-reviewed contract document correspondingly when the user modifies the contract element information on the page.

[0103] In an embodiment, the information extraction large model can recognize a format of docx, and correspondingly, the format conversion module 120 is configured to perform at least one of the following:

[0104] If the format of the to-be-reviewed contract document is doc, the format of the to-be-reviewed contract document is converted into the format of docx.

[0105] If the format of the to-be-reviewed contract document is pdf, the format of the to-be-reviewed contract document is converted into the format of docx by using an optical character recognition technology.

[0106] In an embodiment, the information extraction large model is further configured to analyze whether the extracted contract element information has semantic risks and output an analysis result, wherein the semantic risks include statement ambiguity and / or expression ambiguity.

[0107] In an embodiment, the information extraction large model is further configured to analyze whether the total amount and / or the total quantity in the table are correct if the to-be-reviewed contract document has a table and output an analysis result.

[0108] In an embodiment, the apparatus can further include:

[0109] The labeling feedback module is configured to provide an element extraction feedback function to enable the user to label and feed back contract element information that is incorrectly extracted by the information extraction large model, and to perform tuning processing on the information extraction large model after receiving the labeled data fed back by the user.

[0110] In an embodiment, the contract element information includes at least one of contract party information, a signing amount, a service range, and payment conditions.

[0111] In an embodiment, the apparatus can further include:

[0112] The model training module is configured to perform the following steps: forming a plurality of training samples, and dividing the training samples into a training set, a validation set, and a test set; wherein each training sample is a contract document and contract element information in the contract document; selecting a pre-trained model; selecting an adjustment strategy, the adjustment strategy including partial adjustment and full adjustment; setting hyperparameters used in the training process; initializing parameters of the pre-trained model according to the adjustment strategy; training the pre-trained model using the training set and the adjustment strategy, periodically evaluating the model using the validation set during the training process, and continuing to adjust according to the evaluation results to obtain a converged information extraction large model; and evaluating the performance of the information extraction large model using the test set to obtain performance indicators of the information extraction large model.

[0113] It can be understood that the explanations, specific embodiments, beneficial effects, examples, and the like related to the device provided by the embodiments of the present application can refer to the corresponding parts in the method provided by the first aspect, which will not be repeated here.

[0114] In a third aspect, a computer readable medium is provided, and the computer readable medium stores computer instructions. When the computer instructions are executed by a processor, the processor executes the method provided by the first aspect.

[0115] Specifically, a system or device equipped with a storage medium can be provided, and the storage medium stores software program codes for implementing the functions of any of the above embodiments, and the computer (or CPU or MPU) of the system or device reads and executes the program codes stored in the storage medium.

[0116] In this case, the program codes read from the storage medium can implement the functions of any of the above embodiments, and thus the program codes and the storage medium storing the program codes constitute a part of the present application.

[0117] Embodiments of the storage medium for providing the program codes include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, the program codes can be downloaded from a server computer via a communication network.

[0118] In addition, it should be clear that not only the program codes read by the computer can be executed, but also part or all of the actual operations can be completed by the operating system and the like operating on the computer based on the instructions of the program codes, thereby implementing the functions of any of the above embodiments.

[0119] Further, it is understood that the programs while being read by the storage media are written into the memory provided in the extension board inserted into the computer or the memory provided in the extension module connected to the computer, and then the CPU or the like mounted on the extension board or the extension module is caused to perform part or all of the actual operation based on the instructions of the program codes, thereby realizing the functions of any of the above-described embodiments.

[0120] It is to be understood that the explanation of the content, the detailed description, the beneficial effects, the examples and the like provided in the computer readable medium of the embodiments of the present application can refer to the corresponding parts in the method provided in the first aspect, and will not be repeated here.

[0121] In a fourth aspect, an embodiment of the present specification provides a computing device, comprising a memory and a processor, wherein the memory stores executable codes, and the processor executes the executable codes to implement the method in any of the embodiments of the specification.

[0122] It is to be understood that the explanation of the content, the detailed description, the beneficial effects, the examples and the like provided in the computer readable medium of the embodiments of the present application can refer to the corresponding parts in the method provided in the first aspect, and will not be repeated here.

[0123] Each of the embodiments in the present specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each of the embodiments mainly describes the difference from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the part of the description of the method embodiments.

[0124] Those skilled in the art should realize that, in one or more of the examples described above, the functions described in the present application can be implemented by hardware, software, a plug-in or any combination thereof. When implemented by software, these functions can be stored in a computer readable medium or transmitted as one or more instructions or codes on a computer readable medium.

[0125] The above detailed description further describes the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement and the like made on the basis of the technical solutions of the present application shall be included in the protection scope of the present application.

Claims

1. A contract review method based on a large model, characterized in that: include: Obtain contract documents for review; Converting the format of the contract document to be reviewed into a format that can be recognized by the information extraction model; Inputting the format-converted contract document to be reviewed into the information extraction model to obtain contract element information in the contract document to be reviewed; wherein the information extraction model is used to extract the contract element information from the contract document to be reviewed; Providing a contract preview function and an element positioning function on the page to display the contract document to be reviewed, and positioning the contract element information on the displayed contract document to be reviewed; When the user modifies the contract element information on the page, the contract document to be reviewed is modified accordingly; The information extraction model is also used to analyze whether the extracted contract element information has semantic risks and output the analysis results; wherein the semantic risks include sentence ambiguity and / or vague expression.

2. The method according to claim 1, characterized in that The format that the information extraction model can recognize is the docx format. Correspondingly, converting the format of the contract document to be reviewed into a format that the information extraction model can recognize includes at least one of the following: If the format of the contract document to be reviewed is doc format, convert the format of the contract document to be reviewed into docx format; If the format of the contract document to be reviewed is pdf format, optical character recognition technology is used to convert the format of the contract document to be reviewed into docx format.

3. The method according to claim 1, characterized in that The information extraction model is also used to: if there is a table in the contract document to be reviewed, analyze whether the total amount and / or total quantity in the table is correct, and output the analysis result.

4. The method according to claim 1, wherein Also includes: Providing an element extraction feedback function to enable users to mark and provide feedback on contract element information that is incorrectly extracted by the information extraction model; After receiving the annotated data fed back by the user, the information extraction model is tuned.

5. The method according to claim 1, characterized in that The contract element information includes at least one of the information of the contracting parties, the contract amount, the scope of services and the payment terms.

6. The method according to claim 1, characterized in that The training process of the information extraction model includes: Forming a plurality of training samples and dividing the training samples into a training set, a validation set, and a test set; wherein each training sample is a contract document and contract element information in the contract document; Select a pre-trained model; Select an adjustment strategy, which includes partial adjustment and full adjustment; Set the hyperparameters used in the training process; Initializing the parameters of the pre-trained model according to the adjustment strategy; The pre-trained model is trained using the training set and the adjustment strategy, the model is regularly evaluated using the validation set during the training process, and adjustments are continued based on the evaluation results to obtain a converged information extraction model; The test set is used to perform a performance evaluation on the information extraction model to obtain performance indicators of the information extraction model.

7. A contract review device based on a large model, characterized in that: include: Contract acquisition module, used to obtain contract documents to be reviewed; A format conversion module, used to convert the format of the contract document to be reviewed into a format that can be recognized by the information extraction model; a model extraction module, configured to input the format-converted contract document to be reviewed into the information extraction model to obtain contract element information in the contract document to be reviewed; wherein the information extraction model is configured to extract contract element information from the contract document to be reviewed; A human-computer interaction module is used to provide a contract preview function and an element positioning function on the page to display the contract document to be reviewed, and to locate the position of the contract element information on the displayed contract document to be reviewed; A contract modification module, configured to modify the contract document to be reviewed accordingly after the user modifies the contract element information on the page; The information extraction model is also used to analyze whether the extracted contract element information has semantic risks and output the analysis results; wherein the semantic risks include sentence ambiguity and / or vague expression.

8. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed in a computer, the computer is caused to execute the method according to any one of claims 1 to 6.

9. A computing device, characterized in that The method comprises a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the method according to any one of claims 1 to 6 is implemented.

Images