Contract review method and device, related equipment and computer program product

Abstract

The invention discloses a contract review method and device, related equipment and a computer program product, and the method comprises the steps: determining a target type of a contract and each atomic term contained in the contract for the contract to be reviewed, and determining a review theme corresponding to each atomic term in the contract to be reviewed. And aiming at each atomic clause, according to the corresponding review theme, constructing a review instruction for indicating the large model to review the risk result aiming at the atomic clause according to the corresponding review theme, calling the large model, and obtaining the risk review result of each atomic clause output by the large model. By means of the capacity of the large model, automatic review of the contract is achieved, the review efficiency is improved, and the accuracy of the risk review result of the contract is improved by means of the capacity of the large model.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a contract review method, apparatus, related equipment and computer program product. Background Art

[0002] The contract review task is to review whether there are risks and problems in the contract. The existing contract review solutions are generally based on manual review, that is, they rely on manual review of contract terms one by one, combined with the professional knowledge of multiple departments such as legal, financial, and business to make judgments to identify whether there are risks and problems in the contract.

[0003] Relying on manual contract review places high demands on contract reviewers, who are required to have comprehensive knowledge of law, compliance, finance, business, etc. This method results in low contract review efficiency. Summary of the invention

[0004] In view of the above problems, this application is proposed to provide a contract review method, apparatus, related equipment and computer program product to solve the problems existing in the manual contract review method. The specific solution is as follows:

[0005] First, a contract review method is provided, comprising:

[0006] For the contract to be reviewed, determine the target type of the contract and each atomic clause contained in the contract;

[0007] Query the mapping relationship between the atomic clauses and the review subjects under the configured target type, and determine the review subject corresponding to each atomic clause in the contract;

[0008] For each of the atomic clauses, a review instruction prompt is constructed according to the corresponding review subject, and the prompt is used to instruct the big model to review the risk results of the input atomic clause according to the corresponding review subject;

[0009] Through the prompt corresponding to each of the atomic clauses, the configured big model is called to obtain the risk review results of each of the atomic clauses output by the big model.

[0010] In a possible design, in another implementation of the first aspect of the embodiment of the present application, for each of the atomic clauses, a process of constructing a review instruction prompt according to a corresponding review subject includes:

[0011] Obtain the configured review instruction prompt format template, which includes a task description, a review subject slot, and an input data slot. The task description is used to instruct the big model to review the risk results of the atomic clauses in the input data slot according to the review subject in the review subject slot;

[0012] For each of the atomic clauses, fill its corresponding review subject into the review subject slot, fill the atomic clause into the input data slot, and obtain the review instruction prompt corresponding to the atomic clause.

[0013] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the template further includes: a review knowledge slot, wherein the task description is specifically used to instruct the large model to review the risk results of the atomic clauses in the input data slot according to the review subject in the review subject slot and based on the review knowledge in the review knowledge slot;

[0014] The process of constructing the review instruction prompt also includes:

[0015] For each of the atomic clauses, obtaining the configured review knowledge corresponding to the review subject to which the atomic clause belongs;

[0016] Fill the acquired review knowledge into the review knowledge slot to obtain the review instruction prompt corresponding to the atomic clause.

[0017] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the template further includes: an output requirement slot, and the task description is further used to instruct the large model to output the review result according to the output form specified in the output requirement slot;

[0018] The process of constructing the review instruction prompt also includes:

[0019] For each of the atomic clauses, obtaining a configured output form corresponding to the review subject to which the atomic clause belongs, the output form including a risk type and a risk level;

[0020] Fill the obtained output form into the output requirement slot to obtain the review instruction prompt corresponding to the atomic clause.

[0021] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the template further includes: a sample slot, wherein the task description is specifically used to instruct the large model to refer to the sample given in the sample slot and review the risk results of the atomic clauses in the input data slot according to the review subject in the review subject slot;

[0022] The process of constructing the review instruction prompt also includes:

[0023] For each of the atomic clauses, obtain a reference sample configured to correspond to the examination subject to which the atomic clause belongs;

[0024] Fill the acquired reference sample into the sample slot to obtain the review instruction prompt corresponding to the atomic clause.

[0025] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the configured big model includes two or more different types of big models, and the process of calling the configured big model through the prompt corresponding to each of the atomic clauses to obtain the risk review result of each of the atomic clauses output by the big model includes:

[0026] For each of the atomic clauses:

[0027] Through the prompt corresponding to the atomic clause, each type of large model is called respectively to obtain the risk review result of the atomic clause output by each type of large model;

[0028] According to the weights of various types of large models under the review topics corresponding to the atomic clauses, the risk review results of the atomic clauses output by all types of large models are voted on to obtain the final risk review results of the atomic clauses.

[0029] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the method further includes:

[0030] Obtain the user's adoption status of the final risk review results of each of the atomic clauses, and adjust the weight values ​​of various types of large models under the review topics corresponding to the atomic clauses according to the adoption status.

[0031] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the method further includes:

[0032] Calling the big model to instruct the big model to perform consistency review on the full text of the contract to be reviewed, and obtaining a consistency review result;

[0033] The consistency review includes at least one of the following:

[0034] Consistency review of enterprise name, amount and contract signing parties.

[0035] In a second aspect, a contract review device is provided, comprising:

[0036] A contract type and clause determination unit, used to determine the target type of the contract and each atomic clause contained in the contract for review;

[0037] A review subject determination unit, used for querying the mapping relationship between the atomic clauses and the review subjects under the configured target type, and determining the review subject corresponding to each of the atomic clauses in the contract;

[0038] A review instruction construction unit, used to construct a review instruction prompt for each of the atomic clauses according to the corresponding review subject, wherein the prompt is used to instruct the large model to review the risk results for the input atomic clauses according to the corresponding review subject;

[0039] The clause review unit is used to call the configured big model through the prompt corresponding to each of the atomic clauses to obtain the risk review results of each of the atomic clauses output by the big model.

[0040] In a third aspect, an electronic device is provided, comprising: a memory and a processor;

[0041] The memory is used to store programs;

[0042] The processor is used to execute the program to implement the various steps of the contract review method described in any one of the first aspects of the present application.

[0043] In a fourth aspect, a readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the various steps of the contract review method described in any one of the first aspects of the present application are implemented.

[0044] In a fifth aspect, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the various steps of the contract review method described in any one of the aforementioned first aspects of the present application.

[0045] Through the above technical solution, the present application provides an automated contract review method. For the contract to be reviewed, the target type of the contract and the atomic clauses contained in the contract are first determined, and then the mapping relationship between the atomic clauses and the review subject under the pre-configured target type is queried, so as to determine the review subject corresponding to each atomic clause in the contract to be reviewed. For each atomic clause, a review instruction prompt is constructed according to the corresponding review subject, which is used to instruct the big model to review the risk results of the atomic clause according to the corresponding review subject. On this basis, the big model is called through the prompt to obtain the risk review results of each atomic clause output by the big model. With the help of the capabilities of the big model, the present application realizes the automated review of contracts and improves the review efficiency.

[0046] Furthermore, this application is pre-equipped with a combination of contract types and atomic clauses, and a mapping relationship between the review topics. On this basis, for the contract to be reviewed, the corresponding review topic can be determined according to its target type and each atomic clause in it, and the mapping and matching of each atomic clause with the review topic can be achieved. Based on this, a review instruction prompt can be constructed, which can better stimulate the capabilities of the large model and improve the accuracy of the risk review results of the atomic clauses. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0048] Figure 1 A schematic diagram of an implementation system architecture of the contract review method provided in an embodiment of the present application;

[0049] Figure 2 A schematic diagram of a terminal structure provided in an embodiment of the present application;

[0050] Figure 3 A schematic diagram of a server structure provided in an embodiment of the present application;

[0051] Figure 4 A schematic diagram of a contract review method flow provided for an embodiment of the present application;

[0052] Figure 5 A schematic diagram of the module composition of a contract review system provided in an embodiment of the present application;

[0053] Figure 6 A schematic diagram of a layered architecture of a contract review system provided in an embodiment of the present application;

[0054] Figure 7 A schematic diagram of the structure of a contract review device provided in an embodiment of the present application;

[0055] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] Before introducing this application solution, the relevant concepts involved in this article are first explained:

[0057] prompt: Instructions. When interacting with AI (such as an artificial intelligence model), you need to send instructions to AI. It can be a text description, such as "Please recommend me a pop song" when you interact with AI, or it can be a parameter description in a certain format. For example, if you want AI to draw in a certain format, you need to describe the relevant drawing parameters.

[0058] Large models: In the field of artificial intelligence, large models usually refer to large-scale pre-trained models. Such models are called "large" because they can be pre-trained on a large amount of data and can be transferred to a variety of downstream tasks. Their full English name is Large Pre-Trained Models or Large-Scale Pre-Training Models. Large models are characterized by their large scale and contain billions or even more parameters, which help them learn complex patterns in the data. The emerging capabilities of large models include but are not limited to: contextual learning, instruction following, code generation, step-by-step reasoning capabilities, etc. Large models can include large language models (LLM) and multimodal large models. Large language models are mainly used to process text modal data. Multimodal large models further integrate multimodal capabilities on the basis of large language models and can process information in multiple modalities, such as images, text, audio, etc.

[0059] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0060] This application provides a contract review method that can be applied to Figure 1 The system architecture shown in FIG. 1 may include a terminal 100 and a server 200. The server 200 may include one or more servers ( Figure 1 A server is included as an example for explanation).

[0061] The terminal 100 or the server 200 can be used alone to execute the contract review method provided in the embodiment of the present application. In addition, the terminal 100 and the server 200 can also be used in conjunction to execute the contract review method provided in the embodiment of the present application.

[0062] Next describe Figure 1 The product form of the mid-terminal 100;

[0063] The terminal 100 in the embodiment of the present application can be a mobile phone, a tablet computer, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc., and the embodiment of the present application does not impose any limitation on this.

[0064] Figure 2 An optional hardware structure diagram of the terminal 100 is shown.

[0065] refer to Figure 2 As shown, the terminal 100 may include a radio frequency unit 110, a memory 120, an input unit 130, a display unit 140, a camera 150 (optional), an audio circuit 160, a speaker 161, a microphone 162, an earphone jack 163 (optional), a processor 170, an external interface 180, a power supply 190 and other components. Those skilled in the art will appreciate that Figure 2 These are merely examples of terminals or multi-function devices and do not constitute limitations on the terminals or multi-function devices, which may include more or fewer components than those shown in the figures, or combinations of certain components, or different components.

[0066] The input unit 130 can be used to receive input digital or character information, and generate key signal input related to the user settings and function control of the portable multifunctional device. Specifically, the input unit 130 may include a touch screen 131 and / or other input devices 132. The touch screen 131 can collect the user's touch operations on or near it (such as the user's operation on or near the touch screen using any suitable object such as fingers, joints, stylus, etc.), and drive the corresponding connection device according to a pre-set program. The touch screen can detect the user's touch action on the touch screen, convert the touch action into a touch signal and send it to the processor 170, and can receive and execute the command sent by the processor 170; the touch signal at least includes the touch point coordinate information. The touch screen 131 can provide an input interface and an output interface between the terminal 100 and the user. In addition, the touch screen can be implemented using multiple types such as resistive, capacitive, infrared and surface acoustic wave. In addition to the touch screen 131, the input unit 130 can also include other input devices. Specifically, other input devices 132 may include, but are not limited to, one or more of a physical keyboard, function keys (such as a volume control key, a switch key, etc.), a trackball, a mouse, a joystick, and the like.

[0067] Among them, other input devices 132 can receive input data and so on.

[0068] The display unit 140 may be used to display information input by the user or provided to the user, various menus of the terminal 100, interactive interfaces, file display, and / or playback of any multimedia file. In the embodiment of the present application, the display unit 140 may be used to display various interactive interfaces, processing results, etc. in the contract review method.

[0069] The memory 120 can be used to store instructions and data. The memory 120 can mainly include an instruction storage area and a data storage area. The data storage area can store various data, such as multimedia files, texts, etc.; the instruction storage area can store software units such as operating systems, applications, instructions required for at least one function, or their subsets and extensions. It can also include a non-volatile random access memory; provide the processor 170 with hardware, software and data resources including management of computing and processing equipment, and support control software and applications. It is also used for the storage of multimedia files, and the storage of running programs and applications.

[0070] The processor 170 is the control center of the terminal 100. It uses various interfaces and lines to connect various parts of the entire terminal 100. By running or executing instructions stored in the memory 120 and calling data stored in the memory 120, it executes various functions of the terminal 100 and processes data, thereby controlling the terminal device as a whole. Optionally, the processor 170 may include one or more processing units; preferably, the processor 170 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface and application program, and the modem processor mainly processes wireless communication. It is understandable that the above-mentioned modem processor may not be integrated into the processor 170. In some embodiments, the processor and the memory may be implemented on a single chip, and in some embodiments, they may also be implemented separately on separate chips. The processor 170 may also be used to generate corresponding operation control signals, send them to corresponding components of the computing and processing device, read and process data in the software, especially read and process data and programs in the memory 120, so that each functional module therein performs corresponding functions, thereby controlling the corresponding components to act according to the requirements of the instructions.

[0071] Among them, the memory 120 can be used to store software codes related to the contract review method, the processor 170 can execute the steps of the contract review method, and can also schedule other units (such as the above-mentioned input unit 130 and display unit 140) to achieve corresponding functions.

[0072] The radio frequency unit 110 (optional) can be used for receiving and sending information or receiving and sending signals during a call, for example, after receiving the downlink information of the base station, it is sent to the processor 170 for processing; in addition, the designed uplink data is sent to the base station. Generally, the RF circuit includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (Low Noise Amplifier, LNA), a duplexer, etc. In addition, the radio frequency unit 110 can also communicate with network devices and other devices through wireless communication. The wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile communication (Global System of Mobile communication, GSM), General Packet Radio Service (General Packet Radio Service, GPRS), Code Division Multiple Access (Code Division Multiple Access, CDMA), Wideband Code Division Multiple Access (Wideband Code Division Multip le Access, WCDMA), Long Term Evolution (Long Term Evolution, LTE), email, Short Messaging Service (SMS), etc.

[0073] In the embodiment of the present application, the radio frequency unit 110 can send data to the server 200 and receive the processing result sent by the server 200. Exemplarily, the radio frequency unit 110 sends the received contract to be reviewed to the server 200, and the server 200 reviews the contract, obtains the risk review result of the contract, and returns the risk review result to the terminal 100 for output.

[0074] It should be understood that the radio frequency unit 110 is optional and can be replaced by other communication interfaces, such as a network port.

[0075] The terminal 100 also includes a power supply 190 (such as a battery) for supplying power to various components. Preferably, the power supply can be logically connected to the processor 170 through a power management system, so that the power management system can manage functions such as charging, discharging, and power consumption.

[0076] The terminal 100 further includes an external interface 180 , which may be a standard Micro USB interface or a multi-pin connector, and may be used to connect the terminal 100 to communicate with other devices, or to connect a charger to charge the terminal 100 .

[0077] Although not shown, the terminal 100 may also include a flashlight, a wireless fidelity (WiFi) module, a Bluetooth module, sensors with different functions, etc., which are not described in detail here. Some or all of the methods described below may be applied in the following embodiments. Figure 2 In the terminal 100 shown.

[0078] Next describe Figure 1 The product form of the server 200;

[0079] Figure 3 A structural diagram of a server 200 is provided, such as Figure 3 As shown, the server 200 includes a bus 201, a processor 202, a communication interface 203, and a memory 204. The processor 202, the memory 204, and the communication interface 203 communicate with each other via the bus 201.

[0080] The bus 201 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0081] The processor 202 may be any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0082] The memory 204 may include a volatile memory, such as a random access memory (RAM). The memory 204 may also include a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0083] Among them, the memory 204 can be used to store software codes related to the contract review method, the processor 202 can execute the steps of the contract review method, and can also schedule other units to implement corresponding functions.

[0084] It should be understood that the above-mentioned terminal 100 and server 200 can be centralized or distributed devices, and the processors in the above-mentioned terminal 100 and server 200 (such as processor 170 and processor 202) can be hardware circuits (such as application specific integrated circuit (ASIC), field-programmable gate array (FPGA), general-purpose processor, digital signal processor (DSP), microprocessor or microcontroller, etc.), or a combination of these hardware circuits. For example, the processor can be a hardware system with an instruction execution function, such as a CPU, DSP, etc., or a hardware system without an instruction execution function, such as an ASIC, FPGA, etc., or a combination of the above-mentioned hardware systems without an instruction execution function and hardware systems with an instruction execution function.

[0085] The current method of manually reviewing contracts places high demands on contract reviewers, who are required to have comprehensive knowledge of law, compliance, finance, business, etc. This method results in low efficiency in contract review.

[0086] For this purpose, some possible alternatives are:

[0087] Rule-based contract review. This solution checks specific contract content through a series of pre-defined rules, logic and templates. The core of this solution is to build an actionable rule set for the review task and discover potential problems and risks by executing regular expressions, logical formulas, etc.

[0088] Contract review based on machine learning. This solution adopts a small machine learning model solution to achieve automated review of contract texts through contract data preparation, contract review data annotation, model design and training, evaluation and optimization.

[0089] However, these alternatives still have some drawbacks:

[0090] On the one hand, rule-based contract review has limited review capabilities and can only conduct rule-based reviews on some simple and specific scenarios, such as reviewing content completeness and filling specifications. It cannot handle issues such as semantic understanding and logical reasoning in contracts, and it is difficult to discover some potential risks. On the other hand, the review flexibility and versatility are poor. Rule-based contract review cannot meet the review of contracts of different enterprises, different business scenarios, and different contract types. Developers are required to develop review rules for each category and each item, which increases costs.

[0091] Contract review based on machine learning is time-consuming and laborious. Using small models to review contracts requires annotating a large number of contract texts, identifying key information and risk points through training models, and the types of contracts and the businesses involved in contracts of different enterprises vary greatly. New data standards and model training are required for different enterprises, which is both time-consuming and laborious. Second, data dependence. The effect of small model contract review depends largely on the quality of training data. If the contract text data is incomplete, the format is not standardized, or there are ambiguities, it will greatly affect the results of the contract review. Usually, it is necessary to learn by annotating tens of thousands of contracts to achieve a good review effect. However, in reality, it is difficult to provide tens of thousands of high-quality contracts of different types, businesses, and scenarios, making the solution difficult to implement. Third, the small model has poor flexibility and versatility. The generalization ability of the small model is weak, and it cannot meet the review of contracts of different enterprises, different business scenarios, and different contract types. In practical projects, it is often necessary to carry out tuning training for the review ability of different enterprises, different scenarios, and different types of contracts. Fourth, the processing effect of complex contracts is not good. When faced with complex or special contract terms, the small model cannot fully understand its meaning and potential risks. For example, contracts involving highly specialized terms, complex logical structures, or metaphorical expressions may be difficult to analyze accurately. 5. False positives and false negatives of contract risks. The capabilities of small models are limited, and the accuracy of contract risk review is low, which requires a lot of manual review work, increasing labor costs and time investment.

[0092] In view of this, an embodiment of the present application provides a contract review solution based on a large model, which can at least solve some of the problems existing in the above-mentioned example solution.

[0093] The present application embodiment provides a contract review method, taking the method applied to a computer device as an example. The computer device may be Figure 1 The terminal 100 or the system consisting of the terminal 100 and the server 200. Figure 4 The contract review method specifically includes the following steps:

[0094] Step S100: for the contract to be reviewed, determine the target type of the contract and each atomic clause contained in the contract.

[0095] Specifically, the present application can classify and atomically design contracts in advance to achieve contract deconstruction. The following example illustrates a design model for a contract with two-level classification and three-level atomic terms.

[0096]

[0097] in,

[0098] CONT represents the first-level category to which the contract belongs, n represents the number of categories, cont represents the second-level subcategory, and c represents the third-level atomic clause.

[0099] The following example shows an optional contract type and atomic clause division method:

[0100] Contracts can be divided into 9 primary categories: CONT = {CONT 采购类合同 ,CONT 销售类合同 ,CONT 工程类合同 ,CONT 租赁类合同 ,CONT 合作类合同 ,CONT 劳务类合同 ,CONT 知识产权类合同 ,CONT 金融类合同 ,CONT 其它类合同}.

[0101] Furthermore, each major category of contracts can be classified into two levels: CONT 采购类合同 For example, CONT 采购类合同 ={cont 原材料采购合同 ,cont 设备采购合同 ,cont 宣传制品采购合同 ,…,cont 原易耗品采购合同}.

[0102] Furthermore, each specific contract template is split and designed into atomic clauses. The following is an example of a design scheme for atomic clauses: C = {c 合同基本信息条款 , c 合同标的条款 , c 合同价款条款 , c 合同付款条款 , c 合同履行条款 , c 合同质量条款 , c 合同验收条款 , c 合同权利与义务条款 , c 合同违约责任条款 , c 合同争议解决条款 , c 合同保密条款 , c 合同不可抗力条款 , c 合同生效条款 , …, c 合同其它条款}.

[0103] The above embodiment only uses a two-level classification and three-level atomic clause contract as an example to illustrate a structural method of a contract. In addition, other classification strategies and atomic clause contract methods can also be designed according to different enterprises, business scenarios, etc.

[0104] Based on the pre-designed contract types and atomic clauses, the target type of the contract to be reviewed and the atomic clauses contained in the contract can be determined.

[0105] In one possible implementation, the contract to be reviewed can be parsed in this step to identify the target type to which the contract belongs and the atomic clauses contained in the contract. This process can be implemented through a pre-configured neural network model or by calling the capabilities of a large model.

[0106] In another possible implementation, the present application may also adopt an atomic template contract design scheme, that is, configuring each atomic clause template. When generating a contract, the target type of the contract can be defined according to business needs, and the required atomic clause templates can be selected and assembled, and the atomic clause templates can be assembled into a contract. For a contract with review generated in this way, the target type of the contract and the atomic clauses contained in the contract can be directly obtained.

[0107] Step S110, query the mapping relationship between the atomic clauses and the review topics under the configured target type, and determine the review topic corresponding to each of the atomic clauses in the contract.

[0108] Specifically, when classifying and atomizing contracts, the present application can further configure the review topics corresponding to the contract types and atomic terms. That is, configure the combination of contract types and atomic terms and the mapping relationship between the review topics.

[0109] Taking the design model of the two-level classification and three-level atomic clause contract in the previous step as an example, build a contract review topic model:

[0110]

[0111] in,

[0112] THEME represents the first-level review theme, p represents the number of first-level review themes, and theme represents the second-level review sub-theme under the first-level review theme.

[0113] The following is an example of an optional way to divide the review topics:

[0114] The review topics of contracts can be divided into five primary categories:

[0115] THEME = {Theme合同主体风险类审查主题 , Theme 合同法务类审查主题 , Theme 合同财务类审查主题 , Theme 合同业务类审查主题 , Theme 合同合规类审查主题}.

[0116] Each major review topic can include several secondary review subtopics, starting with Theme. 合同主体风险类审查主题 For example, Theme 合同主体风险类审查主题 ={theme 黑名单审查 , theme 信用审查 , theme 司法风险审查 , theme 空壳情况审查 , theme 合同违约指数审查}.

[0117] The above only illustrates one optional design scheme for a contract review topic. In addition, corresponding review topics can be designed in combination with different corporate business types, contract classifications, and atomic clause contracts.

[0118] Further configure the mapping relationship between contract type, atomic clause and review subject. Specifically, the mapping relationship from CONT (cont) to THEME (theme) can be recorded through the mapping function F (f), so as to realize the mapping of contract text (atomic clause contract) to review subject (review sub-subject).

[0119] F:CONT→THEME

[0120] f:cont→theme

[0121] In this step, the review subject corresponding to each atomic clause in the contract is determined by querying the mapping relationship between the atomic clauses and the review subjects under the pre-configured target type.

[0122] Furthermore, after pre-configuring the mapping relationship between contract types, atomic clauses and review topics, this application can further combine contract types and review topics to construct review risk categories and levels.

[0123] In one possible implementation, risk categories can correspond to review topics one by one. Taking the five review topics in the previous steps as an example, the corresponding risk categories can include: contract subject risk, legal risk, financial risk, business risk and compliance risk. Each risk category can be further subdivided. Taking contract subject risk as an example, it can be subdivided into: blacklist risk items, credit risk items, judicial risk items, shell risk items, and contract default index risk items. The division of other major risk categories is similar.

[0124] The risk level can be set according to business needs. For example, it can be set to three levels: important risk, recommended modification and recommended attention. Of course, the risk level can also include a special level of no risk.

[0125] The following is an example of model design for risk categories and risk levels:

[0126]

[0127] Among them, Risk represents the risk category, risk represents the major category of the risk category, r represents the secondary sub-category of the risk category, RiskRank represents the risk level, and rank represents the risk level. k represents the kth risk level.

[0128] Step S120: for each of the atomic clauses, construct a review instruction prompt according to the corresponding review subject, and the prompt is used to instruct the big model to review the risk results of the input atomic clause according to the corresponding review subject.

[0129] Specifically, in order to be able to call the big model to perform risk review on each atomic clause of the contract, this step requires building a corresponding review instruction prompt for each atomic clause, so as to call the big model's capabilities through the prompt and complete the risk review of the corresponding atomic clause.

[0130] When constructing the review instruction prompt corresponding to the atomic clause, the review instruction prompt can be constructed according to the review subject corresponding to the atomic clause, instructing the large model to review the risk results of the input atomic clause contract according to the corresponding review subject, and obtain the risk review results of each atomic clause.

[0131] Step S130: Call the configured big model through the prompt corresponding to each of the atomic clauses to obtain the risk review results of each of the atomic clauses output by the big model.

[0132] After constructing the review instruction prompt corresponding to each atomic clause in the previous step, the review instruction prompts of each atomic clause in the contract to be reviewed form a prompt set, which corresponds to the contract to be reviewed.

[0133] Use each prompt in the prompt set to call the big model separately to obtain the risk review results of each atomic clause output by the big model.

[0134] The method provided in the embodiment of the present application realizes the automatic review of contracts by calling the big model, thereby improving the review efficiency. By leveraging the big model as a base capability, it can meet the contract review needs of different enterprises, different contract types, and different business scenarios.

[0135] In addition, compared to training a small model to implement contract review, this application is based on the capabilities of a large model, which can avoid problems such as small model training and data dependence. There is no need to label and train contract data, and fine-tuning based on the capabilities of a general large model can support contract review work.

[0136] By constructing a review instruction prompt corresponding to each atomic clause, contract review based on natural language instructions can be implemented, lowering the technical threshold so that ordinary business personnel can also complete the construction of a synthetic review prompt project.

[0137] Furthermore, based on the pre-established mapping relationship between contract types, atomic clauses and review topics, as well as prompts constructed according to review topics, it is possible to better stimulate the capabilities of large models, improve the accuracy of contract review, reduce false positives and omissions of contract risks, and reduce the workload of manual review.

[0138] In some embodiments of the present application, the aforementioned step S120 is described as to the process of constructing a review instruction prompt for each atomic clause according to the corresponding review subject.

[0139] In order to reduce the difficulty of prompt construction, a semi-structured prompt format template is provided in this embodiment. Referring to Table 1 below, the template may include some or all of the following components:

[0140] 1 Background Introduction, Character Setting, Task Description, Review Topic, Review Knowledge, Input Data, Output Requirements, and Sample Example.

[0141] Table 1

[0142]

[0143]

[0144] Among them, items 1-3 are fixed contents, and items 4-8 need to be combined with different review topics, match different review knowledge, input data (atomic clause contracts), output requirements and samples, etc. In this way, the prompt is quickly implemented through the semi-structured prompt design.

[0145] It should be noted that, among the prompt components in the above table 1, the third, fourth and sixth items are required, and the remaining items are optional.

[0146] The background introduction mainly describes the background of the contract review. For example, Background Introduction is "using a large model to complete the risk review of the contract."

[0147] Character setting is mainly to assign roles to large models. For example, Character Setting is "Assume that you are a contract review expert with rich experience in contract review. Your goal is to conduct a rigorous review of the contract content."

[0148] The task description is used to instruct the large model to review the risk results of the atomic clauses in the input data slot according to the review topic in the review topic slot. For example, the Task Description is "output the risk result review of the input data according to the review topic".

[0149] The review topic slot is used to fill the review topic corresponding to the atomic clause to be reviewed.

[0150] The review knowledge slot is used to fill the review knowledge corresponding to the review subject to which the atomic clause belongs. The review knowledge may include knowledge related to assisting the large model in contract risk review. The review knowledge is generally described in natural language. Taking the review results including risk categories and levels as an example, the present application can pre-construct a risk quantification assessment system based on natural language according to the risk categories and risk levels that may exist in the sorted contracts.

[0151] The risk quantification model R can be expressed in JSON format llm , risk quantification model R llm Including risk categories and risk levels, where the risk level needs to be described in natural language. Taking blacklist risk as an example, the risk quantification model R llm as follows:

[0152] R llm ={“Risk”:”Enterprise Blacklist Risk”, “RiskRank”: “If an enterprise is included in the abnormal operation list or the serious illegal dishonesty list of the enterprise credit information disclosure system, the risk level will be a major risk if it is not included in either list. If it is not included in either list, no risk will be prompted.”}.

[0153] The same method can be used to complete the natural language-based risk quantification definition for other risk categories.

[0154] The risk quantification model R constructed above llmCan be added to the review instruction prompt as review knowledge.

[0155] The input data slot is used to populate the atomic clause contract to be reviewed.

[0156] The output requirement slot is used to fill in the specified output form of the specified review result.

[0157] The sample slot is used to fill in the reference sample corresponding to the review topic to which the atomic clause belongs.

[0158] Based on the above introduction to the review instruction prompt format template, the process of constructing a review instruction prompt for atomic clauses is described.

[0159] In a possible implementation, the review instruction prompt format template is obtained, and the template includes the contents of items 3, 4, and 6 in Table 1 above. For each atomic clause, its corresponding review subject is filled into the review subject slot in the format template, and the atomic clause is filled into the input data slot to obtain the review instruction prompt corresponding to the atomic clause.

[0160] In another possible implementation, a review instruction prompt format template is obtained, and the template includes the contents of items 3, 4, 5, and 6 in Table 1 above. Compared with the previous implementation, the template in this embodiment adds a review knowledge slot of item 5. Correspondingly, the task description of item 3 is specifically used to instruct the large model to review the risk results of the atomic clauses in the input data slot according to the review subject in the review subject slot and based on the review knowledge in the review knowledge slot.

[0161] For each atomic clause:

[0162] Get the configured review knowledge corresponding to the review subject to which the atomic clause belongs.

[0163] Fill the review subject corresponding to the atomic clause into the review subject slot in the format template, fill the atomic clause into the input data slot, fill the acquired review knowledge into the review knowledge slot, and obtain the review instruction prompt corresponding to the atomic clause.

[0164] By adding review knowledge information to the prompt, richer review knowledge information can be provided to the large model, assisting the large model in giving better review results.

[0165] In another possible implementation, the review instruction prompt format template is obtained, and the template includes the contents of items 3, 4, 6, and 7 in Table 1. Compared with the first implementation, the template in this embodiment adds an output requirement slot in item 7. Correspondingly, the task description in item 3 is also used to instruct the large model to output the review result in the output form specified in the output requirement slot.

[0166] For each atomic clause:

[0167] Get the configured output form corresponding to the review subject to which the atomic clause belongs, which may include risk type and risk level.

[0168] Fill the review subject corresponding to the atomic clause into the review subject slot in the format template, fill the atomic clause into the input data slot, fill the obtained output form into the output requirement slot, and obtain the review instruction prompt corresponding to the atomic clause.

[0169] By adding specified output requirements in the prompt, the large model can output the risk review results in the specified output format, that is, the risk type and risk level of the atomic clause can be output.

[0170] In another possible implementation, a review instruction prompt format template is obtained, and the template includes the contents of items 3, 4, 6 and 8 in Table 1 above. Compared with the first implementation, the template in this embodiment adds a sample slot of item 8. Correspondingly, the task description of item 3 is specifically used to instruct the large model to refer to the given sample in the sample slot, and to review the risk results of the atomic clauses in the input data slot according to the review subject in the review subject slot.

[0171] For each atomic clause:

[0172] Get the configured reference sample corresponding to the review topic to which the atomic clause belongs.

[0173] Fill the review subject corresponding to the atomic clause into the review subject slot in the format template, fill the atomic clause into the input data slot, fill the obtained reference sample into the sample slot, and obtain the review instruction prompt corresponding to the atomic clause.

[0174] By adding reference examples to the prompt, it is easier for the large model to understand the task requirements and the form of input and output, reducing the probability of hallucination problems in the large model and improving the accuracy of the output risk review results.

[0175] The above examples illustrate several composition examples of the review instruction prompt format template. It can be understood that in addition to this, other combinations of the contents in Table 1 can be made, which are not listed one by one in this embodiment.

[0176] In some embodiments of the present application, for a large model to be configured, the number thereof may be 1 or more.

[0177] When there is only one large model, the output result of the large model can be directly output as the final result.

[0178] When the big model consists of two or more different types of big models, for each atomic clause:

[0179] Through the prompt corresponding to the atomic clause, each type of big model is called separately to obtain the risk review results of the atomic clause output by each type of big model.

[0180] According to the weights of various types of large models under the review topics corresponding to the atomic clauses, the risk review results of the atomic clauses output by all types of large models are voted on to obtain the final risk review results of the atomic clauses.

[0181] That is, when there are more than two large models, weighted voting can be performed on the risk review results output by each large model to determine the final risk review results of the atomic clauses.

[0182] By fusing the output results of multiple large models of different types, the accuracy of risk review results can be improved.

[0183] Different types of big models may focus on different review topics. Therefore, the weight values ​​of different types of big models can be designed in advance for each review topic. When reviewing atomic clauses, the weights of various types of big models can be found according to the review topics to which the atomic clauses belong, and then the risk review results output by various types of big models can be voted according to the weights to obtain the final risk review results of the atomic clauses.

[0184] The specific number of large models can be multiple. Considering the resources and efficiency, three different types of large models can be configured in this embodiment. Then the final risk review result of the atomic clause Result = Vote (F1, F2, F3), where F1-F3 represent the output results of three different large models respectively.

[0185] Assume that the initial weights of each review topic i of F1, F2, and F3 are [α i ,β i ,γ i ]. Among them, the initial weight α i ,β i ,γ i Can be the same or different.

[0186] The risk review results output by the three types of large models are voted on to obtain the final risk review results of the atomic clauses. There are three situations as follows:

[0187] (1) The voting result is 3:0, that is, the risk review results output by the three large models are consistent. i ,β i ,γ i If the two are the same, we can randomly select the result of a large model as the final risk review result: Result = random.choice(α i F1,β i F2,γ i F3).

[0188] (2) The voting result is 2:1, that is, the risk review results output by the two large models are consistent, but inconsistent with the output of the third large model. Then, one seat is selected as the final risk review result from the output results of the two large models with the same voting results. If the weights of the two are the same, one is selected randomly; if the weights are different, the one with the larger weight is selected.

[0189] (3) The voting result is 1:1:1, that is, the risk review results output by the three large models are inconsistent. At this time, the output results of the three large models can be pushed to the user for verification to confirm the final risk review result.

[0190] In one possible implementation, in order to improve the voting effect of the output results of multiple large models, this embodiment can also dynamically adjust the weights of various types of large models according to the user's adoption of the final risk review results.

[0191] Specifically, the user's adoption of the final risk review result of each atomic clause is obtained, and the weight values ​​of various types of large models under the review subject corresponding to the atomic clause are adjusted according to the adoption status.

[0192] Take Result = Vote (F1, F2, F3) as an example:

[0193] There are three situations in which users adopt the final risk review results:

[0194] (1) When the voting result is 3:0, if the user adopts the final risk review result, the α i ,β i ,γ i If the user does not adopt the final risk review result, the α value under the review topic corresponding to the atomic clause is maintained. i ,β i ,γi On this basis, the review instruction prompt of the corresponding review topic can be adjusted to ensure the accuracy of the large model review results.

[0195] (2) When the voting result is 2:1, if the user adopts the final risk review result, the weights of the two large models with a voting result of 2 under the review topic corresponding to the atomic clause will be increased by the set value. If the user does not adopt the final risk review result, the weight of the large model with a voting result of 1 under the review topic corresponding to the atomic clause will be increased by the set value.

[0196] (3) When the voting result is 1:1:1, combined with the results of manual verification, the weight of the correctly verified large model under the review topic corresponding to the atomic clause is increased by a set value.

[0197] In some embodiments of the present application, another implementation scheme of the contract review method is provided. Compared with the contract review method introduced in the aforementioned embodiment, the present embodiment can further add a step of reviewing the consistency of the full text of the contract.

[0198] Specifically:

[0199] The big model is called to instruct the big model to perform a consistency review on the full text of the contract to be reviewed and obtain the consistency review result.

[0200] The consistency review includes but is not limited to the following:

[0201] Consistency review of enterprise name, consistency review of amount, consistency review of contract signing parties, etc.

[0202] In this embodiment, based on the risk review of each atomic clause in the contract introduced in the previous embodiment, a consistency review process of the entire contract is further added, which can check whether there is any consistency risk in the contract based on the entire text.

[0203] In summary:

[0204] The contract review method of this application demonstrates significant technical advantages in terms of versatility and generalization capability, data dependence and training cost, review accuracy and efficiency, as well as technical implementation difficulty and business personnel participation.

[0205] 1. Stronger versatility and generalization: Whether it is rule-based review or small machine learning models, it is usually necessary to customize specific rules for specific types of contracts or a large amount of data annotation and model training, resulting in weak versatility and generalization. The large model technology used in this application can cover the contract review needs of different enterprises, various contract types, various scenarios and different business types through a unified capability base, and has stronger versatility and generalization.

[0206] 2. Reduce data dependence and training costs: The small model contract review method based on machine learning is highly dependent on data and requires a large amount of high-quality, accurately labeled contract data for training. When the model is migrated to a new type of contract, it needs to be retrained, which is costly. However, this application uses large model technology to fine-tune based on the general large model capabilities, without the need for a large amount of labeled data and complex training process, which significantly reduces data dependence and training costs.

[0207] 3. Improve the accuracy and efficiency of review: When dealing with complex contracts, small machine learning models often have problems with insufficient understanding, risk false positives and false negatives, which affect the accuracy and efficiency of the review. This application improves the accuracy of contract review, reduces false positives and false negatives, and reduces the workload of manual review and improves review efficiency by building prompt-based contract review engineering management capabilities, using multi-model fusion intelligent review voting and result recommendation mechanisms, and combining differentiated manual verification and tuning solutions.

[0208] 4. Simplify the difficulty of technical implementation and improve the participation of business personnel: Traditional contract review methods have high technical requirements and it is difficult for business personnel to directly participate. However, this application reduces the technical threshold by building a large model based on the review instruction prompt, allowing business personnel to participate in building their own review instruction prompt project, greatly simplifying the difficulty of technical implementation and improving the participation of business personnel in the contract review process.

[0209] In some embodiments of the present application, a business module design structure of a contract review system is introduced, combined with Figure 5 As shown:

[0210] The main modules of the contract review system based on the big model may include: atomic contract management module 11, risk system management module 12, prompt review engineering module 13 and multi-model fusion review module 14.

[0211] The main functions of the atomic contract management module 11 include contract classification management, atomic clause contract management, contract review subject management, and review subject mapping management. This module mainly implements the classification of contracts and the design of review subjects, and completes the atomic splitting of contract terms and the mapping of atomic contract terms and subjects in combination with the review subjects, thus completing the deconstruction of the contract.

[0212] The risk system management module 12 mainly includes two functions: contract risk classification management and risk quantitative assessment management. Through this module, contract risk classification based on review subject and the construction of contract risk quantitative assessment system based on natural language are realized.

[0213] The prompt review project module 13 has two main functions: prompt project category management and prompt project structured configuration management. This module implements prompt classification and structured management, simplifies the technical implementation difficulty of contract review, and allows business personnel to easily and conveniently build contract review prompt projects, so that business personnel can use large model capabilities to complete automatic contract review work.

[0214] The multi-model fusion review module 14 has the following main functions: LLM multi-model management, review voting management, manual verification management, etc. It mainly completes the contract review based on the large model capabilities, and votes on the review results of multiple models. Finally, it optimizes the weight and prompt engineering of the large model based on the manual verification results.

[0215] In some embodiments of the present application, a layered architecture design of a contract review system is introduced, combined with Figure 6 As shown:

[0216] From top to bottom, it can be divided into application layer, model layer and storage layer. The layered design mode realizes the layered decoupling of each business.

[0217] The application layer mainly includes the atomic contract management module, the risk system management module, the prompt review engineering module and the multi-model fusion review module, and the construction of business functions is completed through the application layer.

[0218] The model layer mainly includes the model scheduling management module and the corresponding large model. The model scheduling management module is mainly used to realize the resource scheduling management of multiple large models. In practice, polling scheduling can be used to achieve efficient utilization of resources. Figure 6 Three large models are used as examples to illustrate.

[0219] The storage layer mainly designs the contract text library, risk management library, prompt review project library and voting management library to achieve persistent storage of data in the business process.

[0220] The contract review device provided in an embodiment of the present application is described below. The contract review device described below and the contract review method described above can be referenced to each other.

[0221] See also Figure 7 , Figure 7 A schematic diagram of the structure of a contract review device disclosed in an embodiment of the present application.

[0222] like Figure 7 As shown, the device may include:

[0223] The contract type and clause determination unit 21 is used to determine the target type of the contract and each atomic clause contained in the contract for the contract to be reviewed;

[0224] The review subject determination unit 22 is used to query the mapping relationship between the atomic clauses and the review subjects under the configured target type, and determine the review subject corresponding to each of the atomic clauses in the contract;

[0225] The review instruction construction unit 23 is used to construct a review instruction prompt for each of the atomic clauses according to the corresponding review subject, and the prompt is used to instruct the large model to review the risk results for the input atomic clauses according to the corresponding review subject;

[0226] The clause review unit 24 is used to call the configured big model through the prompt corresponding to each of the atomic clauses to obtain the risk review result of each of the atomic clauses output by the big model.

[0227] In a possible implementation, the review instruction construction unit constructs a review instruction prompt for each of the atomic clauses according to the corresponding review subject, including:

[0228] Obtain the configured review instruction prompt format template, which includes a task description, a review subject slot, and an input data slot. The task description is used to instruct the big model to review the risk results of the atomic clauses in the input data slot according to the review subject in the review subject slot;

[0229] For each of the atomic clauses, fill its corresponding review subject into the review subject slot, fill the atomic clause into the input data slot, and obtain the review instruction prompt corresponding to the atomic clause.

[0230] In one possible implementation, the review instruction prompt format template also includes: a review knowledge slot, and the task description is specifically used to instruct the large model to review the risk results of the atomic clauses in the input data slot according to the review topic in the review topic slot and based on the review knowledge in the review knowledge slot.

[0231] The process of the review instruction construction unit constructing the review instruction prompt for each of the atomic clauses according to the corresponding review subject also includes:

[0232] For each of the atomic clauses, obtaining the configured review knowledge corresponding to the review subject to which the atomic clause belongs;

[0233] Fill the acquired review knowledge into the review knowledge slot to obtain the review instruction prompt corresponding to the atomic clause.

[0234] In a possible implementation, the review instruction prompt format template also includes: an output requirement slot, and the task description is also used to instruct the large model to output the review result in the output form specified in the output requirement slot.

[0235] The process of the review instruction construction unit constructing the review instruction prompt for each of the atomic clauses according to the corresponding review subject also includes:

[0236] For each of the atomic clauses, obtaining a configured output form corresponding to the review subject to which the atomic clause belongs, the output form including a risk type and a risk level;

[0237] Fill the obtained output form into the output requirement slot to obtain the review instruction prompt corresponding to the atomic clause.

[0238] In one possible implementation, the review instruction prompt format template also includes: a sample slot, and the task description is specifically used to instruct the large model to refer to the sample given in the sample slot and review the risk results of the atomic clauses in the input data slot according to the review topic in the review topic slot.

[0239] The process of the review instruction construction unit constructing the review instruction prompt for each of the atomic clauses according to the corresponding review subject also includes:

[0240] For each of the atomic clauses, obtain a reference sample configured to correspond to the examination subject to which the atomic clause belongs;

[0241] Fill the acquired reference sample into the sample slot to obtain the review instruction prompt corresponding to the atomic clause.

[0242] In a possible implementation, the configured big model includes two or more different types of big models, and the clause review unit calls the configured big model through the prompt corresponding to each of the atomic clauses to obtain the risk review result of each of the atomic clauses output by the big model, including:

[0243] For each of the atomic clauses:

[0244] Through the prompt corresponding to the atomic clause, each type of large model is called respectively to obtain the risk review result of the atomic clause output by each type of large model;

[0245] According to the weights of various types of large models under the review topics corresponding to the atomic clauses, the risk review results of the atomic clauses output by all types of large models are voted on to obtain the final risk review results of the atomic clauses.

[0246] In a possible implementation, the device of the present application may further include:

[0247] The weight adjustment unit is used to obtain the user's adoption of the final risk review result of each atomic clause, and adjust the weight value of each type of large model under the review subject corresponding to the atomic clause according to the adoption status.

[0248] In a possible implementation, the device of the present application may further include:

[0249] A full text consistency review unit, used for calling the big model to instruct the big model to conduct a consistency review on the full text of the contract to be reviewed and obtain a consistency review result;

[0250] The consistency review includes at least one of the following:

[0251] Consistency review of enterprise name, amount and contract signing parties.

[0252] The present application also provides an electronic device in an embodiment. Figure 8 As shown, it shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiment of the present application. The electronic device in the embodiment of the present application may include but is not limited to fixed terminals such as mobile phones, tablet computers, teaching large screens, wearable devices, etc. Figure 8 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0253] like Figure 7As shown, the electronic device may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603 to implement the contract review method of the aforementioned embodiment of the present application. When the electronic device is powered on, various programs and data required for the operation of the electronic device are also stored in the RAM 603. The processing device 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0254] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a memory card, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although Figure 8 An electronic device having various devices is shown, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0255] Also provided in an embodiment of the present application is a computer program product including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any one of the contract review methods provided in the embodiments of the present application.

[0256] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When one or more computer programs are executed by an electronic device, the electronic device can implement any contract review method provided in the embodiment of the present application.

[0257] It should also be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines.

[0258] Through the description of the above implementation mode, the technicians in the field can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. In general, all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits. However, for the present application, software program implementation is a better implementation mode in more cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a ROM, a RAM, a disk or an optical disk, etc., including a number of instructions to enable a computer device (which can be a personal computer, a training device, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0259] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0260] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a training device, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, training device, or data center. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0261] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

Claims

1. A contract review method, characterized in that: include: For the contract to be reviewed, determine the target type of the contract and each atomic clause contained in the contract; Query the mapping relationship between the atomic clauses and the review subjects under the configured target type, and determine the review subject corresponding to each atomic clause in the contract; For each of the atomic clauses, a review instruction prompt is constructed according to the corresponding review subject, and the prompt is used to instruct the big model to review the risk results of the input atomic clause according to the corresponding review subject; Through the prompt corresponding to each of the atomic clauses, the configured big model is called to obtain the risk review results of each of the atomic clauses output by the big model.

2. The method according to claim 1, characterized in that For each of the atomic clauses, the process of constructing the review instruction prompt according to the corresponding review subject includes: Obtain the configured review instruction prompt format template, which includes a task description, a review subject slot, and an input data slot. The task description is used to instruct the big model to review the risk results of the atomic clauses in the input data slot according to the review subject in the review subject slot; For each of the atomic clauses, fill its corresponding review subject into the review subject slot, fill the atomic clause into the input data slot, and obtain the review instruction prompt corresponding to the atomic clause.

3. The method according to claim 2, characterized in that The template further includes: a review knowledge slot, wherein the task description is specifically used to instruct the large model to review the risk results of the atomic clauses in the input data slot according to the review subject in the review subject slot and based on the review knowledge in the review knowledge slot; The process of constructing the review instruction prompt also includes: For each of the atomic clauses, obtaining the configured review knowledge corresponding to the review subject to which the atomic clause belongs; Fill the acquired review knowledge into the review knowledge slot to obtain the review instruction prompt corresponding to the atomic clause.

4. The method according to claim 2, characterized in that: The template further includes: an output requirement slot, and the task description is further used to instruct the large model to output the review result in the output form specified in the output requirement slot; The process of constructing the review instruction prompt also includes: For each of the atomic clauses, obtaining a configured output form corresponding to the review subject to which the atomic clause belongs, the output form including a risk type and a risk level; Fill the obtained output form into the output requirement slot to obtain the review instruction prompt corresponding to the atomic clause.

5. The method according to claim 2, characterized in that: The template further includes: a sample slot, wherein the task description is specifically used to instruct the large model to refer to the sample given in the sample slot and review the risk results of the atomic clauses in the input data slot according to the review subject in the review subject slot; The process of constructing the review instruction prompt also includes: For each of the atomic clauses, obtain a reference sample configured to correspond to the examination subject to which the atomic clause belongs; Fill the acquired reference sample into the sample slot to obtain the review instruction prompt corresponding to the atomic clause.

6. The method according to claim 1, characterized in that If the configured big model includes two or more different types of big models, the process of calling the configured big model through the prompt corresponding to each of the atomic clauses to obtain the risk review result of each of the atomic clauses output by the big model includes: For each of the atomic clauses: Through the prompt corresponding to the atomic clause, each type of large model is called respectively to obtain the risk review result of the atomic clause output by each type of large model; According to the weights of various types of large models under the review topics corresponding to the atomic clauses, the risk review results of the atomic clauses output by all types of large models are voted on to obtain the final risk review results of the atomic clauses.

7. The method according to claim 6, characterized in that Also includes: Obtain the user's adoption status of the final risk review results of each of the atomic clauses, and adjust the weight values ​​of various types of large models under the review topics corresponding to the atomic clauses according to the adoption status.

8. The method according to any one of claims 1 to 7, characterized in that: Also includes: Calling the big model to instruct the big model to perform consistency review on the full text of the contract to be reviewed, and obtaining a consistency review result; The consistency review includes at least one of the following: Consistency review of enterprise name, amount and contract signing parties.

9. A contract review device, characterized in that: include: A contract type and clause determination unit, used to determine the target type of the contract and each atomic clause contained in the contract for review; A review subject determination unit, used for querying the mapping relationship between the atomic clauses and the review subjects under the configured target type, and determining the review subject corresponding to each of the atomic clauses in the contract; A review instruction construction unit, used to construct a review instruction prompt for each of the atomic clauses according to the corresponding review subject, wherein the prompt is used to instruct the large model to review the risk results for the input atomic clauses according to the corresponding review subject; The clause review unit is used to call the configured big model through the prompt corresponding to each of the atomic clauses to obtain the risk review results of each of the atomic clauses output by the big model.

10. An electronic device, characterized in that: include: Memory and processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the contract review method as described in any one of claims 1 to 8.

11. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the contract review method as claimed in any one of claims 1 to 8 is implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, each step of the contract review method as described in any one of claims 1 to 8 is implemented.

Images