Enterprise compliance review methods, devices, equipment and storage media

Abstract

The present application relates to the field of artificial intelligence technology, and discloses a corporate compliance review method, device, equipment, and storage medium. The method includes: obtaining the original text containing the information of the subject to be reviewed; inputting the original text into a preset compliance review model to obtain the compliance review result; generating a corresponding compliance rectification plan based on the compliance review result; calling an API interface to execute the rectification task in the compliance rectification plan. The embodiments of the present application can improve the review efficiency and review accuracy of corporate compliance management.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and storage medium for corporate compliance review. Background Art

[0002] Corporate compliance management is a key link in ensuring the legal operation and risk control of enterprises. It requires enterprises to comply with various laws and regulations, industry standards and internal policies to ensure that their business activities do not violate relevant regulations.

[0003] In the relevant technologies, the review of corporate compliance mainly relies on manual inspection and document management, which is difficult to handle large and complex regulatory texts, resulting in low review efficiency. Secondly, existing automated tools lack in-depth understanding and accurate analysis of regulatory texts, and cannot provide accurate compliance analysis reports. Summary of the invention

[0004] The purpose of this application is to provide a corporate compliance review method, device, equipment and storage medium that can improve the review efficiency and accuracy of corporate compliance management.

[0005] The present application provides a method for corporate compliance review, including:

[0006] Obtaining the original text containing the information of the subject to be reviewed;

[0007] Inputting the original text into a preset compliance review model to obtain a compliance review result; the compliance review model is obtained by training a language model based on a Transformer architecture using sample original texts, sample compliance texts and corresponding sample compliance review results;

[0008] Generate corresponding compliance rectification plans based on the compliance review results;

[0009] Call the API interface to execute the rectification tasks in the compliance rectification plan;

[0010] The inputting the original text into a preset compliance review model to obtain a compliance review result includes:

[0011] Processing the original text to obtain key text in the original text; the key text is text describing information related to the object to be reviewed and the compliance review;

[0012] Encode the key text to obtain a key text semantic vector and a semantic vector to be examined obtained by fitting the key text semantic vectors;

[0013] Comparing the similarity between the semantic vector to be reviewed and the preset compliant semantic vector to obtain a plurality of target compliant semantic vectors; the compliant semantic vector is obtained by encoding the preset compliant text, and the compliant semantic vector is the compliant semantic vector with the highest similarity to the semantic vector to be reviewed;

[0014] Based on the attention mechanism, a compliance comparison is performed on the key text semantic vector and the target compliance semantic vector, and the compliance review result is generated according to the compliance comparison result.

[0015] In some embodiments, before inputting the original text into a preset compliance review model to obtain a compliance review result, the process further includes:

[0016] obtaining the compliance text;

[0017] Performing word segmentation processing on the compliant text to obtain a compliant word text;

[0018] The compliant word text is input into a preset encoding model to encode and normalize the key text to obtain a compliant semantic vector.

[0019] In some embodiments, encoding the key text to obtain the key text semantic vector and the semantic vector to be examined obtained by fitting each of the key text semantic vectors includes:

[0020] The key text is input into a preset encoding model to encode the key text to obtain the key text semantic vector, and a vector average operation or a vector weighted average operation is performed on the key text semantic vector to obtain the semantic vector to be reviewed.

[0021] In some embodiments, the similarity comparison between the semantic vector to be examined and the preset compliant semantic vector to obtain a plurality of target compliant semantic vectors includes:

[0022] Calculating the vector similarity between the semantic vector to be reviewed and the compliant semantic vector;

[0023] According to the vector similarity, a preset number of the target compliant semantic vectors are obtained.

[0024] In some embodiments, the compliance comparison between the key text semantic vector and the target compliance semantic vector is performed based on the attention mechanism, and the compliance review result is generated according to the compliance comparison result, including:

[0025] Taking the target compliance semantic vector as the context, through semantic matching and context understanding, the key text semantic vector and the target compliance semantic vector are compared for compliance and the comparison result is converted into text form to obtain a compliance judgment result;

[0026] Generate corresponding compliance analysis results according to the compliance judgment results;

[0027] Import the compliance judgment result and the compliance analysis result into a preset text template to obtain the compliance review result.

[0028] In some embodiments, generating a corresponding compliance rectification plan according to the compliance review result includes:

[0029] Processing the compliance review results to obtain key text, target compliance text, compliance judgment results and compliance analysis results in the compliance review results;

[0030] Taking the target compliance text, the compliance judgment result and the compliance analysis result as the context, a rectification text containing rectification information is generated, and the key text and the rectification text are fitted to obtain the compliance rectification plan.

[0031] In some embodiments, calling the API interface to execute the rectification task in the compliance rectification plan includes:

[0032] Processing the compliance rectification plan to obtain the rectification text in the compliance rectification plan;

[0033] Generate corresponding rectification tasks according to the rectification text;

[0034] The API interface is called to link the corresponding terminal and execute the rectification task with the terminal.

[0035] The present application also provides a corporate compliance review device, including:

[0036] The first module is used to obtain the original text containing the information of the object to be reviewed;

[0037] The second module is used to input the original text into a preset compliance review model to obtain a compliance review result; the compliance review model is obtained by training a language model based on a Transformer architecture using sample original texts, sample compliance texts and corresponding sample compliance review results;

[0038] The third module is used to generate a corresponding compliance rectification plan based on the compliance review results;

[0039] The fourth module is used to call the API interface to execute the rectification tasks in the compliance rectification plan;

[0040] The inputting the original text into a preset compliance review model to obtain a compliance review result includes:

[0041] Processing the original text to obtain key text in the original text; the key text is text describing information related to the object to be reviewed and the compliance review;

[0042] Encode the key text to obtain a key text semantic vector and a semantic vector to be examined obtained by fitting the key text semantic vectors;

[0043] Comparing the similarity between the semantic vector to be reviewed and the preset compliant semantic vector to obtain a plurality of target compliant semantic vectors; the compliant semantic vector is obtained by encoding the preset compliant text, and the compliant semantic vector is the compliant semantic vector with the highest similarity to the semantic vector to be reviewed;

[0044] Based on the attention mechanism, a compliance comparison is performed on the key text semantic vector and the target compliance semantic vector, and the compliance review result is generated according to the compliance comparison result.

[0045] An embodiment of the present application also provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned corporate compliance review method when executing the computer program.

[0046] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned enterprise compliance review method is implemented.

[0047] The beneficial effects of the present application are as follows: the original text containing the information of the object to be reviewed is processed by using the compliance review model trained by the language model based on the Transformer architecture to obtain the compliance review result, the corresponding compliance rectification plan is generated according to the compliance review result, and the rectification task in the compliance rectification plan is executed by calling the API interface, wherein the compliance review model selects the corresponding target compliance semantic vector based on the key text in the original text, and performs compliance comparison between the key text semantic vector and the target compliance semantic vector based on the attention mechanism, so as to generate the compliance review result according to the compliance comparison result. Since the compliance review model is used to conduct compliance review on the original text, the compliance text of the compliance review clause involved in the object to be reviewed is determined based on the original text, and then the key text in the original text is compared with the compliance text based on the attention mechanism, the review efficiency and accuracy of the enterprise compliance management can be improved. At the same time, the corresponding compliance rectification plan is generated according to the compliance review result and the rectification tasks in the compliance rectification plan are automatically executed, so as to realize the automated execution of compliance rectification and improve the rectification efficiency of the enterprise compliance management. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1It is a flow chart of the enterprise compliance review method provided in the embodiment of the present application.

[0049] Figure 2 It is a flowchart of the method before step S120 provided in an embodiment of the present application.

[0050] Figure 3 It is a flowchart of the specific method of step S124 provided in an embodiment of the present application.

[0051] Figure 4 It is a flowchart of the specific method of step S140 provided in an embodiment of the present application.

[0052] Figure 5 It is a structural diagram of the enterprise compliance review device provided in an embodiment of the present application.

[0053] Figure 6 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0055] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown can be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second" and the like in the specification, claims and drawings are used to distinguish similar objects, and are not used to describe a specific order or sequence.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0057] See also Figure 1 , Figure 1 is a flow chart of the enterprise compliance review method provided in the embodiments of the present application. In some embodiments of the present application, Figure 1 The method includes but is not limited to steps S110 to S140. Figure 1 These four steps are introduced in detail.

[0058] Step S110, obtaining the original text containing the information of the object to be reviewed.

[0059] It should be noted that the objects to be reviewed cover all business areas of the enterprise, and the original text is a text recording the information of the objects to be reviewed. In some embodiments, the objects to be reviewed may include organizational structure, human resource management, financial management, marketing, R&D management, procurement management, production management, environmental protection, information security and corporate culture construction, and the information of the objects to be reviewed may include information on the generation process of the objects to be reviewed, actual event information, personnel information and related subject information.

[0060] In step S110, the original text containing the information of the object to be reviewed is obtained, which can be obtained in the target terminal. By searching for keywords corresponding to the information of the object to be reviewed, the corresponding original text is captured from the target terminal, and then manually checked to eliminate the wrongly selected text and fill in the missed text.

[0061] Step S120, inputting the original text into a preset compliance review model to obtain a compliance review result.

[0062] Among them, the compliance review model is obtained by training a language model based on the Transformer architecture using sample original text, sample compliance text and corresponding sample compliance review results.

[0063] See also Figure 2 , Figure 2 Flowchart of the method before step S120 provided in the embodiment of the present application. In some embodiments of the present application, before step S120, it may also include but is not limited to steps S210 to S230. Figure 2 These three steps are introduced in detail.

[0064] Step S210, obtaining compliance text.

[0065] Step S220, performing word segmentation processing on the compliant text to obtain a compliant word text.

[0066] Step S230, inputting the compliant word text into a preset encoding model to encode and normalize the key text to obtain a compliant semantic vector.

[0067] Among them, the encoding model is a vector encoding model based on the Transformer architecture, which uses the masked language model task for pre-training on large-scale data to obtain high-dimensional semantic vector representation.

[0068] In step S210 to step S230, after obtaining the compliant text, the compliant text is firstly segmented, and the segmentation includes segmentation, sentence segmentation, and stop word removal operations to obtain the compliant word text, and then the compliant word text is input into the pre-trained encoding model, and the compliant word text is encoded using the encoding model. After the encoding is completed, the high-dimensional semantic vector representation of the compliant word text is obtained, and the high-dimensional semantic vector representation of the compliant word text is vector normalized to ensure that the modulus of each vector is 1, and the compliant semantic vector is obtained after normalization. Among them, the normalization formula is:

[0069] ,

[0070] in, is the compliance semantic vector, is a high-dimensional semantic vector representation of the compliant word text, It is the norm of the high-dimensional semantic vector representation of the legal word text.

[0071] After obtaining the compliant semantic vector, store the compliant semantic vector in the vector library, assign a unique identifier to each compliant semantic vector, and create an index to support efficient retrieval. The index type usually selects the IVF_FLAT type, which achieves a good balance between retrieval speed and accuracy. First, determine the index parameters, including the number of cluster centers nlist, which is usually 1000, and then execute the index creation command to complete the index creation, and then insert the compliant semantic vectors and their corresponding identifiers into the database in batches.

[0072] Specifically, step S120 includes step S121 to step S124.

[0073] Step S121, processing the original text to obtain key text in the original text.

[0074] The key text is a text describing information related to the object to be reviewed and the compliance review.

[0075] In step S121, the input original text is processed based on the compliance review model, and the information describing the object to be reviewed and associated with the compliance review in the original text is extracted to obtain the key text.

[0076] Specifically, the compliance review model selects keywords from the original text, and performs multiple rounds of contextual semantic recognition on the selected keyword pairs in different context scenarios to identify the changes in the sensitivity of each keyword in different context scenarios. Based on the changes in the sensitivity of the keywords, the sentences and / or paragraphs containing the keywords are selected to obtain several key texts.

[0077] Step S122, encoding the key texts to obtain key text semantic vectors and semantic vectors to be examined obtained by fitting the semantic vectors of the key texts.

[0078] In step S122, after extracting the key text, each key text is semantically vectorized using a pre-trained vector encoding model to obtain a key text semantic vector. After completing the semantic vectorization of the key text, all key text semantic vectors are combined as the overall semantic vector representation of the object to be reviewed, thereby obtaining the semantic vector to be reviewed.

[0079] In a specific embodiment, step S122 specifically includes the following steps: inputting the key text into a preset encoding model to encode the key text to obtain a key text semantic vector, performing vector averaging operation or vector weighted averaging operation on the key text semantic vector to obtain a semantic vector to be reviewed.

[0080] Specifically, the key text is input into a pre-trained encoding model, and the encoding model is used to encode the key text. After the encoding is completed, the key text semantic vector is obtained, and the high-dimensional semantic vectors of all key texts are combined to form an overall semantic vector representation of the object to be reviewed. The combination method can adopt vector averaging operation or vector weighted averaging operation to obtain the semantic vector to be reviewed.

[0081] In this embodiment, a vector average operation method is used to generate a semantic vector to be examined. The specific formula is:

[0082] ,

[0083] in, is the semantic vector to be examined, is the high-dimensional semantic vector of the j-th key text, j=1,2,...,n, and n is the total number of high-dimensional semantic vectors of the key text.

[0084] Step S123 : performing a similarity comparison between the semantic vector to be reviewed and the preset compliant semantic vector to obtain a plurality of target compliant semantic vectors.

[0085] The compliance semantic vector is obtained by encoding the preset compliance text, and the compliance semantic vector is the compliance semantic vector with the highest similarity to the semantic vector to be reviewed.

[0086] In a specific embodiment, step S123 specifically includes the following steps: calculating the vector similarity between the semantic vector to be reviewed and the compliant semantic vector; and obtaining a preset number of target compliant semantic vectors according to the vector similarity.

[0087] In step S123, the semantic vector to be reviewed is used as a query vector to search in the vector library storing compliant semantic vectors. By comparing the similarity between the semantic vector to be reviewed and the retrieved compliant semantic vector, several compliant semantic vectors with the highest similarity to the semantic vector to be reviewed are selected from each compliant semantic vector to obtain several target compliant semantic vectors, thereby obtaining the most relevant compliant text.

[0088] The vector similarity between the semantic vector to be reviewed and the compliant semantic vector can be described by angular distance, cosine distance, Euclidean distance or Manhattan distance.

[0089] In this embodiment, cosine distance is used to represent vector similarity, and the specific formula is:

[0090] ,

[0091] in, is the vector similarity between the semantic vector to be reviewed and the compliant semantic vector, is the modulus of the semantic vector to be examined, is the modulus of the conforming semantic vector.

[0092] Step S124: Based on the attention mechanism, a compliance comparison is performed on the semantic vector to be reviewed and the target compliant semantic vector, and a compliance review result is generated according to the compliance comparison result.

[0093] In step S124, based on the attention mechanism, semantic matching and comparison are performed on the key text semantic vector and each target compliance semantic vector, and the angular distance between each key text semantic vector and each target compliance semantic vector is calculated. The difference between the calculated vector distance information and the preset vector distance threshold is used as the probability that the object to be reviewed conforms to the compliant text corresponding to the target compliance semantic vector. After traversing each key text semantic vector and each target compliance semantic vector, the compliance review result is obtained.

[0094] See also Figure 3 , Figure 3 is a flowchart of a specific method of step S124 provided in an embodiment of the present application. In some embodiments of the present application, step S124 includes but is not limited to steps S310 to S330. Figure 3 These three steps are introduced in detail.

[0095] Step S310, taking the target compliance semantic vector as the context, through semantic matching and context understanding, the key text semantic vector and the target compliance semantic vector are compared for compliance and the comparison result is converted into text form to obtain a compliance judgment result.

[0096] Before step S310, the user may use the key text (or original text) corresponding to the key text semantic vector, the compliant text corresponding to the target compliance semantic vector and the guide text input by the user to form a first input sequence. After receiving the first input sequence, the compliance review model encodes the elements in the first input sequence into a vector representation. Alternatively, the compliance review model may obtain the corresponding key text semantic vector and target compliance semantic vector by itself.

[0097] For example, the first input sequence is [c; r1, r2, ... r k ;g], c is the original text, r1 to r k is the compliance text, and g is the introductory statement (for example, analyzing whether the case violates the above compliance regulations).

[0098] In step S310, the target compliance semantic vector is used as the context, and the vector distance information between each key text semantic vector and each target compliance semantic vector is calculated through semantic matching and context understanding. The difference between the vector distance information corresponding to the minimum vector distance and the preset vector distance threshold is selected as the probability that the object to be reviewed complies with the compliance text corresponding to the target compliance semantic vector. After traversing each target compliance semantic vector, the probability of the compliance text corresponding to each key text semantic vector is used to determine whether the object to be reviewed complies with the compliance specification clause represented by the corresponding compliance text, thereby obtaining a compliance judgment result. Among them, the compliance judgment result is divided into two states: "compliant" and "non-compliant".

[0099] Step S320: Generate corresponding compliance analysis results based on the compliance judgment results.

[0100] In step S320, for the compliance determination result with a status of "non-compliant", the corresponding cause analysis text is selected from the preset cause entry repository according to the compliance specification clause represented by the compliance text corresponding to the compliance determination result, thereby generating a corresponding compliance analysis result.

[0101] Step S330, import the compliance judgment results and compliance analysis results into a preset text template to obtain the compliance review results.

[0102] In step S330, text labels are pre-assigned to each associated position of the text template. When generating the compliance review result, the compliance judgment result and the compliance analysis result are first converted into text form, and then the text labels corresponding to each associated position where the compliance judgment result or the compliance analysis result is not inserted are obtained. The semantic association between the compliance judgment result and the text label or the compliance analysis result and the text label is identified through a semantic recognition algorithm to obtain a recognition result. According to the matching relationship contained in the recognition result, the corresponding compliance judgment result or compliance analysis result is inserted into the corresponding associated position to obtain a compliance review result.

[0103] Step S130: Generate a corresponding compliance rectification plan based on the compliance review results.

[0104] In step S130, the compliance review results are identified. When there is text that represents that the object to be reviewed does not comply with the corresponding compliance text, the compliance text, the corresponding key text and the corresponding cause analysis text of the corresponding compliance text are combined, and the compliance review model is used to perform a separate correlation analysis on the corresponding compliance text, the corresponding key text and the corresponding cause analysis text to obtain the correlation analysis results. According to the preset mapping relationship, the rectification tasks corresponding to the correlation analysis results are selected from the preset task set, and the rectification tasks and key texts are fitted to generate the corresponding compliance rectification plan.

[0105] In a specific embodiment, step S130 specifically includes the following steps: processing the compliance review results to obtain key text, target compliance text, compliance judgment results and compliance analysis results in the compliance review results; using the target compliance text, compliance judgment results and compliance analysis results as context, generating a rectification text containing rectification information, fitting the key text and the rectification text, and obtaining a compliance rectification plan.

[0106] Specifically, after processing the compliance review results, the second input sequence may be composed of key text, target compliance text, compliance judgment results, compliance analysis results and guide text input by the user. After receiving the second input sequence, the compliance review model encodes the elements in the second input sequence into a vector representation. Alternatively, the compliance review model may obtain the corresponding key text semantic vector, target compliance semantic vector and the vector representation corresponding to the compliance judgment results and the compliance analysis results by itself, and then generate the corresponding rectification text with the target compliance text, compliance judgment results and compliance analysis results as the context, and apply the rectification text to the key text to generate a compliance rectification plan that meets the needs of the object to be reviewed.

[0107] Exemplarily, the second input sequence is [d; r1, p1, a1; r2, p2, a2; ...; r k ,p k ,a k ;g], d is the key text, r1 to r k For compliance text, p1 to p k is the compliance judgment result, a1 to a k is the compliance analysis result, and g is the guide (for example, based on the case information, a rectification plan is provided for each non-compliant compliance judgment result. The rectification plan includes a task list. The task list includes a description of the problem, rectification goals, specific measures, and the tools and parameter settings involved).

[0108] Step S140, calling the API interface to execute the rectification tasks in the compliance rectification plan.

[0109] In step S140, according to the generated compliance rectification plan, by calling the API interfaces of various management systems within the enterprise, the terminals corresponding to the relevant departments and management tools are linked to automatically execute the rectification tasks in the compliance rectification plan.

[0110] See also Figure 4 , Figure 4 is a flowchart of a specific method of step S140 provided in an embodiment of the present application. In some embodiments of the present application, step S140 includes but is not limited to steps S410 to S430. Figure 4 These three steps are introduced in detail.

[0111] Step S410: Process the compliance rectification plan to obtain the rectification text in the compliance rectification plan.

[0112] Step S420: Generate corresponding rectification tasks according to the rectification text.

[0113] Step S430, calling the API interface to link the corresponding terminal and perform the rectification task with the terminal.

[0114] In step S410 to step S430, the corresponding rectification text is identified from the generated compliance rectification plan through text recognition technology, and rectification tasks are generated based on the identified rectification text, the corresponding problem description, rectification goals, specific measures, tools involved and parameter settings. The API interfaces of various management systems within the enterprise are called, and the terminals corresponding to the relevant departments and management tools are linked to execute the generated rectification tasks. For example, the task information of the rectification task can be sent to the corresponding terminal, and the terminal that receives the task information executes the rectification task. Alternatively, an execution instruction for executing the rectification task can be sent to the corresponding terminal, and the terminal that receives the execution instruction can be controlled to execute the rectification task.

[0115] See also Figure 5 The embodiment of the present application also provides a corporate compliance review device, which can implement the above-mentioned corporate compliance review method, and the device includes:

[0116] The first module 510 is used to obtain the original text containing the information of the object to be reviewed;

[0117] The second module 520 is used to input the original text into a preset compliance review model to obtain a compliance review result; the compliance review model is obtained by training a language model based on a Transformer architecture using sample original text, sample compliance text and corresponding sample compliance review results;

[0118] The third module 530 is used to generate a corresponding compliance rectification plan according to the compliance review results;

[0119] The fourth module 540 is used to call the API interface to execute the rectification task in the compliance rectification plan;

[0120] Input the original text into the preset compliance review model to obtain the compliance review results, including:

[0121] Processing the original text to obtain key text in the original text; the key text is text describing information related to the object to be reviewed and the compliance review;

[0122] Encode the key texts to obtain key text semantic vectors and semantic vectors to be reviewed obtained by fitting the semantic vectors of each key text;

[0123] A similarity comparison is performed between the semantic vector to be reviewed and the preset compliant semantic vector to obtain a number of target compliant semantic vectors; the compliant semantic vector is obtained by encoding the preset compliant text, and the compliant semantic vector is the compliant semantic vector with the highest similarity to the semantic vector to be reviewed;

[0124] Based on the attention mechanism, a compliance comparison is performed on the key text semantic vector and the target compliance semantic vector, and a compliance review result is generated based on the compliance comparison result.

[0125] The specific implementation of the enterprise compliance review device is basically the same as the specific implementation of the above-mentioned enterprise compliance review method, and will not be repeated here.

[0126] Figure 6 It is a block diagram of an electronic device according to an exemplary embodiment.

[0127] Refer to the following Figure 6 The electronic device 600 according to this embodiment of the present disclosure is described. Figure 6 The electronic device 600 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0128] like Figure 6 As shown, the electronic device 600 is in the form of a general computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different system components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0129] The storage unit stores program code, which can be executed by the processing unit 610, so that the processing unit 610 executes the steps described in the above-mentioned corporate compliance review method section of this specification according to various exemplary embodiments of the present disclosure.

[0130] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 6201 and / or a cache storage unit 6202 , and may further include a read-only storage unit (ROM) 6203 .

[0131] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0132] Bus 630 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0133] The electronic device 600 may also communicate with one or more external devices 600' (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 650. Furthermore, the electronic device 600 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 660. The network adapter 660 may communicate with other modules of the electronic device 600 via a bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0134] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned enterprise compliance review method is implemented.

[0135] The enterprise compliance review method, device, equipment and storage medium provided in the embodiment of the present application use the compliance review model obtained by training the language model based on the Transformer architecture to process the original text containing the information of the object to be reviewed, obtain the compliance review result, generate the corresponding compliance rectification plan according to the compliance review result, and execute the rectification task in the compliance rectification plan by calling the API interface, wherein the compliance review model selects the corresponding target compliance semantic vector based on the key text in the original text, and performs compliance comparison between the key text semantic vector and the target compliance semantic vector based on the attention mechanism, so as to generate the compliance review result according to the compliance comparison result. Since the compliance review model is used to perform compliance review on the original text, the compliance text of the compliance review clause involved in the object to be reviewed is determined based on the original text, and then the key text in the original text is compared with the compliance text based on the attention mechanism, the review efficiency and review accuracy of the enterprise compliance management can be improved. At the same time, the corresponding compliance rectification plan is generated according to the compliance review result and the rectification task in the compliance rectification plan is automatically executed, so as to realize the automated execution of compliance rectification and improve the rectification efficiency of the enterprise compliance management.

[0136] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, or a network device, etc.) to execute the above method according to the implementation of the present disclosure.

[0137] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0138] Computer readable storage media may include data signals propagated in baseband or as part of a carrier wave, wherein readable program codes are carried. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program codes contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination thereof.

[0139] Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0140] Those skilled in the art will appreciate that the above modules can be distributed in the device according to the description of the embodiment, or can be changed accordingly and only used in one or more devices different from the embodiment. The modules of the above embodiments can be combined into one module, or further divided into multiple sub-modules.

[0141] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.

[0142] The exemplary embodiments of the present disclosure are specifically shown and described above. It should be understood that the present disclosure is not limited to the detailed structures, configurations or implementations described herein; on the contrary, the present disclosure is intended to cover various modifications and equivalent configurations included in the spirit and scope of the appended claims.

Claims

1. A corporate compliance review method, characterized in that: include: Obtaining the original text containing the information of the subject to be reviewed; Inputting the original text into a preset compliance review model to obtain a compliance review result; The compliance review model is obtained by training a language model based on a Transformer architecture using sample original texts, sample compliance texts and corresponding sample compliance review results; Generate corresponding compliance rectification plans based on the compliance review results; Call the API interface to execute the rectification tasks in the compliance rectification plan; The inputting the original text into a preset compliance review model to obtain a compliance review result includes: Processing the original text to obtain key text in the original text; the key text is text describing information related to the object to be reviewed and the compliance review; Encode the key text to obtain a key text semantic vector and a semantic vector to be examined obtained by fitting the key text semantic vectors; Comparing the similarity between the semantic vector to be reviewed and the preset compliant semantic vector to obtain a plurality of target compliant semantic vectors; the compliant semantic vector is obtained by encoding the preset compliant text, and the compliant semantic vector is the compliant semantic vector with the highest similarity to the semantic vector to be reviewed; Based on the attention mechanism, a compliance comparison is performed on the key text semantic vector and the target compliance semantic vector, and the compliance review result is generated according to the compliance comparison result; According to the compliance review results, a corresponding compliance rectification plan is generated, including: Processing the compliance review results to obtain key text, target compliance text, compliance judgment results and compliance analysis results in the compliance review results; Generate a rectification text containing rectification information by taking the target compliance text, the compliance judgment result and the compliance analysis result as context, fit the key text and the rectification text, and obtain the compliance rectification plan; The calling of the API interface to execute the rectification tasks in the compliance rectification plan includes: Processing the compliance rectification plan to obtain the rectification text in the compliance rectification plan; Generate corresponding rectification tasks according to the rectification text; The API interface is called to link the corresponding terminal and execute the rectification task with the terminal.

2. The enterprise compliance review method according to claim 1, characterized in that: Before inputting the original text into a preset compliance review model to obtain a compliance review result, the method further includes: obtaining the compliance text; Performing word segmentation processing on the compliant text to obtain a compliant word text; The compliant word text is input into a preset encoding model to encode and normalize the key text to obtain a compliant semantic vector.

3. The enterprise compliance review method according to claim 1, characterized in that: The step of encoding the key text to obtain the key text semantic vector and the semantic vector to be examined obtained by fitting the key text semantic vectors includes: The key text is input into a preset encoding model to encode the key text to obtain the key text semantic vector, and a vector average operation or a vector weighted average operation is performed on the key text semantic vector to obtain the semantic vector to be reviewed.

4. The enterprise compliance review method according to claim 1, characterized in that: The similarity comparison between the semantic vector to be examined and the preset compliant semantic vector is performed to obtain a plurality of target compliant semantic vectors, including: Calculating the vector similarity between the semantic vector to be reviewed and the compliant semantic vector; According to the vector similarity, a preset number of the target compliant semantic vectors are obtained.

5. The enterprise compliance review method according to claim 1, characterized in that: The compliance comparison between the key text semantic vector and the target compliance semantic vector is performed based on the attention mechanism, and the compliance review result is generated according to the compliance comparison result, including: Taking the target compliance semantic vector as the context, through semantic matching and context understanding, the key text semantic vector and the target compliance semantic vector are compared for compliance and the comparison result is converted into text form to obtain a compliance judgment result; Generate corresponding compliance analysis results according to the compliance judgment results; Import the compliance judgment result and the compliance analysis result into a preset text template to obtain the compliance review result.

6. An enterprise compliance review device, characterized in that: include: The first module is used to obtain the original text containing the information of the object to be reviewed; The second module is used to input the original text into a preset compliance review model to obtain a compliance review result; the compliance review model is obtained by training a language model based on a Transformer architecture using sample original texts, sample compliance texts and corresponding sample compliance review results; The third module is used to generate a corresponding compliance rectification plan based on the compliance review results; The fourth module is used to call the API interface to execute the rectification tasks in the compliance rectification plan; The inputting the original text into a preset compliance review model to obtain a compliance review result includes: Processing the original text to obtain key text in the original text; the key text is text describing information related to the object to be reviewed and the compliance review; Encode the key text to obtain a key text semantic vector and a semantic vector to be examined obtained by fitting the key text semantic vectors; Comparing the similarity between the semantic vector to be reviewed and the preset compliant semantic vector to obtain a plurality of target compliant semantic vectors; the compliant semantic vector is obtained by encoding the preset compliant text, and the compliant semantic vector is the compliant semantic vector with the highest similarity to the semantic vector to be reviewed; Based on the attention mechanism, a compliance comparison is performed on the key text semantic vector and the target compliance semantic vector, and the compliance review result is generated according to the compliance comparison result; According to the compliance review results, a corresponding compliance rectification plan is generated, including: Processing the compliance review results to obtain key text, target compliance text, compliance judgment results and compliance analysis results in the compliance review results; Generate a rectification text containing rectification information by taking the target compliance text, the compliance judgment result and the compliance analysis result as context, fit the key text and the rectification text, and obtain the compliance rectification plan; The calling of the API interface to execute the rectification tasks in the compliance rectification plan includes: Processing the compliance rectification plan to obtain the rectification text in the compliance rectification plan; Generate corresponding rectification tasks according to the rectification text; The API interface is called to link the corresponding terminal and execute the rectification task with the terminal.

7. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the enterprise compliance review method described in any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the enterprise compliance review method described in any one of claims 1 to 5 is implemented.

Images