Internal control information management method and system for administrative institutions based on AI model

Abstract

A method and system for internal control information management of administrative institutions based on an AI model, relating to the technical field of intelligent information management, which obtains internal control information to be sorted; performs word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted; performs word embedding encoding on the sequence of internal control information words to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted; performs internal control information semantic encoding analysis on the sequence of internal control information word embedding vectors to be sorted to obtain semantic features of the internal control information to be sorted; based on the semantic features of the internal control information to be sorted, determines the category label to which the internal control information to be sorted belongs, which can realize automatic classification and labeling of internal control information of administrative institutions, thereby improving the efficiency and accuracy of internal control information management.

Description

Technical Field

[0001] The present application relates to the field of intelligent information management technology, and more specifically, to an internal control information management method and system for administrative institutions based on an AI model. Background Art

[0002] Internal control information management of administrative institutions refers to a series of systems, procedures and methods established by administrative institutions to ensure the authenticity, legality and effectiveness of their business activities. With the increase in the scale and complexity of the business of administrative institutions, the importance of internal control information management has become increasingly prominent. Internal control information management can not only improve the operating efficiency and quality of administrative institutions, but also prevent and control various risks and safeguard the legitimate rights and interests of administrative institutions.

[0003] However, the current internal control information management of administrative institutions usually relies on manual processing and manual classification. This method has many challenges, such as the diversity, dispersion, dynamism and non-standardization of internal control information, which will lead to inefficient collection, organization, classification and use of internal control information, and even misjudgment and underreporting.

[0004] Therefore, an internal control information management solution for administrative institutions based on AI models is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. This application provides an internal control information management method and system for administrative institutions based on an AI model, which obtains internal control information to be sorted; performs word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted; performs word embedding encoding on the sequence of internal control information words to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted; performs internal control information semantic encoding analysis on the sequence of internal control information word embedding vectors to be sorted to obtain semantic features of the internal control information to be sorted; based on the semantic features of the internal control information to be sorted, determines the category label to which the internal control information to be sorted belongs, which can realize the automatic classification and labeling of internal control information of administrative institutions, thereby improving the efficiency and accuracy of internal control information management.

[0006] First, an internal control information management method for administrative institutions based on AI models is provided, which includes:

[0007] Obtain internal control information to be sorted out;

[0008] Performing word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted;

[0009] Performing word embedding encoding on the sequence of internal control information words to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted;

[0010] Performing internal control information semantic coding analysis on the sequence of the internal control information word embedding vectors to be sorted to obtain the semantic features of the internal control information to be sorted, which includes:

[0011] Using a point convolution layer to positionally encode each internal control information word embedding vector to be sorted in the sequence of internal control information word embedding vectors to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted containing position information;

[0012] Passing the sequence of word embedding vectors of the internal control information to be sorted containing position information through a context encoder based on a transformer architecture to obtain a semantic encoding feature vector of the internal control information to be sorted as the semantic feature of the internal control information to be sorted;

[0013] Based on the semantic features of the internal control information to be sorted, a category label to which the internal control information to be sorted belongs is determined.

[0014] In the second aspect, an internal control information management system for administrative institutions based on an AI model is provided, which includes:

[0015] The module for obtaining the internal control information to be sorted is used to obtain the internal control information to be sorted;

[0016] A word segmentation processing module, used for performing word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted;

[0017] A word embedding encoding module, used for performing word embedding encoding on the sequence of the internal control information words to be sorted to obtain a sequence of embedding vectors of the internal control information words to be sorted;

[0018] An internal control information semantic coding analysis module is used to perform internal control information semantic coding analysis on the sequence of the internal control information word embedding vectors to be sorted to obtain semantic features of the internal control information to be sorted;

[0019] The module for determining the category label to which the internal control information to be sorted belongs is used to determine the category label to which the internal control information to be sorted belongs based on the semantic features of the internal control information to be sorted.

[0020] Compared with the prior art, the AI ​​model-based internal control information management method and system for administrative and public institutions provided in the present application obtains internal control information to be sorted; performs word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted; performs word embedding encoding on the sequence of internal control information words to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted; performs internal control information semantic encoding analysis on the sequence of internal control information word embedding vectors to be sorted to obtain semantic features of the internal control information to be sorted; based on the semantic features of the internal control information to be sorted, determines the category label to which the internal control information to be sorted belongs, which can realize the automatic classification and labeling of internal control information of administrative and public institutions, thereby improving the efficiency and accuracy of internal control information management. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 This is a flowchart of the internal control information management method for administrative institutions based on the AI ​​model according to an embodiment of the present application.

[0023] Figure 2 This is a schematic diagram of the architecture of the internal control information management method for administrative institutions based on the AI ​​model according to an embodiment of the present application.

[0024] Figure 3 This is a block diagram of an internal control information management system for administrative institutions based on an AI model according to an embodiment of the present application.

[0025] Figure 4 This is a scenario diagram of the internal control information management method for administrative institutions based on the AI ​​model according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] The following will describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are 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.

[0027] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present application have the same meaning as those commonly understood by those skilled in the art of the present application. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the scope of this application.

[0028] In the embodiments of the present application, it should be noted that, unless otherwise specified and limited, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection or a connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0029] It should be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.

[0030] Internal control information management refers to a series of systems, procedures and methods established by administrative institutions to ensure the authenticity, legality and effectiveness of business activities. These measures are aimed at regulating and supervising the acquisition, storage, processing and transmission of information, as well as ensuring the security, integrity and reliability of information. The importance of internal control information management lies in the fact that as the scale and business complexity of administrative institutions increase, their management activities are becoming increasingly complex, requiring more effective supervision and control.

[0031] The main objectives of internal control information management include:

[0032] 1. Improve operational efficiency and quality: By standardizing information management processes and ensuring the accuracy and reliability of information, administrative institutions can make decisions and manage more effectively, and improve work efficiency and service quality.

[0033] 2. Prevent and control various risks: Internal control information management helps to identify and assess potential risks, including financial risks, security risks and compliance risks, and reduce the probability of these risks by establishing corresponding control measures.

[0034] 3. Safeguarding legitimate rights and interests: Through internal control information management, administrative institutions can safeguard their legitimate rights and interests, including protecting important information assets, preventing information leakage and abuse, and complying with relevant laws, regulations and policy requirements.

[0035] Internal control information management usually involves: internal control system, establishing an internal control system, including clear division of responsibilities, authorization procedures, audit and supervision mechanisms, etc., to ensure the compliance and effectiveness of information management activities. Information system security, ensuring the security of information systems, including data backup and recovery, access control, network security and other measures to prevent information leakage and damage. Information disclosure and transmission, standardizing the process of information disclosure and transmission, ensuring the authenticity and integrity of information, and preventing the misrepresentation and tampering of information. Risk management, establishing a risk management system, including risk identification, assessment and control, and the formulation of emergency plans to deal with various risk events.

[0036] However, there are some defects in the internal control information management of administrative institutions. At present, internal control information management usually relies on manual processing and manual classification, which is inefficient and prone to omissions and errors. At the same time, with the increase in the amount of information, manual processing and classification have become more difficult and time-consuming. The internal control information of administrative institutions usually comes from various departments and business areas, involving various types and sources of information, and the content and form of information are often dynamically changing, which makes the collection, organization and classification of internal control information complicated and difficult. Due to the diversity of sources of internal control information, the degree of standardization and standardization of information varies, and there may be problems such as inconsistent information format, unclear content, and unclear standards, which affect the use and analysis of internal control information. Manual processing and manual classification lead to inefficient internal control information management, and it is impossible to process and analyze a large amount of information in a timely and accurate manner, thus affecting the implementation and effectiveness evaluation of internal control measures. Since manual processing and classification are prone to omissions and errors, it may lead to misjudgment and underreporting of internal control information, affecting the effectiveness and accuracy of internal control management.

[0037] Faced with these shortcomings, administrative institutions need to seek more efficient and accurate internal control information management methods, such as introducing information technology and data analysis tools to improve the efficiency and accuracy of internal control information management.

[0038] Figure 1 This is a flowchart of the internal control information management method for administrative institutions based on the AI ​​model according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of the internal control information management method of administrative institutions based on the AI ​​model according to the embodiment of the present application. Figure 1 and Figure 2 As shown, the internal control information management method for administrative institutions based on the AI ​​model includes: 110, obtaining the internal control information to be sorted; 120, performing word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted; 130, performing word embedding encoding on the sequence of internal control information words to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted; 140, performing internal control information semantic encoding analysis on the sequence of internal control information word embedding vectors to be sorted to obtain semantic features of the internal control information to be sorted; 150, determining the category label to which the internal control information to be sorted belongs based on the semantic features of the internal control information to be sorted.

[0039] In step 110, the internal control information to be sorted is obtained to ensure that the source of the collected internal control information is accurate and complete, and contains all relevant information that needs to be managed and analyzed. In this way, the original internal control information that needs to be processed is obtained, providing basic data for subsequent processing. In step 120, the internal control information to be sorted is segmented to obtain a sequence of internal control information words to be sorted, and appropriate segmentation tools and methods are selected to ensure the accuracy of the segmentation by taking into account the terminology in specific fields and professional terminology within administrative institutions. By converting the original text into a sequence of words, a basis is provided for subsequent semantic processing and analysis. In step 130, the sequence of internal control information words to be sorted is word-embedded to obtain a sequence of internal control information word embedding vectors to be sorted, and an appropriate word embedding model, such as Word2Vec, GloVe or BERT, is selected to obtain word embedding vectors. In this way, the words are converted into vector representations, capturing the semantic associations between words and providing a basis for subsequent semantic analysis. In step 140, the sequence of word embedding vectors of the internal control information to be sorted is subjected to internal control information semantic coding analysis to obtain the semantic features of the internal control information to be sorted. Clustering, topic model or deep learning method can be used to perform semantic analysis on word embedding vectors to obtain the semantic features of the internal control information. Through semantic coding analysis, the semantic similarities and differences between internal control information can be discovered, which helps to understand the meaning and internal associations of internal control information. In step 150, based on the semantic features of the internal control information to be sorted, the category label to which the internal control information to be sorted belongs is determined. A classification algorithm or clustering method can be used to assign category labels to the internal control information according to its semantic features. By determining the category label of the internal control information, it can be classified and managed, providing a basis for subsequent monitoring and decision-making.

[0040] In response to the above technical problems, the technical concept of this application is to collect the internal control information of administrative institutions and introduce artificial intelligence-based natural language processing technology at the back end to perform semantic analysis and processing on the internal control information, so as to determine the category label to which the internal control information belongs, such as financial information, operation information, compliance information, control activity information, information and communication information, internal supervision information, etc. In this way, the automatic classification and labeling of the internal control information of administrative institutions can be realized, thereby improving the efficiency and accuracy of internal control information management.

[0041] Through natural language processing technology, internal control information can be automatically classified and labeled, and the automatic management of internal control information can be realized, which will greatly reduce the burden of manual processing and improve the efficiency of internal control information management. Natural language processing technology can realize real-time processing and response to internal control information, making internal control information management more timely and flexible, and helping to quickly discover and deal with internal control risks. Natural language processing technology based on artificial intelligence can perform in-depth semantic analysis of internal control information, so as to more accurately determine the category label to which the internal control information belongs, avoid the bias of human subjective judgment, and improve the accuracy of internal control information management. Natural language processing technology can convert a large amount of internal control information into structured data, provide more data-supported decision-making basis for administrative units, help the leadership better understand the internal control situation, and formulate more scientific management strategies. Through semantic analysis, the potential correlation between different categories of internal control information can be discovered, which helps to fully understand and grasp the correlation of internal control information and provide more reference basis for internal control management decisions.

[0042] Specifically, in the technical solution of the present application, first, the internal control information to be sorted is obtained. Next, considering that in the internal control information management, the internal control information to be sorted is usually a continuous text, such as a report, a document or a description. That is to say, there is a large amount of semantic information in the internal control information to be sorted, and the smallest semantic understanding unit is a word. Therefore, in order to better understand and analyze the semantic information of the internal control information, in the technical solution of the present application, it is necessary to perform word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted. It is worth mentioning that word segmentation processing can provide a basis for subsequent steps such as text processing and semantic encoding. By converting the internal control information to be sorted into a sequence of internal control information words to be sorted, natural language processing and deep learning techniques can be better utilized to analyze and process this information, thereby achieving more accurate and efficient internal control information management.

[0043] Then, in order to use the professional terminology of the internal control information of administrative institutions to map each information word in the sequence of the internal control information words to be sorted into a high-dimensional feature space, so as to convert each information word into a vector representation to facilitate subsequent semantic analysis and understanding, in the technical solution of the present application, the sequence of the internal control information words to be sorted is further passed through a word encoder based on a word embedding layer, so as to linearly project each of the internal control information words to be sorted into a one-dimensional embedding vector through a learnable embedding matrix, so as to obtain a sequence of embedding vectors of the internal control information words to be sorted. In the internal control information management, each word can be converted into a word embedding vector by encoding the sequence of the internal control information words to be sorted through the word encoder based on the word embedding layer. These word embedding vectors can retain the semantic relationship between words, which helps to encode and understand the semantics of the internal control information in the subsequent process, so as to better represent and process the semantic features of the internal control information.

[0044] In a specific embodiment of the present application, word embedding encoding is performed on the sequence of internal control information words to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted, including: passing the sequence of internal control information words to be sorted through a word encoder based on a word embedding layer to obtain a sequence of internal control information word embedding vectors to be sorted.

[0045] Through word embedding technology, words will be mapped to a high-dimensional vector space, so that semantically similar words are close in the vector space, which can better capture the semantic association between words and help understand the meaning of words in internal control information and their mutual relationship in the semantic space. Word embedding vectors can better represent the diversity of words, that is, a word may have different meanings in different contexts. Word embedding vectors can capture this diversity and help to more comprehensively understand the meaning of words in internal control information. Word embedding vectors map high-dimensional word space to low-dimensional vector space, reducing the dimension of features, facilitating subsequent processing and analysis, and also reducing the complexity of calculations. Through word embedding vectors, words in internal control information can be converted into continuous vector representations. This continuous representation helps to improve the generalization ability of the model, making the model more accurate and robust when processing internal control information.

[0046] That is, by obtaining a sequence of word embedding vectors of the internal control information to be sorted through a word encoder based on the word embedding layer, it is possible to better capture the semantic associations between words, improve the diverse representation of internal control information, reduce the dimension and enhance the generalization ability of the model, thereby providing richer and more accurate semantic information for subsequent semantic analysis and classification.

[0047] In natural language processing, the position information of words is very important for understanding the semantics and structure of texts and sentences. The same word may have different meanings or functions in different positions. For example, in a sentence, the positions of the subject and the object are different, and their semantic roles are also different. Therefore, in order to better capture the position information of words, it is necessary to introduce position encoding into the word embedding vector. Based on this, in the technical solution of the present application, in order to integrate the position information into the word embedding vector to better represent the position relationship of the words in the sentence, it is necessary to further use a point convolution layer to position encode each of the internal control information word embedding vectors to be sorted in the sequence of the internal control information word embedding vectors to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted containing position information. In particular, here, point convolution is used for position encoding, and each element in the sequence of the internal control information word embedding vectors to be sorted can be assigned a new vector representation, which contains the position information of the element. This method can not only effectively perform position encoding, but also has low computational cost and is easy to parallelize. In this way, the position relationship of words in the sentence can be better preserved, providing richer information for subsequent context encoding.

[0048] In a specific embodiment of the present application, an internal control information semantic coding analysis is performed on the sequence of internal control information word embedding vectors to be sorted to obtain semantic features of the internal control information to be sorted, including: using a point convolution layer to perform position encoding on each internal control information word embedding vector to be sorted in the sequence of internal control information word embedding vectors to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted containing position information; passing the sequence of internal control information word embedding vectors to be sorted containing position information through a context encoder based on a converter architecture to obtain a semantic coding feature vector of the internal control information to be sorted as the semantic feature of the internal control information to be sorted.

[0049] By performing position encoding through the point convolution layer, the position information can be incorporated into the word embedding vector, so that the model can better understand the positional relationship of words in the sentence when processing the internal control information to be sorted, which helps to improve the understanding and analysis capabilities of the context. The context encoder based on the transformer architecture can perform a global contextual understanding of the internal control information to be sorted, so that each word embedding vector can better combine the contextual information of the surrounding words, thereby more accurately expressing the semantics of the internal control information to be sorted. The context encoder can extract the semantic features of the internal control information to be sorted and convert the word embedding vector into a higher-level semantic encoding feature vector, which helps to better capture the semantic information of the internal control information to be sorted and provide a richer and more accurate feature representation for subsequent classification and label determination.

[0050] Furthermore, considering that in the actual process of semantic understanding of the internal control information to be sorted, contextual information is very important for understanding the semantics of words and the semantics of the whole text, this is because the same word may have different meanings in different contexts. Therefore, in order to better analyze the semantic information of the internal control information to achieve accurate understanding and management of the internal control information, in the technical solution of the present application, the sequence of word embedding vectors of the internal control information to be sorted containing position information is further encoded through a context encoder based on a converter architecture to extract the global contextual semantic association feature information between the word embedding encoding features containing position information in the internal control information to be sorted, thereby obtaining the semantic encoding feature vector of the internal control information to be sorted.

[0051] Then, the semantic encoding feature vector of the internal control information to be sorted is passed through a classifier to obtain a classification result, and the classification result is used to represent the category label to which the internal control information to be sorted belongs. Specifically, in the technical solution of the present application, the label of the classifier is the category label to which the internal control information to be sorted belongs. In particular, the category label includes but is not limited to financial information, operation information, compliance information, control activity information, information and communication information, and internal supervision information.

[0052] In a specific embodiment of the present application, based on the semantic features of the internal control information to be sorted, the category label to which the internal control information to be sorted belongs is determined, including: passing the semantic encoding feature vector of the internal control information to be sorted through a classifier to obtain a classification result, and the classification result is used to represent the category label to which the internal control information to be sorted belongs.

[0053] Among them, the category labels include: financial information, operational information, compliance information, control activity information, information and communication information, and internal supervision information.

[0054] By using a classifier, the internal control information to be sorted can be automatically divided into corresponding category labels, thereby realizing automatic classification of internal control information and saving labor costs and time. The classification results can help organizations and managers more quickly understand the category labels to which the internal control information to be sorted belongs, which helps to make decisions quickly and take corresponding actions. Based on the classification results, data analysis and mining can be performed to achieve data-driven decision-making and management, which helps to discover potential associations and trends. Through the classifier, real-time processing and response to internal control information can be achieved, which helps to timely discover and solve problems and improve work efficiency and management level.

[0055] In one embodiment of the present application, the internal control information management method for administrative institutions based on the AI ​​model also includes a training step: for training the word encoder based on the word embedding layer, the point convolution layer, the context encoder based on the converter architecture, and the classifier. The training step includes: obtaining training internal control information to be sorted; performing word segmentation processing on the training internal control information to be sorted to obtain a sequence of training internal control information words to be sorted; performing word embedding encoding on the sequence of training internal control information words to be sorted to obtain a sequence of training internal control information word embedding vectors to be sorted; using the point convolution layer to positionally encode each training internal control information word embedding vector in the sequence of training internal control information word embedding vectors to be sorted to obtain a sequence of training internal control information word embedding vectors containing position information; passing the sequence of training internal control information word embedding vectors containing position information through the context encoder based on the converter architecture to obtain a semantic encoding feature vector of training internal control information to be sorted; performing clustering optimization on the semantic encoding feature vector of the training content information to be sorted to obtain an optimized semantic encoding feature vector of the training internal control information to be sorted; passing the optimized semantic encoding feature vector of the training internal control information to be sorted through the classifier to obtain a classification loss function value; and training the word encoder based on the word embedding layer, the point convolution layer, the context encoder based on the converter architecture, and the classifier based on the classification loss function value.

[0056] In the above technical scheme, the sequence of training word embedding vectors of internal control information to be sorted expresses the text encoding features of the internal control information to be sorted based on word segmentation semantics, and the point convolution layer is used to perform position encoding on each training word embedding vector of internal control information to be sorted in the sequence of training word embedding vectors of internal control information to be sorted to obtain a sequence of training word embedding vectors of internal control information to be sorted containing position information, and after the sequence of training word embedding vectors of internal control information to be sorted containing position information is passed through a context encoder based on a converter architecture to obtain the semantic encoding feature vector of the internal control information to be sorted, the global associated text semantic features based on the context of the training word segmentation encoding semantic features containing position information can be further extracted.

[0057] However, when the sequence of training content information word embedding vectors containing position information includes text encoding features and position encoding features based on word segmentation semantics, and the training content information semantic encoding feature vector integrates local-global correlation text semantic features, the training content information semantic encoding feature vector will also have the local feature distribution discreteness based on the specific feature distribution of each training content information word embedding vector containing position information, so that when the training content information semantic encoding feature vector is classified as a whole by a classifier, it will be difficult to converge to the probability of a predetermined class when class regression is performed by the classifier due to its local feature distribution discreteness, thereby affecting the training speed of the classifier and the accuracy of the classification result finally obtained.

[0058] Based on this, the present application performs clustering optimization on the semantic coding feature vector of the training content information to be sorted, that is, firstly clustering the various eigenvalues ​​of the semantic coding feature vector of the training content information to be sorted, for example, clustering based on the distance between eigenvalues, and then optimizing based on the in-class and out-of-class representations of the clustered features to obtain the optimized semantic coding feature vector of the training internal control information to be sorted.

[0059] In one embodiment of the present application, the following optimization formula is used to optimize the in-class and out-of-class representations of the clustered features to obtain the optimized training semantic encoding feature vector of the internal control information to be sorted; wherein the optimization formula is:

[0060]

[0061] in, are the eigenvalues ​​of the semantic encoding feature vector of the training content information to be sorted, is the number of feature sets corresponding to the semantic encoding feature vector of the training content information to be sorted, is the number of clustering features, represents the clustering feature set, It is each eigenvalue of the semantic coding feature vector of the internal control information to be sorted in the optimization training.

[0062] Specifically, by taking the in-class features and out-of-class features of the semantic coding feature vector of the training content information to be sorted as different instance roles to perform class instance description based on clustering ratio distribution, and introducing clustering response history based on in-class and out-of-class dynamic context, a coordinated global perspective is maintained for the in-class distribution and out-of-class distribution of the overall features of the semantic coding feature vector of the training content information to be sorted, so that the optimized feature clustering operation for the semantic coding feature vector of the training content information to be sorted can maintain a consistent response of the in-class and out-of-class features, so that the regression convergence path based on feature clustering remains consistent during the class regression process, and the convergence effect of the semantic coding feature vector of the training content information to be sorted towards the predetermined class probability is improved, so as to improve the training speed of the classifier and the accuracy of the classification results. In this way, the automatic classification and labeling of the internal control information of administrative institutions can be realized, thereby improving the efficiency and accuracy of internal control information management.

[0063] In summary, the internal control information management method and system of administrative institutions based on the AI ​​model according to the embodiment of the present application are explained, which collects the internal control information of administrative institutions and introduces the natural language processing technology based on artificial intelligence at the back end to perform semantic analysis and processing on the internal control information, so as to determine the category label to which the internal control information belongs, such as financial information, operation information, compliance information, control activity information, information and communication information, internal supervision information, etc. In this way, the automatic classification and labeling of the internal control information of administrative institutions can be realized, thereby improving the efficiency and accuracy of internal control information management.

[0064] In one embodiment of the present application, Figure 3 FIG. 1 is a block diagram of an internal control information management system for administrative institutions based on an AI model according to an embodiment of the present application. Figure 3 As shown, according to the embodiment of the present application, the internal control information management system 200 for administrative institutions based on the AI ​​model includes: an internal control information acquisition module 210 for acquiring the internal control information to be sorted; a word segmentation processing module 220 for performing word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted; a word embedding coding module 230 for performing word embedding coding on the sequence of internal control information words to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted; an internal control information semantic coding analysis module 240 for performing internal control information semantic coding analysis on the sequence of internal control information word embedding vectors to be sorted to obtain semantic features of the internal control information to be sorted; a category label determination module 250 for the internal control information to be sorted is used to determine the category label to which the internal control information to be sorted belongs based on the semantic features of the internal control information to be sorted.

[0065] In the administrative and public institution internal control information management system based on the AI ​​model, the word embedding encoding module is used to: pass the sequence of internal control information words to be sorted through a word encoder based on a word embedding layer to obtain a sequence of internal control information word embedding vectors to be sorted.

[0066] In the administrative and public institution internal control information management system based on the AI ​​model, the internal control information semantic coding analysis module includes: a position coding unit, which is used to use a point convolution layer to perform position coding on each internal control information word embedding vector to be sorted in the sequence of internal control information word embedding vectors to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted containing position information; a context coding unit, which is used to pass the sequence of internal control information word embedding vectors to be sorted containing position information through a context encoder based on a converter architecture to obtain a semantic coding feature vector of the internal control information to be sorted as the semantic feature of the internal control information to be sorted.

[0067] Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the above-mentioned administrative and public institution internal control information management system based on the AI ​​model have been referred to above. Figure 1 to Figure 2 The internal control information management method of administrative institutions based on AI model has been introduced in detail, and therefore, its repeated description will be omitted.

[0068] As described above, the administrative and public institutions internal control information management system 200 based on the AI ​​model according to the embodiment of the present application can be implemented in various terminal devices, such as servers for internal control information management of administrative and public institutions based on the AI ​​model. In one example, the administrative and public institutions internal control information management system 200 based on the AI ​​model according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the administrative and public institutions internal control information management system 200 based on the AI ​​model can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the administrative and public institutions internal control information management system 200 based on the AI ​​model can also be one of the many hardware modules of the terminal device.

[0069] Alternatively, in another example, the administrative and public institution internal control information management system 200 based on the AI ​​model and the terminal device may also be separate devices, and the administrative and public institution internal control information management system 200 based on the AI ​​model may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0070] Figure 4 Schematic diagram of the scenario of the internal control information management method of administrative institutions based on the AI ​​model according to the embodiment of the present application. Figure 4As shown, in this application scenario, first, the internal control information to be sorted is obtained (for example, Figure 4 Then, the acquired internal control information to be sorted is input into a server that is deployed with an AI model-based administrative unit internal control information management algorithm (for example, Figure 4 In S) shown in , the server is capable of processing the internal control information to be sorted based on the administrative and public institution internal control information management algorithm of the AI ​​model to determine the category label to which the internal control information to be sorted belongs.

[0071] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0072] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0073] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. The internal control information management method of administrative institutions based on AI model is characterized by: include: Obtain internal control information to be sorted out; Performing word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted; Performing word embedding encoding on the sequence of internal control information words to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted; Performing internal control information semantic coding analysis on the sequence of the internal control information word embedding vectors to be sorted to obtain the semantic features of the internal control information to be sorted, which includes: Using a point convolution layer to positionally encode each internal control information word embedding vector to be sorted in the sequence of internal control information word embedding vectors to be sorted to obtain a sequence of internal control information word embedding vectors to be sorted containing position information; Passing the sequence of word embedding vectors of the internal control information to be sorted containing position information through a context encoder based on a transformer architecture to obtain a semantic encoding feature vector of the internal control information to be sorted as the semantic feature of the internal control information to be sorted; Determining the category label to which the internal control information to be sorted belongs based on the semantic features of the internal control information to be sorted; The method further comprises a training step for training a word encoder based on a word embedding layer, the point convolution layer, the context encoder based on a transformer architecture, and a classifier; The training step comprises: Obtain internal control information to be sorted out during training; Performing word segmentation processing on the training internal control information to be sorted to obtain a sequence of training internal control information words to be sorted; Performing word embedding encoding on the sequence of training internal control information words to be sorted to obtain a sequence of training internal control information words to be sorted embedding vectors; Using the point convolution layer to positionally encode each training internal control information word embedding vector in the sequence of training internal control information word embedding vectors to be organized to obtain a sequence of training internal control information word embedding vectors containing position information; Passing the sequence of training to-be-organized internal control information word embedding vectors containing position information through the context encoder based on the transformer architecture to obtain a training to-be-organized internal control information semantic encoding feature vector; Performing cluster optimization on the semantic coding feature vector of the training internal control information to be sorted to obtain an optimized semantic coding feature vector of the training internal control information to be sorted; Passing the optimized training semantic coding feature vector of the internal control information to be sorted through the classifier to obtain a classification loss function value; Training the word encoder based on the word embedding layer, the point convolution layer, the context encoder based on the transformer architecture, and the classifier based on the classification loss function value; Clustering and optimizing the semantic coding feature vector of the training content information to be sorted, including: clustering each feature value of the semantic coding feature vector of the training content information to be sorted, and optimizing the in-class and out-of-class representations of the clustered features to obtain the optimized semantic coding feature vector of the training internal control information to be sorted; The following optimization formula is used to optimize the in-class and out-of-class representations of the clustered features to obtain the semantic encoding feature vector of the optimized training internal control information to be sorted; wherein the optimization formula is: ； in, are the eigenvalues ​​of the semantic encoding feature vector of the training content information to be sorted, is the number of feature sets corresponding to the semantic encoding feature vector of the training content information to be sorted, is the number of clustering features, represents the clustering feature set, It is each eigenvalue of the semantic coding feature vector of the internal control information to be sorted in the optimization training.

2. According to the AI ​​model-based internal control information management method for administrative institutions in claim 1, it is characterized in that: The sequence of the internal control information words to be sorted is subjected to word embedding encoding to obtain a sequence of the internal control information word embedding vectors to be sorted, including: passing the sequence of the internal control information words to be sorted through a word encoder based on a word embedding layer to obtain a sequence of the internal control information word embedding vectors to be sorted.

3. The internal control information management method for administrative institutions based on AI model according to claim 2 is characterized in that: Based on the semantic features of the internal control information to be sorted, determining the category label to which the internal control information to be sorted belongs includes: passing the semantic encoding feature vector of the internal control information to be sorted through a classifier to obtain a classification result, and the classification result is used to represent the category label to which the internal control information to be sorted belongs.

4. The internal control information management method for administrative institutions based on AI model according to claim 3 is characterized in that: The category labels include: financial information, operational information, compliance information, control activity information, information and communication information, and internal oversight information.

5. An administrative and public institution internal control information management system based on an AI model, used to execute the administrative and public institution internal control information management method based on an AI model as claimed in claim 1, characterized in that: include: The module for obtaining the internal control information to be sorted is used to obtain the internal control information to be sorted; A word segmentation processing module, used for performing word segmentation processing on the internal control information to be sorted to obtain a sequence of internal control information words to be sorted; A word embedding encoding module, used for performing word embedding encoding on the sequence of the internal control information words to be sorted to obtain a sequence of embedding vectors of the internal control information words to be sorted; The internal control information semantic coding analysis module is used to perform internal control information semantic coding analysis on the sequence of the internal control information word embedding vectors to be sorted to obtain the semantic features of the internal control information to be sorted.

6. The internal control information management system for administrative institutions based on AI model according to claim 5 is characterized in that: It also includes a category label determination module to which the internal control information to be sorted belongs, which is used to determine the category label to which the internal control information to be sorted belongs based on the semantic features of the internal control information to be sorted.

7. The internal control information management system for administrative institutions based on AI model according to claim 6 is characterized in that: The word embedding encoding module is used to: pass the sequence of the internal control information words to be sorted through a word encoder based on a word embedding layer to obtain a sequence of embedding vectors of the internal control information words to be sorted.

8. The internal control information management system for administrative institutions based on AI model according to claim 7 is characterized in that: The internal control information semantic coding analysis module includes: A position encoding unit, used for performing position encoding on each of the internal control information word embedding vectors to be sorted in the sequence of internal control information word embedding vectors to be sorted using a point convolution layer to obtain a sequence of internal control information word embedding vectors to be sorted containing position information; The context encoding unit is used to pass the sequence of the word embedding vectors of the internal control information to be sorted containing the position information through a context encoder based on a converter architecture to obtain a semantic encoding feature vector of the internal control information to be sorted as the semantic feature of the internal control information to be sorted.

Images