Business archive data intelligent management method and device and electronic equipment

Abstract

The invention relates to the field of business archive management, and particularly discloses a business archive data intelligent management method and device and electronic equipment, and the method comprises the steps: firstly obtaining a paper business archive image and electronic archive text data collected by OCR scanning equipment, and then carrying out the feature extraction and correlation analysis of the two through a deep learning technology, and finally, judging whether errors exist in the system business archives or not through a classifier, thereby identifying possible systematic errors or deviations, further improving the accuracy of data, reducing the requirements of manual inspection, and enhancing the efficiency of archive management.

Description

Technical Field

[0001] The present application relates to the field of business archive management, and more specifically, to a method, device and electronic equipment for intelligent management of business archive data. Background Art

[0002] Business records management involves the systematic organization, storage, and processing of all business documents and records within an organization. Effective business records management not only ensures the security and integrity of information but also improves work efficiency, supports compliance requirements, and facilitates rapid retrieval and analysis. This process encompasses the creation, classification, archiving, maintenance, and destruction of records. Its goal is to improve transparency and efficiency in organizational operations by optimizing the flow and management of information.

[0003] In the tax field, intelligent management of business archive data is becoming increasingly important. Traditional business archive data usually relies on paper documents and basic data entry systems. This method is not only inefficient but also prone to data errors and loss.

[0004] Therefore, a method, device and electronic device for intelligent management of business archive data are desired. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a method, device and electronic device for intelligent management of business archive data. The method first obtains paper business archive images and electronic archive text data collected by an OCR scanning device, then uses deep learning technology to perform feature extraction and association analysis on the two, and finally uses a classifier to determine whether there are errors in the system business archives, thereby identifying possible systematic errors or deviations, thereby improving data accuracy, reducing the need for manual inspection, and enhancing the efficiency of archive management.

[0006] According to one aspect of the present application, a method for intelligent management of business archive data is provided, which includes: Obtain paper business file images and electronic file text data captured by OCR scanning equipment; Extracting a paper business file association feature vector and an electronic file semantic global feature vector from the paper business file image collected by the OCR scanning device and the electronic file text data; Based on the paper business file association feature vector and the electronic file semantic global feature vector, it is determined whether there is an error in the system business file.

[0007] According to another aspect of the present application, there is provided a business archive data intelligent management device, comprising: Business file data acquisition module, used to obtain paper business file images and electronic file text data collected by OCR scanning equipment; A business archive data extraction module is used to extract paper business archive association feature vectors and electronic archive semantic global feature vectors from the paper business archive image collected by the OCR scanning device and the electronic archive text data; The business file error judgment and classification module is used to judge whether there is an error in the system business file based on the paper business file association feature vector and the electronic file semantic global feature vector.

[0008] According to another aspect of the present application, an electronic device is provided, comprising: a processor; and a memory, wherein computer program instructions are stored in the memory, and when the computer program instructions are executed by the processor, the processor executes the business archive data intelligent management device as described above.

[0009] Compared with the existing technology, the present application provides a method, device and electronic device for intelligent management of business archive data, which first obtains paper business archive images and electronic archive text data collected by OCR scanning equipment, and then uses deep learning technology to perform feature extraction and correlation analysis on the two. Finally, a classifier is used to determine whether there are errors in the system business archives, thereby identifying possible systematic errors or deviations, thereby improving data accuracy, reducing the need for manual inspection, and enhancing the efficiency of archive management. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0011] Figure 1 Flowchart of a method for intelligent management of business archive data according to an embodiment of the present application.

[0012] Figure 2 This is a flowchart of extracting paper business archive association feature vectors and electronic archive semantic global feature vectors from the paper business archive image collected by the OCR scanning device and the electronic archive text data in the business archive data intelligent management method according to an embodiment of the present application.

[0013] Figure 3This is a flowchart of performing feature extraction on the paper business file image collected by the OCR scanning device to obtain the paper business file associated feature vector in the business file data intelligent management method according to an embodiment of the present application.

[0014] Figure 4 This is a flowchart of performing data preprocessing on the paper business archive image collected by the OCR scanning device to obtain a paper business archive image block embedding vector sequence in the business archive data intelligent management method according to an embodiment of the present application.

[0015] Figure 5 2 is a block diagram of a business archive data intelligent management device according to an embodiment of the present application.

[0016] Figure 6 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0018] Figure 1 FIG. 1 is a flow chart of a method for intelligent management of business archive data according to an embodiment of the present application. Figure 1 As shown, the business archive data intelligent management method according to the embodiment of the present application includes: S110, obtaining a paper business archive image and electronic archive text data collected by an OCR scanning device; S120, extracting a paper business archive association feature vector and an electronic archive semantic global feature vector from the paper business archive image collected by the OCR scanning device and the electronic archive text data; S130, judging whether there is an error in the system business archive based on the paper business archive association feature vector and the electronic archive semantic global feature vector.

[0019] In the above-mentioned intelligent business archive data management method, step S110 acquires paper business archive images and electronic archive text data captured by an OCR scanner. It should be understood that business archive management is the systematic organization, storage, and processing of all business documents and records of an enterprise or organization. Efficient business archive management not only ensures the security and integrity of information, but also improves work efficiency, ensures compliance, and facilitates rapid information retrieval and analysis. This management process covers multiple aspects, including archive creation, classification, archiving, maintenance, and destruction. Its core purpose is to improve the transparency and efficiency of organizational operations by optimizing information flow and management. Traditionally, business archive data relies primarily on paper documents and basic data entry systems. This approach is not only inefficient but also prone to data errors and loss. Therefore, in the technical solution of this application, by acquiring paper business archive images and electronic archive text data captured by an OCR scanner and combining them with deep learning technology, the accuracy of the data is improved by determining whether there are errors in the system's business archives and identifying potential systemic errors or deviations. This approach not only reduces reliance on manual inspection but also enhances the overall efficiency of archive management.

[0020] Specifically, the primary purpose of acquiring paper business archive images and electronic archive text data captured by OCR scanning equipment is to achieve comprehensive and efficient management of business archives. Paper archives typically exist in traditional document formats, such as contracts, reports, and invoices. Manually processing these paper archives is time-consuming and prone to errors. OCR (Optical Character Recognition) technology can convert the text information in paper archives into electronic data, facilitating further calculations and analysis. Meanwhile, electronic archive text data already exists in digital form, which can be directly processed by the system and analyzed in conjunction with other information sources. Combining paper archive images with electronic archive text data can comprehensively cover the archive content, effectively utilizing information in both paper and digital forms. This improves the management efficiency of business archives, enhances the intelligence level of information processing, helps the system detect and correct potential errors, and ensures the quality and accuracy of business archives.

[0021] In the aforementioned intelligent business archive data management method, step S120 extracts paper business archive association feature vectors and electronic archive semantic global feature vectors from the paper business archive image captured by the OCR scanning device and the electronic archive text data. This allows business archives to be analyzed from different perspectives and levels, identifying potential errors and inconsistencies, achieving more accurate data comparison and error detection, and thus improving the intelligence level and processing efficiency of the business archive management system.

[0022] Figure 2The flowchart of extracting paper business archive association feature vectors and electronic archive semantic global feature vectors from the paper business archive image collected by the OCR scanning device and the electronic archive text data in the business archive data intelligent management method according to the embodiment of the present application is as follows. Figure 2 As shown, in a specific embodiment of the present application, the step S120 includes: S121, performing feature extraction on the paper business file image collected by the OCR scanning device to obtain the paper business file associated feature vector; S122, performing feature extraction on the electronic file text data to obtain the electronic file semantic global feature vector.

[0023] It's understandable that paper business records are converted into image data through OCR technology, but these images themselves are still pixel-level representations, making direct analysis of these image data complex and inefficient. The feature extraction process abstracts key information in an image (such as text, graphics, and layout structure) into feature vectors. These vectors represent the main features of the image, such as text layout, font style, and graphics type, making it easier for computer systems to understand and process this data.

[0024] Furthermore, feature extraction of electronic archival text data aims to gain a deeper understanding and leverage the complex semantic information contained within this data. Although electronic archival text data is already in digital format, directly processing the raw text often faces information sparsity and difficulties in understanding its semantics. The feature extraction process, by applying natural language processing (NLP) techniques, transforms words, phrases, and their contextual information within the text into semantic feature vectors. These vectors not only capture the basic meaning of each word but also consider the relationships between words and the overall context of the text, providing a more comprehensive and in-depth understanding of the text's content.

[0025] Figure 3 The flowchart of the business archive data intelligent management method according to the embodiment of the present application is to extract the features of the paper business archive image collected by the OCR scanning device to obtain the paper business archive associated feature vector. Figure 3 As shown, in a specific embodiment of the present application, the step S121 includes: S1211, performing data preprocessing on the paper business archive image collected by the OCR scanning device to obtain a paper business archive image block embedding vector sequence; S1212, passing the paper business archive image block embedding vector sequence through a paper business archive converter model based on the ViT model to obtain the paper business archive associated feature vector.

[0026] It's understandable that information contained in paper business archive images often exists in complex formats, including text, tables, charts, and other elements. The data preprocessing stage involves operations such as image sharpening, denoising, segmentation, and standardization. These steps aim to enhance image quality, remove noise, and make text and structural elements more prominent, thereby improving the accuracy of subsequent feature extraction. This clear and standardized image data is better suited for processing by computer vision algorithms.

[0027] Furthermore, the ViT (Vision Transformer)-based model embeds the image blocks of paper business archives into a vector sequence. This approach leverages advanced computer vision technology to extract more meaningful features from image data and achieve more efficient and accurate information processing. After data preprocessing and embedding vector generation, the image blocks of paper business archives have been converted into structured, high-dimensional feature representations. However, these vectors remain localized and primitive, failing to fully capture the more complex structural and semantic information within the image. By feeding these image block embedding vector sequences into the Transformer architecture, the ViT-based model leverages the Transformer's powerful modeling capabilities to capture global relationships and contextual information between image blocks. The ViT model is particularly adept at processing global information in image data. Through its self-attention mechanism, it learns long-range dependencies between image blocks, thereby improving the accuracy and completeness of feature extraction. The ViT model's advantage lies in its ability to process both spatial and semantic information within image data. Traditional convolutional neural networks (CNNs) rely primarily on layer-by-layer extraction of local features when processing images. The ViT model, however, divides images into fixed-size blocks and processes these blocks directly as sequences. This approach better captures long-range structural and semantic relationships within images, offering significant advantages for processing complex paper archival images, particularly documents rich in information (such as tables, diagrams, and complex typography).

[0028] Figure 4 The flowchart of the method for intelligent management of business archive data according to the embodiment of the present application is to pre-process the paper business archive image collected by the OCR scanning device to obtain the paper business archive image block embedding vector sequence. Figure 4 As shown, in a specific embodiment of the present application, the step S1211 includes: S1211-1, performing paper business archive image segmentation on the paper business archive image collected by the OCR scanning device to obtain a paper business archive image block sequence; S1211-2, performing paper business archive embedding coding on the paper business archive image block sequence to obtain the paper business archive image block embedding vector sequence.

[0029] It's understandable that the primary purpose of image segmentation for paper business files is to improve information extraction accuracy. Information within an image may be distributed across different locations and exist in various forms. By segmenting the image into blocks, the system can more accurately locate and extract information within each area. For example, text blocks and table blocks can be processed separately to more accurately identify and parse their content, avoiding errors caused by information mixing. Image segmentation can help streamline data processing. Breaking down a large image into smaller blocks allows each block to be processed in parallel, which not only speeds up processing but also reduces computing resource requirements.

[0030] Furthermore, embedding coding is performed on the paper business archive image block sequence to obtain a sequence of embedding vectors for the paper business archive image blocks. This is done to convert the image data into an efficient and manageable mathematical representation, facilitating further data processing and analysis. After segmentation, each image block contains information about a portion of the document's content. To effectively process this information and utilize it for intelligent analysis, the visual information of each image block must be converted into a vectorized feature representation. This is the primary purpose of embedding coding. The key to embedding coding for paper business archive image blocks lies in converting visual information into a computer-processable format. The raw data of image blocks is two-dimensional pixel values, which is difficult to directly apply to complex machine learning models and algorithms. Through embedding coding, these image blocks are mapped into a high-dimensional feature space, converting the image's visual information into digital vectors. Each image block embedding vector contains important features of the image block, such as the font, size, and color of text or the structure and style of a chart. These features effectively reflect the content and structure of the image block.

[0031] In a specific embodiment of the present application, step S122 includes: passing the electronic archive text data through an electronic archive text semantic encoder to obtain multiple electronic archive text feature vectors; and feature encoding the multiple electronic archive text feature vectors to obtain the electronic archive semantic global feature vector.

[0032] It should be understood that the main function of the electronic archive text semantic encoder is to convert natural language text into high-dimensional feature vectors that can compactly represent the semantic content of the text. Semantic encoders typically utilize deep learning models, such as Transformer or BERT, which can learn the underlying semantic structure of the language by training on large-scale corpora. After each text data is processed by the encoder, one or more feature vectors are generated. These vectors not only contain the basic lexical information of the text, but also capture the deeper meaning of the context and language. Specifically, the electronic archive text data is segmented to obtain an electronic archive word sequence; the embedding layer of the electronic archive text semantic encoder is used to map each electronic archive word in the electronic archive word sequence to a word embedding vector to obtain a sequence of electronic archive word embedding vectors; and the transformer-based BERT model of the electronic archive text semantic encoder is used to perform global contextual semantic encoding on the sequence of electronic archive word embedding vectors to obtain multiple electronic archive text feature vectors.

[0033] Furthermore, feature encoding is performed on multiple electronic archive text feature vectors to obtain a global semantic feature vector for the electronic archive. This is done to fuse the local features of multiple texts into a comprehensive, high-dimensional global representation, thereby fully capturing the overall semantic information of the electronic archive. A single electronic archive text feature vector only represents a portion of the text's semantic information and may not fully express the content and structure of the entire archive. Each text feature vector extracts specific semantic features from a local perspective, but these features may only be local and fragmentary. By integrating multiple text feature vectors through feature encoding, these local features can be synthesized into a global feature vector that covers the entire electronic archive.

[0034] In a specific embodiment of the present application, feature encoding is performed on the multiple electronic archive text feature vectors to obtain the electronic archive semantic global feature vector, including: arranging the multiple electronic archive text feature vectors into electronic archive text association feature vectors; and passing the electronic archive text association feature vectors through an electronic archive text association feature encoder based on a multi-layer perceptron to obtain the electronic archive semantic global feature vector.

[0035] It should be understood that arranging text feature vectors into associated feature vectors helps preserve and reflect the semantic relationships between text features. In real electronic archives, different text blocks or document sections often have complex internal connections, such as similarities in subject matter, coherence in context, and interdependence of information. Arranging multiple text feature vectors into associated feature vectors effectively preserves this associated information, enabling the system to understand and process the relationships between texts.

[0036] Furthermore, the electronic archive text-related feature vectors are processed through a multi-layer perceptron-based electronic archive text-related feature encoder to obtain a global feature vector for the electronic archive semantics. This is done to further refine and integrate the complex semantic information of the text, forming a comprehensive, high-dimensional global feature representation. As a deep learning model, the multi-layer perceptron (MLP) possesses powerful nonlinear mapping and feature learning capabilities. Encoding the electronic archive text-related feature vectors through the MLP effectively captures and refines the complex semantic relationships within text data. The MLP consists of multiple layers of neural networks, with each layer learning more abstract features from the output of the previous layer. This enables the MLP to process and integrate information from multiple related feature vectors, generating a global feature vector in a high-dimensional space that comprehensively represents the text semantics. The MLP-based feature encoder process can better preserve the semantic relevance and contextual information of text features. While the electronic archive text-related feature vectors already structurally reflect the relationships between texts, these vectors may still be local and fragmentary. The MLP encoding process integrates this local feature information into a global feature vector through processing across multiple neural network layers. This global feature vector not only includes the individual features of each text, but also incorporates the semantic relationships and contextual information between them. The MLP training process optimizes the representation of the global feature vector, making it more generalizable. During training, the MLP continuously adjusts network weights using a backpropagation algorithm to minimize prediction error. In this way, the encoder learns how to combine different text feature vectors into a representative global feature, thereby improving the feature vector's adaptability to practical tasks. Specifically, each fully connected layer of the multi-layer perceptron-based electronic archive text-related feature encoder performs fully connected encoding on the input vector during the forward pass of the layer to extract high-dimensional implicit features of the feature values ​​at each position in the input vector. The input to the first layer of the multi-layer perceptron-based electronic archive text-related feature encoder is the electronic archive text-related feature vector. The input to the second to last layers of the multi-layer perceptron-based electronic archive text-related feature encoder is the output vector of the previous layer. The output of the last layer of the multi-layer perceptron-based electronic archive text-related feature encoder is the electronic archive semantic global feature vector.

[0037] In the above-mentioned intelligent business archive data management method, step S130 determines whether the system business archive contains errors based on the paper business archive association feature vector and the electronic archive semantic global feature vector. It should be understood that paper business archives are typically digitized to become electronic archives, and the information in the paper archives is converted into electronic data for storage and analysis. The paper business archive association feature vector is obtained by encoding this converted data and can reflect the structured information and local features of the paper archive. These feature vectors contain the specific content and organization of the paper archive. The electronic archive semantic global feature vector is obtained by comprehensively encoding the electronic archive using a deep learning model. It integrates all local features and semantic relationships of the text to form a high-dimensional global feature representation that can accurately reflect the overall content and structure of the archive. Through systematic comparison and analysis, the specific information of the paper archive and the comprehensive semantics of the electronic archive can be combined to form a multi-level, multi-angle error detection feature.

[0038] In a specific embodiment of the present application, the step S130 includes: fusing the paper business archive association feature vector and the electronic archive semantic global feature vector to obtain a system business archive error judgment classification feature vector; performing local feature adjustment based on backpropagation representation on the system business archive error judgment classification feature vector to obtain an optimized system business archive error judgment classification feature vector; passing the optimized system business archive error judgment classification feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether there is an error in the system business archive.

[0039] It should be understood that the system's business archive error classification feature vector generated by fusing these two feature vectors can more accurately reflect the actual state of the business archive and identify issues such as inconsistencies, incorrect entries, and missing information. This fusion approach also leverages machine learning and data mining techniques to optimize the accuracy of error judgment by training the classifier. The system can be trained based on the fused feature vector to identify and classify different types of errors, thereby providing more precise error judgments and remediation suggestions.

[0040] In particular, in the technical solution of the present application, the system business archive error judgment classification feature vector is obtained by feature processing and feature fusion of paper business archive images and electronic archive text data collected by the OCR scanning device. During the paper business archive image segmentation process, the image cutting may not be precise enough, resulting in inconsistent image block boundaries or noise. Similarly, after semantic encoding, electronic archive text data may be affected by text parsing errors, which introduce additional noise during the arrangement and encoding process. Ultimately, the disturbances in these local features will be amplified in the fused system business archive error judgment classification feature vector. Considering that the classifier works based on the input system business archive error judgment classification feature vector, if the system business archive error judgment classification feature vector is unstable due to disturbances during the data processing process, the classifier may encounter insufficient classification accuracy when faced with these features with noise and errors. Therefore, in the technical solution of the present application, the system business archive error judgment classification feature vector is subjected to local feature adjustment based on backpropagation representation to obtain an optimized system business archive error judgment classification feature vector.

[0041] Among them, performing local feature adjustment based on backpropagation representation on the system business file error judgment classification feature vector to obtain an optimized system business file error judgment classification feature vector includes: determining a first system business file error judgment weight matrix and a second system business file error judgment weight matrix of the classifier before and after each iterative update; performing matrix multiplication of the first system business file error judgment weight matrix and the second system business file error judgment weight matrix with the system business file error judgment classification feature vector respectively to obtain a first system business file error judgment classification modulation feature vector and a second system business file error judgment classification modulation feature vector; calculating the position difference between the first system business file error judgment classification modulation feature vector and the second system business file error judgment classification modulation feature vector to obtain a system business file error judgment shift information representation. vector; calculate the F norm of the position mean vector between the first system business file error judgment classification modulation feature vector and the second system business file error judgment classification modulation feature vector as the system business file error judgment shift compensation scaling factor; linearly scale the system business file error judgment shift information representation vector with the system business file error judgment shift compensation scaling factor to obtain a scaled system business file error judgment shift information representation vector, and input the scaled system business file error judgment shift information representation vector into a Sigmoid activation function to obtain a system business file error judgment backpropagation shift compensation representation vector; calculate the position point multiplication of the system business file error judgment backpropagation shift compensation representation vector and the system business file error judgment classification feature vector to obtain the optimized system business file error judgment classification feature vector.

[0042] The optimization steps are specifically expressed as follows: ; ;in, represents the classification feature vector of the system business file error judgment, Indicates the system business file error judgment shift compensation scaling factor, represents the error judgment weight matrix of the second system business file, represents the error judgment weight matrix of the first system business file, represents the activation function, Indicates point multiplication by position, represents matrix multiplication, represents the difference by position, It means adding by position. represents the Frobenius norm, Represents the optimized system business archive error judgment classification feature vector.

[0043] In the technical solution of this application, to enhance the classifier's robustness to local perturbations in the input feature vector, a local feature adjustment based on backpropagation representation is performed on the system service profile error classification feature vector. The key to this backpropagation-based local feature adjustment lies in capturing information about changes in the weight matrix over the training process and utilizing this information to generate a system service profile error shift information representation vector. This information is then used to generate a system service profile error backpropagation shift compensation representation vector through scaling and nonlinear transformation (sigmoid function). In this way, the local feature shift description of the system service profile error classification feature vector is supported by the distribution differences of different classification scenarios during the classifier's training process. This not only quantifies the feature vector shift caused by the weight matrix change, but also generates a compensation vector based on the shift information. This compensation vector weights each element in the feature vector according to the degree of perturbation, thereby reducing the impact of the perturbation on the classification result. In this way, the classifier's robustness to local disturbances in the input data can be significantly improved, the classifier's stability and generalization ability can be enhanced, and by reducing the impact of local disturbances on the classification results, the classifier can maintain a high classification accuracy when facing noisy or variant data.

[0044] Furthermore, the optimized system business archive error judgment classification feature vectors integrate key information from both paper and electronic archives, reflecting the characteristics and potential error patterns of business archives from different data sources. These feature vectors incorporate both local information and global semantics of the text, enabling them to capture subtle differences and potential issues within the data. However, the optimized system business archive error judgment classification feature vectors themselves are only a highly abstract representation of the data, and directly interpreting them for error judgment can be challenging. Therefore, using a classifier to process the optimized system business archive error judgment classification feature vectors can transform complex feature information into specific classification results, thereby simplifying the error detection process. Once the classifier is trained and applied to the optimized system business archive error judgment classification feature vectors, it can generate classification results based on the content of the optimized system business archive error judgment classification feature vectors. These results indicate whether the system business archives contain errors, such as data inconsistencies, missing information, or typos.

[0045] In summary, the embodiment of the present application first obtains paper business file images and electronic file text data collected by the OCR scanning device, then uses deep learning technology to perform feature extraction and association analysis on the two, and finally uses a classifier to determine whether there are errors in the system business files, thereby identifying possible systematic errors or deviations, thereby improving data accuracy, reducing the need for manual inspection, and enhancing the efficiency of file management.

[0046] Figure 5 FIG is a block diagram of a business archive data intelligent management device according to an embodiment of the present application. Figure 5 As shown, the business archive data intelligent management device 100 according to the embodiment of the present application includes a business archive data acquisition module 110, which is used to obtain paper business archive images and electronic archive text data collected by an OCR scanning device; a business archive data extraction module 120, which is used to extract paper business archive association feature vectors and electronic archive semantic global feature vectors from the paper business archive images collected by the OCR scanning device and the electronic archive text data; a business archive error judgment and classification module 130, which is used to judge whether there are errors in the system business archive based on the paper business archive association feature vectors and the electronic archive semantic global feature vectors.

[0047] As described above, the business archive data intelligent management device 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the business archive data intelligent management device 100 can be integrated into the terminal device as a software module and / or hardware module. For example, the business archive data intelligent management device 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the business archive data intelligent management device 100 can also be one of the many hardware modules of the terminal device.

[0048] Alternatively, in another example, the business archive data intelligent management device 100 and the terminal device can also be separate devices, and the business archive data intelligent management device 100 can be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0049] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned business archive data intelligent management device have been described in detail in the above reference. Figures 1 to 4 The description of the business archive data intelligent management method has been introduced in detail, and therefore, its repeated description will be omitted.

[0050] Below, reference Figure 6 To describe the electronic device according to the embodiment of the present application.

[0051] See also Figure 6 , Figure 6 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes: The processor 11 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application; the memory 12 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM).

[0052] The memory 12 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program codes are stored in the memory 12 and are called by the processor 11 to execute the business archive data intelligent management method of the embodiments of this application. Input / output interface 13, used for information input and output; Communication interface 14, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.); bus 15 , which transmits information between the various components of the device (e.g., processor 11 , memory 12 , input / output interface 13 , and communication interface 14 ); The processor 11 , the memory 12 , the input / output interface 13 and the communication interface 14 are connected to each other in communication within the device via a bus 15 .

[0053] An embodiment of the present application also provides a storage medium, which is a computer-readable storage medium. The storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned business archive data intelligent management method is implemented.

[0054] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0055] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0056] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0057] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0058] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0059] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable 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. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0060] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0061] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0062] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0063] In addition, the functional units in the various embodiments of the present application can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method of each embodiment of the present application. The aforementioned storage medium includes various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk. The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, but the scope of rights of the embodiments of the present application is not limited thereby. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the rights of the embodiments of the present application.

Claims

1. A business archive data intelligent management method, characterized in that: include: Obtain paper business archive images and electronic archive text data collected by OCR scanning equipment; Extracting paper business archive association feature vectors and electronic archive semantic global feature vectors from the paper business archive image collected by the OCR scanning device and the electronic archive text data; Based on the paper business archive associated feature vector and the electronic archive semantic global feature vector, it is determined whether there is an error in the system business archive.

2. The intelligent management method for business archive data according to claim 1, characterized in that: Extracting a paper business archive association feature vector and an electronic archive semantic global feature vector from the paper business archive image collected by the OCR scanning device and the electronic archive text data, including: Performing feature extraction on the paper business archive image collected by the OCR scanning device to obtain the paper business archive associated feature vector; Feature extraction is performed on the electronic archive text data to obtain the electronic archive semantic global feature vector.

3. The intelligent management method for business archive data according to claim 2, characterized in that: Extracting features from the paper business archive image collected by the OCR scanning device to obtain the paper business archive associated feature vector includes: Performing data preprocessing on the paper business archive image collected by the OCR scanning device to obtain a paper business archive image block embedding vector sequence; The paper business archive image block embedding vector sequence is passed through a paper business archive converter model based on the ViT model to obtain the paper business archive associated feature vector.

4. The intelligent management method for business archive data according to claim 3 is characterized in that: The paper business archive image collected by the OCR scanning device is subjected to data preprocessing to obtain a paper business archive image block embedding vector sequence, including: Performing paper business archive image segmentation on the paper business archive image collected by the OCR scanning device to obtain a paper business archive image block sequence; The paper business archive image block sequence is subjected to paper business archive embedding coding to obtain the paper business archive image block embedding vector sequence.

5. The business archive data intelligent management method according to claim 4, characterized in that: Extracting features from the electronic archive text data to obtain the electronic archive semantic global feature vector includes: Passing the electronic archive text data through an electronic archive text semantic encoder to obtain a plurality of electronic archive text feature vectors; Feature encoding is performed on the multiple electronic archive text feature vectors to obtain the electronic archive semantic global feature vector.

6. The intelligent management method for business archive data according to claim 5, characterized in that: Performing feature encoding on the plurality of electronic archive text feature vectors to obtain the electronic archive semantic global feature vector includes: Arranging the plurality of electronic archive text feature vectors into an electronic archive text associated feature vector; The electronic archive text-related feature vector is passed through an electronic archive text-related feature encoder based on a multi-layer perceptron to obtain the electronic archive semantic global feature vector.

7. The intelligent management method for business archive data according to claim 6, characterized in that: Based on the paper business archive association feature vector and the electronic archive semantic global feature vector, judging whether there is an error in the system business archive includes: The paper business archive association feature vector and the electronic archive semantic global feature vector are integrated to obtain a system business archive error judgment classification feature vector; Performing local feature adjustment based on back-propagation representation on the system business file error judgment classification feature vector to obtain an optimized system business file error judgment classification feature vector; The optimized system business file error judgment classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to judge whether there is an error in the system business file.

8. The intelligent management method for business archive data according to claim 7, characterized in that: The system business file error judgment classification feature vector is locally adjusted based on the back propagation representation to obtain an optimized system business file error judgment classification feature vector, including: Determine a first system business file error judgment weight matrix and a second system business file error judgment weight matrix of the classifier before and after each iterative update; Performing matrix multiplication of the first system business file error judgment weight matrix and the second system business file error judgment weight matrix with the system business file error judgment classification feature vector respectively to obtain a first system business file error judgment classification modulation feature vector and a second system business file error judgment classification modulation feature vector; Calculating the position difference between the first system service file error judgment classification modulation feature vector and the second system service file error judgment classification modulation feature vector to obtain a system service file error judgment shift information representation vector; Calculating the F-norm of the position mean vector between the first system service file error judgment classification modulation feature vector and the second system service file error judgment classification modulation feature vector as a system service file error judgment shift compensation scaling factor; Linearly scaling the system business file error judgment shift information representation vector by the system business file error judgment shift compensation scaling factor to obtain a scaled system business file error judgment shift information representation vector, and inputting the scaled system business file error judgment shift information representation vector into a Sigmoid activation function to obtain a system business file error judgment backpropagation shift compensation representation vector; The optimized system business file error judgment classification feature vector is obtained by calculating the point-by-point multiplication of the system business file error judgment back-propagation shift compensation representation vector and the system business file error judgment classification feature vector.

9. A business archive data intelligent management device, characterized in that: include: The business archive data acquisition module is used to acquire paper business archive images and electronic archive text data collected by the OCR scanning device; A business archive data extraction module, used to extract paper business archive associated feature vectors and electronic archive semantic global feature vectors from the paper business archive image collected by the OCR scanning device and the electronic archive text data; The business archive error judgment classification module is used to judge whether there is an error in the system business archive based on the paper business archive associated feature vector and the electronic archive semantic global feature vector.

10. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 8 when executing the computer program.

Images