Electronic document classification method and device and electronic equipment

Abstract

The invention discloses an electronic document classification method and device and electronic equipment. Relates to the field of artificial intelligence. The method comprises the steps of obtaining a target electronic document; information included in the target electronic document is identified, N groups of information are obtained, N is an integer larger than or equal to 2, and the N groups of information correspond to different information types; processing the N groups of information by adopting N networks in a document classification model to obtain a target classification result of the target electronic document, the N networks being in one-to-one correspondence with the N groups of information, and the document classification model being obtained through machine learning based on the plurality of electronic documents and classification results corresponding to the plurality of electronic documents. Through the method and the device, the problem of low classification accuracy of the electronic documents caused by classification processing of the electronic documents based on a single network model in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and more specifically, to an electronic document classification method, device and electronic device. Background Art

[0002] When classifying electronic documents in the related art, the categories of electronic documents are mainly based on a single neural network model (such as a BP neural network model), which results in poor model applicability and, in turn, low accuracy in electronic document classification. In addition, the classification method for electronic documents in the related art is a general method, which has poor applicability to the classification and storage of electronic documents related to financial services (such as banking services).

[0003] Currently, no effective solution has been proposed to solve the problem that the classification processing of electronic documents based on a single network model in related technologies leads to low accuracy of electronic document classification and poor applicability for document classification processing and storage. Summary of the invention

[0004] The main purpose of the present application is to provide an electronic document classification method, device and electronic device to solve the problem in the related art that the classification processing of electronic documents is based on a single network model, resulting in low accuracy of electronic document classification.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, an electronic file classification method is provided. The method includes: obtaining a target electronic document; identifying the information included in the target electronic document to obtain N groups of information, wherein N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; using N networks in a document classification model to process the N groups of information to obtain a target classification result of the target electronic document, wherein the N networks correspond to the N groups of information one by one, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, and is obtained through machine learning.

[0006] Optionally, identifying the information included in the target electronic document to obtain N groups of information includes: converting the target electronic document into an electronic document image; identifying the information included in the electronic document image from N levels to obtain the N groups of information, wherein the N levels include at least a feature layer, an object layer, and a semantic concept layer, the feature layer is used to describe appearance features of the electronic document image, the object layer is used to indicate elements included in the target electronic document, and the semantic concept layer is used to indicate business-related features included in the target electronic document, and the N groups of information include at least feature layer information, object layer information, and semantic concept layer information.

[0007] Optionally, the adopting N networks in the document classification model to process the N groups of information to obtain a target classification result of the target electronic document includes: adopting the first network in the document classification model to process the feature layer information to obtain a first classification result; adopting the second network in the document classification model to process the object layer information to obtain a second classification result; adopting the third network in the document classification model to process the semantic concept layer information to obtain a third classification result; and obtaining the target classification result based on the first classification result, the second classification result and the third classification result.

[0008] Optionally, the first network includes a first convolutional neural network, the second network includes a second convolutional neural network and a recurrent neural network, and the third network includes a long short-term memory network.

[0009] Optionally, the second network in the document classification model is used to process the object layer information to obtain a second classification result, including: using the second convolutional neural network to process the image information in the object layer information to obtain a fourth classification result; using the recurrent neural network to process the text information in the object layer information to obtain a fifth classification result; and obtaining the second classification result based on the fourth classification result and the fifth classification result.

[0010] Optionally, obtaining the target classification result based on the first classification result, the second classification result and the third classification result includes: determining a first weight corresponding to the first classification result, a second weight corresponding to the second classification result, and a third weight corresponding to the third classification result; obtaining the target classification result based on the first classification result, the second classification result, the third classification result, the first weight, the second weight and the third weight.

[0011] Optionally, after the N networks in the document classification model are used to process the N groups of information to obtain a target classification result of the target electronic document, the method further includes: renaming the target electronic document based on the target classification result to obtain a renaming result; determining a target compression algorithm from multiple compression algorithms based on the target classification result, wherein the multiple compression algorithms correspond to different document types; and compressing the target electronic document using the target compression algorithm based on the renaming result to obtain a compressed document.

[0012] In order to achieve the above-mentioned purpose, according to another aspect of the present application, an electronic document classification device is provided. The device includes: an acquisition module for a target electronic document; an identification module for identifying information included in the target electronic document to obtain N groups of information, wherein N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; a processing module for processing the N groups of information using N networks in a document classification model to obtain a target classification result of the target electronic document, wherein the N networks correspond to the N groups of information one by one, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, and is obtained through machine learning.

[0013] Optionally, the recognition module includes: a first conversion submodule, used to convert the target electronic document into an electronic document image; a first recognition submodule, used to identify information included in the electronic document image from N levels to obtain the N groups of information, wherein the N levels at least include a feature layer, an object layer, and a semantic concept layer, the feature layer is used to describe the appearance characteristics of the electronic document image, the object layer is used to indicate elements included in the target electronic document, and the semantic concept layer is used to indicate business-related characteristics included in the target electronic document, and the N groups of information at least include feature layer information, object layer information, and semantic concept layer information.

[0014] Optionally, the processing module includes: a first processing sub-module, used to use the first network in the document classification model to process the feature layer information to obtain a first classification result; a second processing sub-module, used to use the second network in the document classification model to process the object layer information to obtain a second classification result; a third processing sub-module, used to use the third network in the document classification model to process the semantic concept layer information to obtain a third classification result; a first result sub-module, used to obtain the target classification result based on the first classification result, the second classification result and the third classification result.

[0015] Optionally, the second processing submodule includes: a fourth processing submodule, used to use the second convolutional neural network to process the image information in the object layer information to obtain a fourth classification result; a fifth processing submodule, used to use the recurrent neural network to process the text information in the object layer information to obtain a fifth classification result; and a second result submodule, used by the user to obtain the second classification result based on the fourth classification result and the fifth classification result.

[0016] Optionally, the first result submodule includes: a first determination submodule, used to determine a first weight corresponding to the first classification result, a second weight corresponding to the second classification result, and a third weight corresponding to the third classification result; a first target submodule, used to obtain the target classification result based on the first classification result, the second classification result, the third classification result, the first weight, the second weight and the third weight.

[0017] Optionally, the device also includes: a first naming submodule, used to rename the target electronic document based on the target classification result to obtain a renaming result; a second determination submodule, used to determine a target compression algorithm from multiple compression algorithms based on the target classification result, wherein the multiple compression algorithms correspond to different document types; and a first compression submodule, used to compress the target electronic document using the target compression algorithm based on the renaming result to obtain a compressed document.

[0018] In an embodiment of the present application, a target electronic document is obtained; information included in the target electronic document is identified to obtain N groups of information, wherein N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; N networks in a document classification model are used to process the N groups of information to obtain a target classification result of the target electronic document, wherein the N networks correspond one-to-one to the N groups of information, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, respectively, and is obtained through machine learning, thereby achieving the purpose of classifying and identifying the information included in the electronic document, and using different network layers in the document classification model to carry out targeted processing of different types of information, thereby achieving the technical effect of improving the accuracy of electronic document classification, and further solving the technical problem of low accuracy of electronic document classification due to the classification processing of electronic documents based on a single network model in the related art and incomplete consideration of factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 A hardware structure block diagram of a computer terminal for implementing an electronic document classification method is shown;

[0021] Figure 2 is a flow chart of an electronic document classification method according to an embodiment of the present application;

[0022] Figure 3 is a schematic diagram of an electronic document classification device provided according to an embodiment of the present application;

[0023] Figure 4 It is a structural block diagram of an electronic document classification electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims 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 sequence. 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 of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising 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.

[0026] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following explanations:

[0027] Neural Network (NN) refers to a computational model that mimics the structure and function of neurons in the human brain. It processes information through a large number of nodes (or "neurons") and edges (or "weights") connecting these nodes. These nodes are usually organized into layers, including input layers, hidden layers, and output layers. Neural networks are able to learn complex relationships and patterns between input data and optimize their performance by adjusting weights. They are widely used in machine learning, deep learning, image recognition, natural language processing and other fields.

[0028] Feature Layer (FL) refers to the layer used to extract basic image features (such as edges, corners, textures, etc.) in image processing and computer vision. These features are the basis for image analysis and understanding, and can be used for further image recognition and classification tasks.

[0029] The object layer (OL) refers to the level of identifying and understanding specific objects in the image. At this level, the system not only recognizes the basic features in the image, but also combines these features into higher-level, meaningful objects, such as people, cars, buildings, etc.

[0030] The Semantic Concept Layer (SCL) refers to the layer in image understanding that links the identified objects to broader semantic information (such as scenes, events, behaviors, etc.). The goal of this layer is to understand the relationship between objects in the image and their meaning in a specific context.

[0031] Convolutional Neural Network (CNN) refers to a deep learning algorithm that is mainly used to process data with a grid structure, such as images. It performs well in tasks such as image recognition, classification, and segmentation.

[0032] Recurrent Neural Network (RNN) is a neural network used to process sequence data. It can process sequences of any length. RNN introduces a loop structure into the network, which enables the network to consider previous input information when processing the current input. Therefore, it is particularly suitable for tasks such as natural language processing and time series analysis.

[0033] Long Short-Term Memory (LSTM) is a special type of RNN. It solves the gradient vanishing and gradient exploding problems of traditional RNN when processing long sequence data by introducing a gating mechanism (including input gate, forget gate and output gate). LSTM can learn long-term dependencies, so it performs better when processing long sequence data and is widely used in language models, speech recognition and other fields.

[0034] It should be noted that the collected information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data are in compliance with relevant laws, regulations and standards, necessary confidentiality measures are taken, and public order and good customs are not violated, and corresponding operation entrances are provided for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or institutions to provide users with corresponding operation entrances for users to choose to agree or refuse the results of automated decision-making; if the user chooses to refuse, the expert decision-making process will be entered.

[0035] Example 1

[0036] According to an embodiment of the present application, a method embodiment of electronic document classification is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0037] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computer terminal (or mobile device) for implementing an electronic document classification method. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0038] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0039] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the electronic document classification method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the above-mentioned electronic document classification method is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 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.

[0040] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0041] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0042] In the above operating environment, the present application provides an optional electronic document classification method. Figure 2 is a flow chart of an electronic document classification method according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0043] Step S202, obtaining a target electronic document;

[0044] Optionally, when a certain electronic document needs to be classified, the document is first obtained. This step can activate a series of subsequent processing processes such as recognition, classification, naming, compression, and storage. After obtaining the target electronic document, further information recognition and feature extraction can be performed on it to provide specific data objects for subsequent classification processing.

[0045] Optionally, in the process of acquiring the target electronic document, it is necessary to ensure the integrity of the document to avoid data loss or format changes during transmission or storage, which may affect the accuracy of subsequent processing. Therefore, some pre-processing, such as format conversion and resolution adjustment, can be performed while acquiring the document to meet the needs of subsequent processing steps and improve the smoothness and efficiency of the entire process.

[0046] Step S204, identifying the information included in the target electronic document to obtain N groups of information, where N is an integer greater than or equal to 2, and the N groups of information correspond to different information types;

[0047] Optionally, for each group of information types, corresponding technical means can be used, such as optical character recognition (OCR) technology to recognize text, image processing technology to recognize graphics and images, etc., to extract key features that are helpful for classification of target electronic documents, and provide high-quality input data for subsequent neural network processing. The steps before information recognition and classification prepare data for subsequent model processing, ensuring that each network model can receive information of the corresponding type, so that learning and classification can be carried out more effectively.

[0048] Optionally, the information included in the target electronic document can be identified by feature extraction, that is, by extracting features from the information included in the target electronic document to obtain N groups of information. Feature extraction is an important concept in the field of machine learning and computer vision, which can be used to extract key information from raw data that is helpful for subsequent tasks (such as classification, recognition, clustering, etc.). The purpose of feature extraction is to reduce the dimension of data, remove redundant information, and improve the efficiency and performance of the algorithm.

[0049] In an optional embodiment, information included in a target electronic document is identified to obtain N groups of information, including: converting the target electronic document into an electronic document image; identifying information included in the electronic document image from N levels to obtain N groups of information, wherein the N levels at least include a feature layer, an object layer, and a semantic concept layer, the feature layer is used to describe appearance features of the electronic document image, the object layer is used to indicate elements included in the target electronic document, and the semantic concept layer is used to indicate business-related features included in the target electronic document, and the N groups of information at least include feature layer information, object layer information, and semantic concept layer information.

[0050] Optionally, by converting electronic documents into images, image processing technology can be used to more comprehensively identify the document content. This conversion helps to extract the appearance features, element composition, and business-related conceptual features of the document, ensuring that the classification model can receive multi-dimensional information of the document.

[0051] Optionally, the information recognition process can be decomposed into N levels, corresponding to the appearance, constituent elements and business concepts of the document, which helps the classification model to perform hierarchical analysis of document information. This analysis method can understand the structure and content of the document in more detail and improve the accuracy and robustness of classification. Extracting information from the feature layer, object layer and semantic concept layer can integrate multi-dimensional features and provide richer and more comprehensive data support for subsequent classification processing. This multi-angle feature integration enables the classification model to make decisions based on more comprehensive information.

[0052] By converting electronic documents into images and identifying information from N levels, this embodiment can not only comprehensively extract the multi-dimensional features of the document, but also enhance the classification model's understanding of the document content and business background, thereby significantly improving the accuracy and efficiency of document classification and providing a solid foundation for subsequent document management processes.

[0053] It should be noted that the electronic document image may include, but is not limited to: personal financial account information, corporate financial account information, financial business contracts, transaction records, business reports, business application materials, and business notifications and confirmations, among which financial business contracts may include, but are not limited to, loan contracts, deposit contracts, remittance contracts, credit card contracts, etc. Transaction records include records of various transactions conducted by customers in financial institutions, such as transfer records, withdrawal records, payment records, etc. Business reports include statistical reports of various businesses of financial institutions, such as deposit reports, loan reports, remittance reports, etc. Customer information inquiry authorization is an authorization letter from a customer to authorize a financial institution to inquire about or use his or her personal information and transaction information. Business application materials are business application materials submitted by a customer to a bank, such as an account opening application, a loan application, a credit card application, etc. Business notifications and confirmations are business notifications sent by financial institutions to customers and customer confirmations of business, such as bills and statements.

[0054] These documents contain important information such as transaction information, contract information, and customer personal information between financial institutions and customers, and are an indispensable part of financial services. When using and processing these documents, it is necessary to comply with relevant regulations to ensure the confidentiality and security of customer information.

[0055] The feature layer includes information about the color, shape, layout, etc. of the relevant content included in the electronic document image. Color refers to the color of the electronic document image, such as the color of the text and the color of the background. Shape refers to the shape of the objects in the document image, such as the shape of the text and the shape of the graphics. Layout refers to the layout and arrangement of elements in the document, such as the arrangement of text and the layout of tables.

[0056] The object layer includes information on text, graphics, images, tables, etc. in the electronic document image. Among them, text refers to the text information in the document, including letters, numbers, punctuation marks, etc.; graphics refer to graphic elements in the document, such as lines, boxes, arrows, graphic symbols, etc.; images refer to images embedded or associated in the document, such as seals, signatures, photos, etc.; tables refer to table elements in the document, including the structure and content of the table.

[0057] The semantic concept layer includes information such as personal information, business information, contract terms, and business processes included in the electronic document image. Among them, personal information refers to the customer's personal information contained in the document; business information refers to the banking business information involved in the document, such as transaction amount, transaction type, business status, etc.; contract terms refer to important terms and agreements in business contracts such as loan contracts and deposit contracts; business processes refer to the processes and steps involved in the business process, such as application, review, approval, etc.

[0058] Step S206, using N networks in the document classification model to process N groups of information to obtain a target classification result of the target electronic document, wherein the N networks correspond one-to-one to the N groups of information, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, and is obtained through machine learning.

[0059] Alternatively, machine learning is a branch of artificial intelligence that enables computer systems to learn and improve from data without being explicitly programmed. Machine learning algorithms and models analyze large amounts of data to identify patterns and features to make predictions or decisions about new, unseen data. Document classification models are implementations of machine learning algorithms that can make predictions or classifications based on input data. By using N networks to process N groups of information, hierarchical and targeted processing of document information can be achieved. Each network focuses on processing a specific type of information, such as color, texture, text, graphics, semantic concepts, etc., which helps improve the accuracy and efficiency of information processing. After each network processes a group of information separately, its output results are integrated to make the final classification decision. This comprehensive decision-making mechanism can make full use of the advantages of each network to ensure the comprehensiveness and accuracy of the classification results. Document classification models are trained for machine learning based on a large number of electronic documents and their corresponding classification results. This data-driven model training method enables the model to learn the complex laws of document classification, so that it can make accurate classifications when faced with new documents.

[0060] In an optional embodiment, N groups of information are processed using N networks in a document classification model to obtain a target classification result of a target electronic document, including: using a first network in the document classification model to process feature layer information to obtain a first classification result; using a second network in the document classification model to process object layer information to obtain a second classification result; using a third network in the document classification model to process semantic concept layer information to obtain a third classification result; and obtaining a target classification result based on the first classification result, the second classification result and the third classification result.

[0061] Optionally, by decomposing document information into feature layer, object layer and semantic concept layer, and using different network models to process them respectively, it can ensure that each type of information is analyzed and processed in a targeted manner. This helps to capture the multi-level features of the document and improve the accuracy and comprehensiveness of the classification.

[0062] Optionally, different information layers can be processed by the first network, the second network, and the third network, respectively, and the obtained first classification results, second classification results, and third classification results can be regarded as different dimensions of classification decisions. Comprehensive analysis based on these multi-dimensional results can improve the rationality and accuracy of decision-making and avoid affecting the overall classification due to the deviation of a single information source. Each network model is trained based on information at the feature layer, object layer, or semantic concept layer, and can more effectively learn and understand the classification patterns of information at a specific level. This targeted training and optimization helps to improve the performance of the model in its respective field, thereby improving the efficiency and accuracy of the overall classification model.

[0063] This embodiment uses N networks to perform hierarchical processing and comprehensive decision-making on N groups of information, which can not only improve the accuracy and reliability of classification results, but also enhance the depth of understanding of document content, improve processing efficiency, and promote continuous optimization of the model, thereby achieving efficient and accurate document classification.

[0064] In an optional embodiment, the first network includes a first convolutional neural network, the second network includes a second convolutional neural network and a recurrent neural network, and the third network includes a long short-term memory network.

[0065] Optionally, the first network is mainly used to process feature layer information, such as appearance features such as color, texture, shape and layout. CNN is good at processing image data and can effectively extract local features from images. The second network includes a second convolutional neural network and a recurrent neural network, which are used to process graphics, images in the object layer and text and table information in the document respectively. CNN can recognize objects in images, while RNN is good at processing sequence data, such as text and tables, and can capture the sequence dependency of object layer information for more comprehensive information analysis. The third network uses LSTM, focusing on processing information at the semantic concept layer, such as personal information, business information, contract terms, etc. LSTM can process long sequence data and retain long-term dependencies, which is very effective for in-depth understanding of semantic concepts in documents, especially contextual information and semantic relationships in texts. This embodiment configures a specific type of neural network to process information at different levels, improve classification accuracy, enhance model generalization capabilities, improve processing efficiency, and adapt to complex business needs.

[0066] In an optional embodiment, the second network in the document classification model is used to process the object layer information to obtain a second classification result, including: using a second convolutional neural network to process the image information in the object layer information to obtain a fourth classification result; using a recurrent neural network to process the text information in the object layer information to obtain a fifth classification result; and obtaining a second classification result based on the fourth classification result and the fifth classification result.

[0067] Optionally, the fourth classification result is the image information processing result, and the fifth classification result is the text information processing result. Obtaining the second classification result based on the fourth classification result and the fifth classification result means fusing the information of the two dimensions of image and text, which can provide a more comprehensive understanding and decision-making basis for the classification model and improve the quality of classification decisions. By using CNN and RNN in the second network to process the image information and text information in the object layer respectively, the accuracy and efficiency of classification can be improved, and the robustness and adaptability of the model can be enhanced.

[0068] In an optional embodiment, a target classification result is obtained based on the first classification result, the second classification result and the third classification result, including: determining a first weight corresponding to the first classification result, a second weight corresponding to the second classification result, and a third weight corresponding to the third classification result; based on the first classification result, the second classification result, the third classification result, the first weight, the second weight and the third weight, the target classification result is obtained.

[0069] Optionally, the first classification result, the second classification result, and the third classification result can respectively reflect the classification of the electronic document in different dimensions, and by weighted fusion of these results, a more comprehensive and integrated decision-making basis can be constructed. This fusion takes into account the appearance, constituent elements, and business concepts of the document to ensure the comprehensiveness and accuracy of the classification results. By assigning different weights to each classification result (feature layer, object layer, semantic concept layer), this embodiment can quantify the contribution of different information types to the final classification decision. It helps to focus on more critical or more decisive information when processing complex documents. In addition, by introducing weights, the classification model can also be better adapted to the uncertainty of information. Even if a certain information type (such as the feature layer) performs poorly in individual documents, by adjusting its weight, the model can still rely on the classification results of other information types to make more accurate decisions.

[0070] In an optional embodiment, after using N networks in the document classification model to process N groups of information to obtain a target classification result of the target electronic document, the method also includes: renaming the target electronic document based on the target classification result to obtain a renaming result; determining a target compression algorithm from multiple compression algorithms based on the target classification result, wherein the multiple compression algorithms correspond to different document types; and compressing the target electronic document using the target compression algorithm based on the renaming result to obtain a compressed document.

[0071] Optionally, renaming electronic documents based on the target classification results can ensure that the naming logic of each document is consistent, which is convenient for storage and retrieval. The renaming process can include but is not limited to key document attributes such as creation date, document type, business classification, etc., which is conducive to the efficient operation of the document management system.

[0072] Optionally, the selection of compression algorithms can be based on the type of document. Document classification results directly affect the compression method. Different types of documents (such as text, images, tables, etc.) can be suitable for different compression algorithms. Selecting the most suitable algorithm can ensure a balance between compression efficiency and document quality. Document classification combined with renaming and compression can achieve efficient storage of documents. Renaming can ensure the identifiability and clear classification of documents, and compression can reduce the storage space occupied.

[0073] This embodiment can achieve efficient storage and retrieval of electronic documents through renaming and optimized compression processing after document classification, while ensuring the security and ease of reuse of documents, greatly improving the efficiency and standardization of document management, and has significant practical value for fields such as banking that need to process a large number of electronic documents.

[0074] It should be noted that the electronic document classification method provided in the embodiment of the present application obtains the target electronic document; identifies the information included in the target electronic document to obtain N groups of information, wherein N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; uses N networks in the document classification model to process the N groups of information to obtain the target classification result of the target electronic document, wherein the N networks correspond to the N groups of information one by one, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, obtained through machine learning, which solves the problem in the related art that the classification processing of electronic documents based on a single network model leads to low accuracy of electronic document classification and poor applicability to document classification processing and storage. Thus, the effect of improving the accuracy of electronic document classification is achieved.

[0075] Based on the above embodiments and optional embodiments, the present application proposes an optional implementation of an electronic document classification method, which includes:

[0076] Step S1, dividing the information contained in the electronic document image into three levels, including a feature level, an object level, and a semantic concept level;

[0077] Step S2, configuring a neural network model, configuring a different neural network for each feature layer, object layer and semantic concept layer, specifically including:

[0078] The neural network model includes a first CNN network, a second CNN network, an RNN network, and an LSTM network, wherein: the first CNN network takes a feature layer as input, and the feature layer contains features such as color, texture, shape, and layout. These features can be extracted and classified by CNN. CNN can extract local features from the original image and classify according to these features.

[0079] The second CNN network takes the graphics and images in the object layer as input, and the RNN network takes the text and tables in the object layer as input. CNN can be used to recognize objects in images, while RNN can be used to process text and table data and capture dependencies in sequences.

[0080] LSTM networks take as input the semantic concept layer, which contains high-level semantic concepts such as personal information, business information, contract terms, and business processes. These concepts can be processed by LSTM. These network structures can capture contextual information and semantic relationships in texts and classify them according to the text content.

[0081] Step S3, model training, to obtain a classification model, specifically includes:

[0082] A plurality of (e.g., 4000) electronic document images are classified according to the method of step S1, 3500 of them are randomly selected as a training set, and the remaining 1500 are selected as a verification set. The neural network model constructed in step S2 is trained and verified. When the model training is completed and the verification is passed, the trained classification model is obtained.

[0083] Step S4, classifying the electronic documents based on the classification model, specifically includes:

[0084] By classifying electronic documents based on the classification model, multiple electronic files can be classified to determine the category to which each electronic document belongs.

[0085] Step S5, arranging and naming the classified electronic documents, specifically includes:

[0086] Step S51, standardize the naming rules: To ensure the uniqueness and identifiability of the document, formulate a set of naming rules, such as using the creation date, number, keyword, document category, etc. of the electronic document as the naming rules, and name each electronic document.

[0087] Step S52, automatic renaming: based on the naming rules, use scripts or tools to automatically rename the classified electronic documents to ensure that the naming conventions are followed.

[0088] Step S6, electronic document performance compression, specifically includes:

[0089] Step S61, select compression algorithm: select a suitable compression algorithm according to the document type, keywords and other features of each electronic document. For example, pictures can be compressed using Joint Photographic Experts Group JPEG or Portable Network Graphics PNG, and text files can be compressed using Regional Improvement Program ZIP or compression program GZIP.

[0090] Step S62, compression processing: using corresponding tools or libraries to compress the document to reduce storage space occupation.

[0091] Step S7, electronic documents are matched and stored in a suitable proportion, specifically including:

[0092] Step S71, determining the storage ratio: determining the storage ratio of different types of electronic documents according to business requirements and storage capacity.

[0093] Step S72, automatic screening and storage: write a script or use a tool to automatically screen and store electronic documents that meet the ratio requirements.

[0094] Step S8, electronic document storage management, specifically includes:

[0095] Step S81, database management: using a database management system (such as MySQL, Oracle, etc.) to store and index electronic document information.

[0096] Step S82, metadata storage: storing the metadata of the electronic document (such as title, creation date, category, etc.) in a database for subsequent retrieval and management.

[0097] Step S9, retrieval and access of electronic documents, specifically includes:

[0098] Step S91, creating an index: To improve retrieval efficiency, a full-text index or a keyword index may be created for the electronic document content.

[0099] Step S92, authority control: controlling the access and operation authority to different types of electronic documents according to the user role and authority settings.

[0100] Step S10, monitoring and maintenance of electronic documents, specifically includes:

[0101] Step S1011, storage space monitoring: regularly monitor storage space usage to ensure that there is enough space to store new electronic documents.

[0102] Step S1012, backup and restore: For important electronic documents, special tags need to be added and important electronic documents need to be backed up regularly to ensure data security. When necessary, data can be quickly restored.

[0103] Through the above steps, the arrangement and naming, performance compression, matching of appropriate proportions for storage and subsequent retrieval, access, monitoring and maintenance of classified banking business-related electronic documents can be realized, which helps to improve the efficiency and standardization of document management and ensure the integrity and security of data.

[0104] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0105] Example 2

[0106] The embodiment of the present application also provides an electronic document classification device. It should be noted that the electronic document classification device of the embodiment of the present application can be used to execute the electronic document classification method provided by the embodiment of the present application. The electronic document classification device provided by the embodiment of the present application is introduced below.

[0107] According to an embodiment of the present application, a device for implementing the above-mentioned electronic document classification method is also provided. Figure 3As shown, the device includes: an acquisition module 300, for a target electronic document; an identification module 302, connected to the acquisition module 300, for identifying information included in the target electronic document to obtain N groups of information, wherein N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; a processing module 304, connected to the identification module 302, for processing the N groups of information using N networks in a document classification model to obtain a target classification result of the target electronic document, wherein the N networks correspond one-to-one to the N groups of information, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, and is obtained through machine learning.

[0108] The electronic document classification device provided in the embodiment of the present application is provided with an acquisition module 300 for a target electronic document; an identification module 302 connected to the acquisition module 300 for identifying the information included in the target electronic document to obtain N groups of information, wherein N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; a processing module 304 connected to the identification module 302 for processing the N groups of information using N networks in the document classification model to obtain a target classification result of the target electronic document, wherein the N networks correspond to the N groups of information one by one, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, obtained through machine learning, which solves the problem of low accuracy of electronic document classification caused by the classification processing of electronic documents based on a single network model in the related art. Thus, the effect of improving the accuracy of electronic document classification is achieved.

[0109] Optionally, in the electronic document classification device provided in the embodiment of the present application, the device includes: a first conversion submodule, used to convert the target electronic document into an electronic document image; a first identification submodule, used to identify information included in the electronic document image from N levels to obtain N groups of information, wherein the N levels include at least a feature layer, an object layer, and a semantic concept layer, the feature layer is used to describe the appearance characteristics of the electronic document image, the object layer is used to indicate the elements included in the target electronic document, and the semantic concept layer is used to indicate the business-related features included in the target electronic document, and the N groups of information include at least feature layer information, object layer information, and semantic concept layer information.

[0110] Optionally, in the electronic document classification device provided in the embodiment of the present application, the device also includes: a first processing submodule, used to use the first network in the document classification model to process the feature layer information to obtain a first classification result; a second processing submodule, used to use the second network in the document classification model to process the object layer information to obtain a second classification result; a third processing submodule, used to use the third network in the document classification model to process the semantic concept layer information to obtain a third classification result; a first result submodule, used to obtain a target classification result based on the first classification result, the second classification result and the third classification result.

[0111] Optionally, in the electronic document classification device provided in the embodiment of the present application, the device also includes: a fourth processing submodule, used to use a second convolutional neural network to process the image information in the object layer information to obtain a fourth classification result; a fifth processing submodule, used to use a recurrent neural network to process the text information in the object layer information to obtain a fifth classification result; a second result submodule, used for the user to obtain a second classification result based on the fourth classification result and the fifth classification result.

[0112] Optionally, in the electronic document classification device provided in the embodiment of the present application, the device also includes: a first determination submodule, used to determine a first weight corresponding to the first classification result, a second weight corresponding to the second classification result, and a third weight corresponding to the third classification result; a first target submodule, used to obtain a target classification result based on the first classification result, the second classification result, the third classification result, the first weight, the second weight and the third weight.

[0113] Optionally, in the electronic document classification device provided in the embodiment of the present application, the device also includes: a first naming submodule, used to rename the target electronic document based on the target classification result to obtain a renaming result; a second determination submodule, used to determine a target compression algorithm from multiple compression algorithms based on the target classification result, wherein the multiple compression algorithms correspond to different document types; a first compression submodule, used to compress the target electronic document using the target compression algorithm based on the renaming result to obtain a compressed document.

[0114] It should be noted that the acquisition module 300, the identification module 302, and the processing module 304 correspond to steps S202 to S206 in Example 1, and the examples and application scenarios implemented by the modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in a memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n), and the above-mentioned modules can also be run in the computer terminal 10 provided in Example 1 as part of the device.

[0115] Example 3

[0116] An embodiment of the present application may provide an electronic device, Figure 4 is a structural block diagram of an electronic device according to an embodiment of the present application. Figure 4 As shown, the electronic device may include: one or more ( Figure 4 (only one is shown) processor 1002, memory 1004, storage controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0117] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal 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.

[0118] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain the target electronic document; identify the information included in the target electronic document to obtain N groups of information, wherein N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; use N networks in the document classification model to process the N groups of information to obtain the target classification result of the target electronic document, wherein the N networks correspond one-to-one to the N groups of information, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, and is obtained through machine learning.

[0119] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: converting the target electronic document into an electronic document image; identifying the information included in the electronic document image from N levels to obtain N groups of information, wherein the N levels at least include a feature layer, an object layer, and a semantic concept layer, the feature layer is used to describe the appearance characteristics of the electronic document image, the object layer is used to indicate the elements included in the target electronic document, and the semantic concept layer is used to indicate the business-related characteristics included in the target electronic document, and the N groups of information at least include feature layer information, object layer information, and semantic concept layer information.

[0120] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: use the first network in the document classification model to process the feature layer information to obtain a first classification result; use the second network in the document classification model to process the object layer information to obtain a second classification result; use the third network in the document classification model to process the semantic concept layer information to obtain a third classification result; based on the first classification result, the second classification result and the third classification result, obtain the target classification result.

[0121] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: the first network includes a first convolutional neural network, the second network includes a second convolutional neural network and a recurrent neural network, and the third network includes a long short-term memory network.

[0122] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: use the second convolutional neural network to process the image information in the object layer information to obtain a fourth classification result; use the recurrent neural network to process the text information in the object layer information to obtain a fifth classification result; based on the fourth classification result and the fifth classification result, obtain a second classification result.

[0123] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: determine the first weight corresponding to the first classification result, the second weight corresponding to the second classification result, and the third weight corresponding to the third classification result; based on the first classification result, the second classification result, the third classification result, the first weight, the second weight and the third weight, obtain the target classification result.

[0124] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: based on the target classification result, rename the target electronic document to obtain a renaming result; based on the target classification result, determine the target compression algorithm from multiple compression algorithms, wherein the multiple compression algorithms correspond to different document types; based on the renaming result, use the target compression algorithm to compress the target electronic document to obtain a compressed document.

[0125] The embodiment of the present application provides a solution for electronic document classification. By acquiring a target electronic document; identifying the information included in the target electronic document to obtain N groups of information, wherein N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; using N networks in a document classification model to process the N groups of information to obtain a target classification result of the target electronic document, wherein the N networks correspond to the N groups of information one by one, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, respectively, obtained through machine learning, thereby achieving the purpose of improving the accuracy of electronic document classification, and further solving the technical problems of low accuracy of electronic document classification and poor applicability to document classification processing and storage caused by incomplete consideration of factors in a single network model.

[0126] It can be understood by those skilled in the art that Figure 4 The structure shown is for illustration only, and the electronic device may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, or other terminal devices. Figure 4 The structure of the electronic device is not limited. Figure 4 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Figure 4 Different configurations shown.

[0127] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0128] Example 4

[0129] The embodiment of the present application further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the electronic document classification method provided in the first embodiment.

[0130] Optionally, in this embodiment, the above storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0131] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing the steps of the electronic document classification method.

[0132] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0133] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0134] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, 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 units or modules, which can be electrical or other forms.

[0135] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0136] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0137] 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, in essence, 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, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0138] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. An electronic grading and classification method, characterized in that: include: Obtain target electronic documents; Identify the information included in the target electronic document to obtain N groups of information, where N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; The N groups of information are processed using N networks in a document classification model to obtain a target classification result of the target electronic document, wherein the N networks correspond one-to-one to the N groups of information, and the document classification model is obtained through machine learning based on multiple electronic documents and the classification results corresponding to the multiple electronic documents.

2. The method according to claim 1, characterized in that: The information included in the target electronic document is identified to obtain N groups of information, including: Converting the target electronic document into an electronic document image; The information included in the electronic document image is identified from N levels to obtain the N groups of information, wherein the N levels at least include a feature layer, an object layer, and a semantic concept layer, the feature layer is used to describe the appearance characteristics of the electronic document image, the object layer is used to indicate the elements included in the target electronic document, and the semantic concept layer is used to indicate business-related characteristics included in the target electronic document, and the N groups of information at least include feature layer information, object layer information, and semantic concept layer information.

3. The method according to claim 2, characterized in that The adopting N networks in the document classification model to process the N groups of information to obtain the target classification result of the target electronic document includes: Using the first network in the document classification model to process the feature layer information to obtain a first classification result; Using the second network in the document classification model to process the object layer information to obtain a second classification result; Using a third network in the document classification model to process the semantic concept layer information to obtain a third classification result; The target classification result is obtained based on the first classification result, the second classification result and the third classification result.

4. The method according to claim 3, characterized in that The first network includes a first convolutional neural network, the second network includes a second convolutional neural network and a recurrent neural network, and the third network includes a long short-term memory network.

5. The method according to claim 4, characterized in that The adopting the second network in the document classification model to process the object layer information to obtain a second classification result includes: Using the second convolutional neural network to process the image information in the object layer information to obtain a fourth classification result; Using the recurrent neural network to process the text information in the object layer information to obtain a fifth classification result; Based on the fourth classification result and the fifth classification result, the second classification result is obtained.

6. The method according to claim 3, characterized in that The obtaining the target classification result based on the first classification result, the second classification result and the third classification result includes: Determine a first weight corresponding to the first classification result, a second weight corresponding to the second classification result, and a third weight corresponding to the third classification result; The target classification result is obtained based on the first classification result, the second classification result, the third classification result, the first weight, the second weight and the third weight.

7. The method according to any one of claims 1 to 6, characterized in that After the N networks in the document classification model are used to process the N groups of information to obtain the target classification result of the target electronic document, the method further includes: Based on the target classification result, rename the target electronic document to obtain a renaming result; Based on the target classification result, determining a target compression algorithm from a plurality of compression algorithms, wherein the plurality of compression algorithms correspond to different document types; Based on the renaming result, the target electronic document is compressed using the target compression algorithm to obtain a compressed document.

8. An electronic document classification device, characterized in that: include: An acquisition module for a target electronic document; An identification module, used to identify the information included in the target electronic document to obtain N groups of information, where N is an integer greater than or equal to 2, and the N groups of information correspond to different information types; The processing module is used to use N networks in the document classification model to process N groups of information to obtain target classification results of target electronic documents, wherein the N networks correspond one-to-one to the N groups of information, and the document classification model is based on multiple electronic documents and the classification results corresponding to the multiple electronic documents, which are obtained through machine learning.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the electronic filing classification method described in any one of claims 1 to 7.

10. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 7 when running.

Images