Electronic official document automatic filing method and device based on large model training
Through the automatic archiving method of electronic official documents based on large-model training, and using pre-trained language model and feature fusion technology, the problems of low efficiency and poor accuracy of traditional manual archiving are solved, and efficient and standardized automated archiving and management of electronic official documents are achieved.
Patent Information
- Application Number
- CN202411621744.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-14
- Publication Date
- 2025-03-28
AI Technical Summary
Traditional electronic file archiving mainly relies on manual operations, which is time-consuming and labor-intensive, and is prone to errors. Different archivers have inconsistent archiving standards for the same type of files, resulting in inconsistent archiving results, affecting the standardization of file management and search efficiency. As the number of electronic files increases, manual archiving is difficult to meet the needs of efficient management.
The automatic archiving method of electronic official documents based on large model training is adopted. By obtaining completed electronic official documents, it converts them into preset format text, extracting text information, processing noise data, identifying archived years, and storing electronic official documents into the corresponding year directory. Use pre-trained language models and transfer learning technology, combining the content characteristics and metadata of electronic documents, and perform multi-dimensional feature fusion to improve the accuracy and efficiency of archives.
It realizes the automated archiving of electronic documents, improves the standardization and search efficiency of document management, reduces the errors and time-consuming of manual operations, and meets the needs of efficient management of electronic documents. Through transfer learning and feature fusion technology, the effectiveness and efficiency of archive age classification are improved, and the electronic file management needs of different scales are adapted to the needs of electronic file management.
Smart Images
Figure CN119719250A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of electronic document archiving, and in particular to a method and device for automatically archiving electronic documents based on large model training. Background Art
[0002] In the daily operations of modern enterprises and organizations, electronic document management is an important task. With the development of informatization and digitalization, the number and types of electronic files have increased rapidly. How to manage these electronic files efficiently and securely has become a key issue. Archiving of electronic files is one of the important links in file management. Archiving can achieve long-term preservation, quick search and effective use of files. However, traditional electronic file archiving mainly relies on manual operation, which is not only time-consuming and labor-intensive, but also prone to errors. In addition, different archiving personnel may have different archiving standards for the same type of files, resulting in inconsistent archiving results, affecting the standardization of file management and search efficiency. At the same time, with the increase in the number of electronic files, manual archiving can no longer meet the needs of efficient management. Therefore, it needs to be improved. Summary of the invention
[0003] Based on this, it is necessary to provide an automatic archiving method and device for electronic documents based on large model training, in view of the fact that traditional electronic document archiving mainly relies on manual operation, which is not only time-consuming and labor-intensive and prone to errors, but different archiving personnel may have different archiving standards for the same type of files, resulting in inconsistent archiving results, affecting the standardization of file management and search efficiency. At the same time, with the increase in the number of electronic files, manual archiving can no longer meet the needs of efficient management.
[0004] The present invention provides a method for automatically archiving electronic documents based on large model training, the method comprising:
[0005] Obtain completed electronic documents;
[0006] Convert electronic documents into pre-formatted text and extract text information;
[0007] Based on text information, process the noisy data to form clean text data;
[0008] Identify the filing year of the clean text data, output the year classification result of the electronic document, and store the electronic document in the corresponding year directory.
[0009] In one embodiment, identifying the archiving year of the clean text data includes:
[0010] Get clean text data;
[0011] The cleaned text data is input into a pre-trained language model to output a filing year result, wherein the pre-trained language model is based on a deep neural network and is trained using electronic document samples and filing year samples as training data.
[0012] In one embodiment, inputting the cleaned text data into a pre-trained language model and outputting a filing year result comprises:
[0013] Inputting the cleaned text data into a pre-trained language model and outputting a predicted filing year result;
[0014] Compare the predicted filing year results with the actual filing year results to generate a reward signal or a penalty signal;
[0015] Adjust the internal parameters of the pre-trained language model based on the reward signal or penalty signal;
[0016] In response to the predicted filing year result being the same as the actual filing year result, the filing year result is output.
[0017] In one embodiment, identifying the archiving year of the clean text data further includes:
[0018] Fix all layers of the pre-trained language model;
[0019] Extract semantic features of clean text data and obtain high-level feature representations of pre-trained language models to be used as input for electronic document archiving tasks and to connect custom classifiers for training.
[0020] In one embodiment, identifying the archiving year of the clean text data further includes:
[0021] Based on the actual archive year directory of clean text data, the training data is annotated to form a labeled dataset;
[0022] Fine-tune the pre-trained language model based on the labeled dataset.
[0023] In one embodiment, identifying the archiving year of the clean text data further includes:
[0024] Obtain metadata of electronic documents;
[0025] Based on feature fusion technology, clean text data and metadata are combined to construct multi-dimensional classification features.
[0026] In one embodiment, the method further comprises:
[0027] The year classification results of electronic documents are pushed to users via the cloud platform in the form of text messages and emails, and are connected to the big data platform in real time through the data interface.
[0028] The present invention also provides an automatic electronic document filing device based on large model training, comprising:
[0029] The acquisition module is used to obtain completed electronic documents;
[0030] A conversion module, used to convert electronic documents into a preset format text and extract text information;
[0031] A processing module, used for processing noise data based on text information to form clean text data;
[0032] The output module is used to identify the filing year of the clean text data, output the year classification result of the electronic document, and store the electronic document in the corresponding year directory.
[0033] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements any of the above-mentioned methods for automatic archiving of electronic documents based on large model training.
[0034] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-described methods for automatically archiving electronic documents based on large model training.
[0035] The above-mentioned automatic archiving method and device for electronic documents based on large model training, by converting the acquired completed electronic documents into a unified text format, facilitates the improvement of the standardization of file management, and by automatically identifying and outputting the archiving year of the electronic documents, facilitates the unification of archiving standards. At the same time, by automatically storing the electronic documents in the corresponding year directory, it is convenient to improve the search efficiency and meet the needs of efficient management of electronic documents.
[0036] On the one hand, by using pre-trained language models and transfer learning, the effect and efficiency of archive age classification can be significantly improved in terms of reducing feature engineering workload, improving the ability to process complex semantics, maintaining good performance in data sparse conditions, and reducing dependence on large amounts of labeled data. At the same time, through the generated reward signals or penalty signals, enhanced learning technology can be used to continuously optimize the internal parameters and archiving strategies of the pre-trained language model based on user feedback and actual application effects, thereby improving intelligence and adaptability. On the other hand, the content features and metadata of electronic documents are comprehensively utilized for classification, providing multi-dimensional feature inputs to improve the accuracy and robustness of the pre-trained language model. At the same time, through feature fusion technology, text features and metadata features are combined to facilitate the construction of a more comprehensive and effective classification feature space. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0038] Figure 1 The figure is a schematic diagram of the process of an automatic filing method of electronic documents based on large model training in one embodiment;
[0039] Figure 2 A schematic diagram of a process for identifying the filing year of an electronic document in one embodiment;
[0040] Figure 3 A schematic diagram of the output process of the electronic document filing year result in one embodiment;
[0041] Figure 4 A schematic diagram of a process for obtaining high-level features in an embodiment;
[0042] Figure 5 A schematic diagram of a fine-tuning process of a pre-trained language model in one embodiment;
[0043] Figure 6 A schematic diagram of a process for constructing multi-dimensional classification features in an embodiment;
[0044] Figure 7 A schematic diagram of the structure of an automatic electronic document filing device based on large model training in one embodiment;
[0045] Figure 8 The figure is a diagram showing the internal structure of an electronic device according to an embodiment. DETAILED DESCRIPTION
[0046] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0047] Combine the following Figure 1-Figure 8 The present invention describes the method and device for automatically archiving electronic documents based on large model training.
[0048] like Figure 1 As shown, in one embodiment, a method for automatically filing electronic documents based on large model training includes the following steps:
[0049] Step S100, obtaining the completed electronic document.
[0050] Step S200, converting the electronic document into a preset format text and extracting text information.
[0051] Step S300, based on the text information, processing the noise data to form clean text data.
[0052] It should be added that processing noisy data is an important part of the data analysis process, which aims to reduce the adverse effects of noise data on analysis and prediction, facilitate the removal of irrelevant content, and ensure the quality of training data.
[0053] Step S400, identifying the filing year of the clean text data, outputting the year classification result of the electronic document, and storing the electronic document in the corresponding year directory.
[0054] This automatic archiving method for electronic documents based on large model training converts the acquired completed electronic documents into a unified text format, thereby facilitating the improvement of the standardization of file management, and automatically identifies and outputs the archiving year of the electronic documents, thereby facilitating the unification of archiving standards. At the same time, by automatically storing the electronic documents in the corresponding year directory, it is convenient to improve the search efficiency and meet the needs of efficient management of electronic documents.
[0055] Since traditional classification algorithms usually require manual design and feature extraction, this process is complex and time-consuming, and requires the participation of domain experts to select and optimize features. In addition, when the data is sparse, the model performance of traditional classification algorithms is easily affected, especially when the amount of training data is insufficient, it is difficult to obtain a good classification effect. Figure 2 , identify the archive year of clean text data, including:
[0056] Step S410, obtaining clean text data.
[0057] Step S420, input the cleaned text data into a pre-trained language model and output the filing year result. The pre-trained language model is based on a deep neural network and is trained using electronic document samples and filing year samples as training data.
[0058] Pre-trained language model is a commonly used method in transfer learning. By using pre-trained language model and transfer learning, the effect and efficiency of archive age classification can be significantly improved by reducing the workload of feature engineering, improving the ability to process complex semantics, maintaining good performance in data-sparse conditions, and reducing dependence on large amounts of labeled data.
[0059] In this embodiment, see Figure 3 , input the clean text data into the pre-trained language model, and output the archive year results, including:
[0060] Step S421, input the cleaned text data into the pre-trained language model and output the predicted filing year result.
[0061] Step S422, compare the predicted filing year result with the actual filing year result, and generate a reward signal or a penalty signal.
[0062] Step S423: adjusting internal parameters of the pre-trained language model based on the reward signal or the penalty signal.
[0063] Step S424, in response to the predicted filing year result being the same as the actual filing year result, outputting the filing year result.
[0064] By generating reward signals or penalty signals and enhancing learning technology, it is convenient to continuously optimize the internal parameters and archiving strategies of the pre-trained language model according to user feedback and actual application effects, thereby improving intelligence and adaptability.
[0065] In this embodiment, see Figure 4 , identifying the archive year of clean text data, and also including:
[0066] Step S430, fixing all layers of the pre-trained language model.
[0067] Step S440, extracting semantic features of the cleaned text data and obtaining high-level feature representations of a pre-trained language model to be used as input for the electronic document archiving task and to be connected to a custom classifier for training.
[0068] By learning nonlinear combinations of low-level features, high-level feature representations are automatically learned. These high-level features have stronger representation capabilities and can better capture complex patterns in the data, thereby facilitating the processing of complex natural language semantics and capturing deep semantic relationships in text.
[0069] In this embodiment, see Figure 5 , identifying the archive year of clean text data, and also including:
[0070] Step S450, annotating the training data based on the actual archive year directory of the cleaned text data to form a labeled data set.
[0071] Step S460: fine-tune the pre-trained language model based on the labeled data set.
[0072] By fine-tuning the pre-trained language model, the pre-trained language model can better adapt to the text classification needs of specific fields and improve classification accuracy.
[0073] In this embodiment, see Figure 6 , identifying the archive year of clean text data, and also including:
[0074] Step S470, obtaining metadata of the electronic document.
[0075] Step S480, based on feature fusion technology, the clean text data and metadata are combined to construct multi-dimensional classification features.
[0076] Comprehensively utilize the content features and metadata (such as document type and drafting department) of electronic documents for classification, provide multi-dimensional feature input, improve the accuracy and robustness of the pre-trained language model, and at the same time, combine text features and metadata features through feature fusion technology to facilitate the construction of a more comprehensive and effective classification feature space.
[0077] In this embodiment, the method for automatically archiving electronic documents based on large model training also includes:
[0078] Step S500: The year classification result of the electronic document is pushed to the user via the cloud platform in the form of SMS or email, and is connected to the big data platform in real time via the data interface.
[0079] This facilitates real-time monitoring of the archiving process, allowing archive system administrators to provide real-time manual feedback on archiving results and continuously improve the classification capabilities of pre-trained language models.
[0080] The following is a description of the automatic electronic document archiving device based on large model training provided by the present invention. The automatic electronic document archiving device based on large model training described below and the automatic electronic document archiving method based on large model training described above can be referenced to each other.
[0081] like Figure 7 As shown, in one embodiment, an automatic electronic document archiving device based on large model training includes an acquisition module 710, a conversion module 720, a processing module 730 and an output module 740.
[0082] The acquisition module 710 is used to acquire completed electronic documents.
[0083] The conversion module 720 is used to convert the electronic document into a text in a preset format and extract text information.
[0084] The processing module 730 is used to process the noise data based on the text information to form clean text data.
[0085] The output module 740 is used to identify the archiving year of the clean text data, output the year classification result of the electronic document, and store the electronic document in the corresponding year directory.
[0086] In this embodiment, the archiving year of the clean text data is identified, specifically for:
[0087] Get clean text data;
[0088] The clean text data is input into the pre-trained language model and the filing year result is output. The pre-trained language model is based on a deep neural network and is trained using electronic document samples and filing year samples as training data.
[0089] In this embodiment, clean text data is input into a pre-trained language model and the filing year result is output, which is specifically used for:
[0090] Input the clean text data into the pre-trained language model and output the predicted filing year result;
[0091] Compare the predicted filing year results with the actual filing year results to generate a reward signal or a penalty signal;
[0092] Adjust the internal parameters of the pre-trained language model based on the reward signal or penalty signal;
[0093] In response to the predicted filing year result being the same as the actual filing year result, the filing year result is output.
[0094] In this embodiment, identifying the archiving year of the clean text data is also specifically used for:
[0095] Fix all layers of the pre-trained language model;
[0096] Extract semantic features of clean text data and obtain high-level feature representations of pre-trained language models to be used as input for electronic document archiving tasks and to connect custom classifiers for training.
[0097] In this embodiment, identifying the archiving year of the clean text data is also specifically used for:
[0098] Based on the actual archive year directory of clean text data, the training data is annotated to form a labeled dataset;
[0099] Fine-tune the pre-trained language model based on the labeled dataset.
[0100] In this embodiment, identifying the archiving year of the clean text data is also specifically used for:
[0101] Obtain metadata of electronic documents;
[0102] Based on feature fusion technology, clean text data and metadata are combined to construct multi-dimensional classification features.
[0103] In this embodiment, the device further includes:
[0104] The push module 750 is used to push the year classification results of electronic documents to users through the cloud platform in the form of text messages and emails, and access the big data platform in real time through the data interface.
[0105] The automatic electronic document archiving device based on large model training converts the acquired completed electronic documents into a unified text format, thereby facilitating the improvement of the standardization of file management. It automatically identifies and outputs the archiving year of the electronic documents, thereby facilitating the unification of archiving standards. At the same time, it automatically stores the electronic documents in the corresponding year directory, thereby facilitating the improvement of search efficiency and meeting the needs of efficient management of electronic documents.
[0106] Figure 8 An example of a physical structure diagram of an electronic device is shown. The electronic device may be a smart terminal, and its internal structure diagram may be as follows: Figure 8As shown. The electronic device includes a processor, a memory and a network interface connected via a system bus. The processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the electronic device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for automatically filing electronic documents based on large model training is implemented, and the method includes:
[0107] Obtain completed electronic documents;
[0108] Convert electronic documents into pre-formatted text and extract text information;
[0109] Based on text information, process the noisy data to form clean text data;
[0110] Identify the filing year of the clean text data, output the year classification result of the electronic document, and store the electronic document in the corresponding year directory.
[0111] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present invention, and does not constitute a limitation on the electronic device to which the scheme of the present invention is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0112] On the other hand, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, implements a method for automatically filing electronic documents based on large model training, the method comprising:
[0113] Obtain completed electronic documents;
[0114] Convert electronic documents into pre-formatted text and extract text information;
[0115] Based on text information, process the noisy data to form clean text data;
[0116] Identify the filing year of the clean text data, output the year classification result of the electronic document, and store the electronic document in the corresponding year directory.
[0117] In another aspect, a computer program product or a computer program is provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and when the processor executes the computer instructions, a method for automatically filing electronic documents based on large model training is implemented, the method comprising:
[0118] Obtain completed electronic documents;
[0119] Convert electronic documents into pre-formatted text and extract text information;
[0120] Based on text information, process the noisy data to form clean text data;
[0121] Identify the filing year of the clean text data, output the year classification result of the electronic document, and store the electronic document in the corresponding year directory.
[0122] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory.
[0123] By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0124] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0125] The above-described embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.
Claims
1. A method for automatic filing of electronic documents based on large model training, characterized in that: The method comprises: Obtain completed electronic documents; Convert electronic documents into pre-formatted text and extract text information; Based on text information, process the noisy data to form clean text data; Identify the filing year of the clean text data, output the year classification result of the electronic document, and store the electronic document in the corresponding year directory.
2. The method for automatic filing of electronic documents based on large model training according to claim 1 is characterized in that: The identification of the archived year of the clean text data includes: Get clean text data; The cleaned text data is input into a pre-trained language model to output a filing year result, wherein the pre-trained language model is based on a deep neural network and is trained using electronic document samples and filing year samples as training data.
3. The method for automatic filing of electronic documents based on large model training according to claim 2 is characterized in that: The step of inputting the cleaned text data into a pre-trained language model and outputting a filing year result comprises: Inputting the cleaned text data into a pre-trained language model and outputting a predicted filing year result; Compare the predicted filing year results with the actual filing year results to generate a reward signal or a penalty signal; Adjust the internal parameters of the pre-trained language model based on the reward signal or penalty signal; In response to the predicted filing year result being the same as the actual filing year result, the filing year result is output.
4. The method for automatically filing electronic documents based on large model training according to claim 2 is characterized in that: The identifying the archiving year of the clean text data also includes: Fix all layers of the pre-trained language model; Extract semantic features of clean text data and obtain high-level feature representations of pre-trained language models to be used as input for electronic document archiving tasks and to connect custom classifiers for training.
5. The method for automatically filing electronic documents based on large model training according to claim 4 is characterized in that: The identifying the archiving year of the clean text data also includes: Based on the actual archive year directory of clean text data, the training data is annotated to form a labeled dataset; Fine-tune the pre-trained language model based on the labeled dataset.
6. The method for automatically filing electronic documents based on large model training according to any one of claims 1 to 5, characterized in that: The identifying the archiving year of the clean text data also includes: Obtain metadata of electronic documents; Based on feature fusion technology, clean text data and metadata are combined to construct multi-dimensional classification features.
7. The method for automatically filing electronic documents based on large model training according to claim 6 is characterized in that: The method further includes: The year classification results of electronic documents are pushed to users via the cloud platform in the form of text messages and emails, and are connected to the big data platform in real time through the data interface.
8. An automatic electronic document filing device based on large model training, characterized in that: include: The acquisition module is used to obtain completed electronic documents; A conversion module, used to convert electronic documents into a preset format text and extract text information; A processing module, used for processing noise data based on text information to form clean text data; The output module is used to identify the filing year of the clean text data, output the year classification result of the electronic document, and store the electronic document in the corresponding year directory.
9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for automatic archiving of electronic documents based on large model training as described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for automatic archiving of electronic documents based on large model training as described in any one of claims 1 to 7 are implemented.