Intelligent archive management method and system based on large language model

Abstract

The invention provides an intelligent archive management method and system based on a large language model. The method comprises the following steps: in response to any current operation end, performing current scanning on an archive label corresponding to a current archive, calling an original archive, and performing random extraction to determine a content checking group; triggering an archive checking plug-in to check the current archive based on the content checking group, and determining the document state of the current archive based on a checking result; in response to the fact that the current file is in a tampered state, determining tampered content of the current file based on a large language model, and determining a proximity operation end for performing proximity scanning on the current file based on a called operation record table corresponding to the current file; and establishing an initial archive comparison layer, respectively filling the tampered content and the adjacent operation end into a content indication area and a terminal indication area in the initial archive comparison layer, and sending the obtained current archive comparison layer to a management end for display. The file management efficiency is at least improved.

Description

Technical Field

[0001] The present invention relates to data processing technology, and in particular to an intelligent archive management method and system based on a large language model. Background Art

[0002] At present, in the current archive management practice, ensuring the authenticity and integrity of archives is a crucial issue. However, due to various reasons, archives are at risk of being tampered with, and once tampering occurs, it is often difficult to accurately identify the tamperer. This situation not only seriously threatens the authority and credibility of archives, but also brings great challenges to archive management.

[0003] Therefore, how to develop a method and system for archive management that can fully utilize large language models to achieve efficient and intelligent management has become an urgent problem to be solved. Summary of the invention

[0004] Based on the above problems, the present invention is proposed to provide an intelligent archive management method and system based on a large language model that overcomes the above problems or at least partially solves the above problems.

[0005] According to one aspect of the present invention, there is provided an intelligent archive management method based on a large language model, comprising the following steps: In response to any current operation terminal performing a current scan on a file tag corresponding to a current file, an original file corresponding to the current file is retrieved, and original files included in the original file are randomly extracted based on a preset extraction strategy, and the extracted original files are determined as a content verification group; Triggering the archive verification plug-in to perform integrity-based file verification on the current archive based on the content verification group, and determining a document status corresponding to the current archive based on the verification result, wherein the document status includes a complete status and a tampered status; In response to the current file being in a tampered state, determining the tampered content corresponding to the current file, and determining an adjacent operation terminal for adjacent scanning of the current file based on the retrieved operation record table corresponding to the current file; An initial file comparison layer is established based on the current file, the tampered content and the adjacent operation terminal are respectively filled into the content indication area and the terminal indication area in the initial file comparison layer, and the obtained current file comparison layer is sent to the management terminal for display.

[0006] Optionally, in the method according to the present invention, randomly extracting each original file included in the original archive based on a preset extraction strategy, and determining each extracted original file as a content verification group, including: Determine each original document in the original file that has been pre-checked as a necessary check group, and determine all other original documents in the original file except the necessary check group as a random check group; Performing content evaluation on each original file in the random check group, and summing up the extraction priority values ​​corresponding to each original file to obtain a total extraction value; Calculate the ratio of each extraction priority value to the total extraction value to obtain the extraction probabilities corresponding to the original files; Determine the number of files corresponding to each original file in the content verification group, and calculate the product of the number of files and the preset extraction ratio to obtain the current extraction number; Creating a random drawing wheel, and dividing the random drawing wheel into regions based on the respective drawing probabilities corresponding to the respective original files, to obtain respective sector-shaped regions corresponding to the respective original files; The random drawing turntable is controlled to rotate and draw corresponding to the current drawing quantity, and based on the drawing result, each drawn original file and all original files in the necessary check group are determined as a content check group.

[0007] Optionally, in the method according to the present invention, performing content evaluation on each original file in the random check group includes: Determine each original file in the random check group based on the file attribute, and determine each original file with sensitive attributes as a sensitive file group, and determine each original file with common attributes as a common file group; Retrieving the first attribute evaluation value to respectively assign scores to the original files in the sensitive file group, and retrieving the second attribute evaluation value to respectively assign scores to the original files in the common file group, wherein the first attribute evaluation value is greater than the second attribute evaluation value; Perform pixel identification on each original file in the random check group to obtain pixel points of each file in the same original file, and obtain the number of pixels corresponding to each file pixel point to obtain the number of pixels corresponding to each original file; Compare the number of pixels with a first number threshold and a second number threshold respectively, wherein the first number threshold is less than the second number threshold; In response to the number of pixels corresponding to any original file being less than the first number threshold, calling the first number evaluation value to perform score configuration on the original file based on the number dimension; In response to the number of pixels corresponding to any original file being greater than or equal to the first number threshold and less than the second number threshold, calling the second number evaluation value to perform score configuration based on the number dimension on the original file; In response to the number of pixels corresponding to any original file being greater than or equal to the second number threshold, calling a third quantity evaluation value to perform score configuration on the original file based on the quantity dimension; The attribute evaluation value and the quantity evaluation value corresponding to the same original file are summed up to obtain the extraction priority values ​​corresponding to each original file.

[0008] Optionally, in the method according to the present invention, determining each original file in the random check group based on the file attribute, and determining each original file with sensitive attributes as a sensitive file group, and determining each original file with ordinary attributes as a ordinary file group, comprises: Obtaining the page numbers of the files respectively corresponding to the original files, and determining the original files having the same page number as the retrieved preset directory page number as the directory files in the original files; Extracting the directory file to obtain the directory titles and the page number intervals corresponding to the directory titles; Classifying each directory title hierarchically to obtain a hierarchical structure tree including nodes of different hierarchical levels, wherein each hierarchical node corresponds to a different directory title; Retrieving a preset sensitive record table, wherein the preset sensitive record table includes various sensitive characters; Determining the directory titles corresponding to the level nodes based on the level order of the level structure tree, and in response to at least a portion of the directory title corresponding to any level node being located in the preset sensitive record table, determining the level node and all level nodes connected downward to the level node as having sensitive attributes; Based on the page number intervals corresponding to each hierarchical node with sensitive attributes, all original files whose corresponding file page numbers are located in the page number intervals in the random verification group are determined to have sensitive attributes, and all remaining original files in the random verification group are determined to have ordinary attributes.

[0009] Optionally, in the method according to the present invention, each file line arranged in a vertical order from top to bottom in the directory file is obtained, and all file lines with leading characters are determined as directory titles; Perform character recognition based on a large language model on each directory title to obtain characters of each file in the same directory title; Grouping the file characters in the same directory title by category to obtain a text grouping of the file characters corresponding to the text category and a digital grouping of the file characters corresponding to the digital category; Determine the title page number corresponding to the directory title based on each file character in the digital group, and obtain each title page number corresponding to each directory title; Determine the size of any file character in the text group respectively, and obtain the size of each title corresponding to each text group; Arrange the titles of each directory in a hierarchical manner from large to small according to the corresponding title size, and obtain the hierarchical levels corresponding to each directory title; Based on the vertical order, two directory titles at adjacent positions and corresponding to the same hierarchical level are respectively determined as adjacent title groups, and the corresponding first directory title in the same adjacent title group is determined as the first title, and the corresponding last directory title is determined as the last title; Determine the title page number corresponding to the first title and the title page numbers corresponding to all the directory titles between the first title and the last title as an interval determination group, and determine the maximum page number with the largest corresponding value and the minimum page number with the smallest corresponding value based on all the title page numbers in the interval determination group; The page number interval consisting of the minimum page number and the maximum page number is determined to correspond to the first title, and each page number interval corresponding to each directory title is obtained.

[0010] Optionally, in the method according to the present invention, the method further comprises: Performing handwriting recognition on each original document in the random check group based on a handwriting recognition model, and in response to determining that a handwriting area exists in any original document based on the recognition result, determining that the original document has a handwriting attribute; Determine a handwriting extension direction corresponding to the handwriting area, and compare the handwriting extension direction with a horizontal direction corresponding to the original document to obtain a handwriting angle between the handwriting extension direction and the horizontal direction; In response to the handwriting angle being within a preset angle interval, determining a first handwriting weight of a corresponding direction dimension based on the handwriting angle; Determine the area ratio of the handwriting area to the original document, and determine a second handwriting weight corresponding to the ratio dimension based on the area ratio; The first handwriting weight and the second handwriting weight are summed to obtain a handwriting reduction weight corresponding to the original document with handwriting attributes; The extraction priority value corresponding to the original file is reduced and updated based on the handwriting reduction weight to obtain an updated extraction priority value.

[0011] Optionally, in the method according to the present invention, determining the handwriting extension direction corresponding to the handwriting area includes: Establishing a file coordinate system corresponding to the original file, wherein an X-axis of the file coordinate system is parallel to a horizontal direction corresponding to the original file; Performing pixel processing on the handwritten area to obtain handwritten pixel points constituting the handwritten area; Based on the file coordinate system, each handwritten coordinate point corresponding to each handwritten pixel point is obtained, and each handwritten coordinate point corresponding to the maximum horizontal coordinate value, the minimum horizontal coordinate value, the maximum vertical coordinate value, and the minimum vertical coordinate value is determined as the maximum horizontal coordinate point, the minimum horizontal coordinate point, the maximum vertical coordinate point, and the minimum vertical coordinate point; Taking the maximum horizontal coordinate point and the minimum horizontal coordinate point as starting points, respectively, a first longitudinal line segment and a second longitudinal line segment extending in the vertical direction are constructed, and taking the maximum vertical coordinate point and the minimum vertical coordinate point as starting points, respectively, a first horizontal line segment and a second horizontal line segment extending in the horizontal direction are constructed; Controlling the first longitudinal line segment, the second longitudinal line segment, the first transverse line segment and the second transverse line segment to extend to form an initial rectangle, and obtaining the corner points corresponding to the initial rectangle; Determine the center point of the rectangle corresponding to the initial rectangle, and obtain the connecting lines between each corner point and the center point of the rectangle; Taking each corner point as a starting point, establishing each corner extension line perpendicular to each point connection line, and moving each corner extension line along the corresponding point connection line toward the center point of the rectangle; In response to each corner extension line moving to contact different handwritten pixel points, each corner extension line is controlled to stop moving to form a current rectangle; Rectangular frame lines constituting the current rectangle are acquired, and a handwriting extension direction corresponding to the handwriting area is determined based on extension directions of frame lines corresponding to the rectangular frame lines.

[0012] Optionally, in the method according to the present invention, the method further comprises: Acquire each historical verification information corresponding to each original file in the random verification group, wherein each historical verification information includes each verification operation arranged in a forward order based on the verification time, and each verification operation has corresponding verification results; Performing reverse extraction based on the forward arrangement for each piece of historical verification information, and determining the verification operations obtained in a continuous arrangement and corresponding to a preset number as a continuous verification group; In response to all verification results corresponding to the same continuous verification group being verified, the preset reduction weight is called to reduce and update the extraction priority value corresponding to the original file to obtain an updated extraction priority value.

[0013] Optionally, in the method according to the present invention, triggering the archive verification plug-in to perform integrity-based file verification on the current archive based on the content verification group, and determining a document status corresponding to the current archive based on the verification result, wherein the document status includes an intact status and a tampered status, including: Based on the archive verification plug-in, each current file having a traceability relationship with each original file in the content verification group is retrieved from the current archive, and each original file and each current file are binarized to obtain each original image and each current image, wherein any of the original images includes each first image pixel point corresponding to the first pixel value, and any of the current images includes each second image pixel point corresponding to the second pixel value; Creating a transparent fusion layer, and sequentially stacking the original file and the current file having the same traceability relationship on the transparent fusion layer; In response to the transparent fusion layer including any first image pixel point and / or any second image pixel point, the file status corresponding to the current archive is determined to be a tampered status.

[0014] Optionally, in the method according to the present invention, an initial archive comparison layer is established based on the current archive, the tampered content and the adjacent operation terminal are respectively filled into a content indication area and a terminal indication area located in the initial archive comparison layer, and the obtained current archive comparison layer is sent to the management terminal for display, including: Establishing an initial file comparison layer based on the current file, wherein the initial file comparison layer includes a vertically arranged content indication area and a terminal indication area for filling an adjacent operating terminal; Determine a current file in the current archive that contains the tampered content, obtain each file line in the current file that is arranged in a vertical order from top to bottom, and determine all file lines that contain a blank group consisting of two adjacent blank characters as paragraph start lines; Based on the vertical order, two paragraph start lines in adjacent positions are respectively determined as adjacent paragraph groups, and the corresponding first paragraph start line in the same adjacent paragraph group is determined as the first paragraph, and the corresponding last paragraph start line is determined as the last paragraph; Determine the first paragraph and all file lines between the first paragraph and the last paragraph as the same file paragraph, and obtain the file paragraphs in the current file; Determine a file paragraph including the tampered content as a tampered paragraph, and determine an original paragraph corresponding to the tampered paragraph in an original file having the same traceability relationship as the current file; Generating a first bounding box enclosing the tampered paragraph in the current file, and generating a second bounding box enclosing the original paragraph in the original file; Generate a first indication area and a second indication area arranged horizontally in the content indication area, and fill the first indication area with the current file having the first enclosing frame, and fill the second indication area with the original file having the second enclosing frame; The adjacent operating end is filled into the terminal indication area, and the obtained current file comparison layer is sent to the management end for display.

[0015] According to another aspect of the present invention, there is provided an intelligent archive management system based on a large language model, comprising: an extraction module, configured to retrieve an original file corresponding to the current file in response to any current operation terminal performing a current scan on a file tag corresponding to the current file, and randomly extract each original file included in the original file based on a preset extraction strategy, and determine each extracted original file as a content verification group; a verification module configured to trigger the archive verification plug-in to perform an integrity-based file verification on the current archive based on the content verification group, and determine a document status corresponding to the current archive based on the verification result, wherein the document status includes an intact status and a tampered status; a determination module, configured to determine the tampered content corresponding to the current file in response to the current file being in a tampered state, and determine an adjacent operation terminal for adjacent scanning of the current file based on the retrieved operation record table corresponding to the current file; The establishment module is configured to establish an initial archive comparison layer based on the current archive, fill the tampered content and the adjacent operation terminal into the content indication area and the terminal indication area located in the initial archive comparison layer respectively, and send the obtained current archive comparison layer to the management terminal for display.

[0016] According to the solution of the present invention, the present invention can randomly extract original archives in response to the scanning of archive labels by the operating terminal to form a content verification group, which significantly improves the efficiency of archive verification; subsequently, the present invention can realize efficient file verification based on integrity of the current archive through the archive verification plug-in, and quickly and accurately determine the document status (complete status or tampered status) of the archive, thereby effectively ensuring the originality and authenticity of the archive; and once the present invention detects that the archive is in a tampered state, it can accurately identify and locate the tampered content, and quickly lock the adjacent operating terminal that is performing an adjacent scan of the current archive based on the operation record table; this double confirmation mechanism not only helps to timely discover and correct false information in the archive, but also facilitates subsequent responsibility It provides strong support for traceability; finally, the present invention can present key information such as tampered content and adjacent operation terminals in a graphical manner by establishing an initial file comparison layer, so that the management end can intuitively understand the true status and potential problems of the files; this visual management method not only facilitates management personnel to quickly grasp the results of file verification, but also greatly improves the convenience and transparency of file management; the present invention integrates a variety of technical means such as random sampling, integrity verification, large language model analysis and graphical display, and realizes efficient verification, precise positioning, intuitive display and intelligent management of files, which not only helps to improve the accuracy and efficiency of file management, but also lays a solid foundation for building a safer and more reliable file management system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flow chart of an intelligent archive management method based on a large language model according to an embodiment of the present invention is shown; Figure 2 A schematic diagram of a random drawing turntable in this embodiment is shown; Figure 3 A structural block diagram of an intelligent archive management system based on a large language model according to another embodiment of the present invention is shown. DETAILED DESCRIPTION

[0018] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0019] To solve the problems in the background technology, the inventor proposes the solution of the present invention. An embodiment of the present invention provides an intelligent archive management method based on a large language model, which can be executed in a computing device, wherein the computing device can be understood as a terminal with data processing function.

[0020] Figure 1 A flowchart of an intelligent archive management method based on a large language model according to an embodiment of the present invention is shown, and the method is suitable for being executed in a computing device.

[0021] like Figure 1 As shown, the intelligent archive management method based on the large language model proposed in this embodiment begins with step S102. In step S102, the following contents are included: In response to any current operation terminal performing a current scan on an archive tag corresponding to a current archive, an original archive corresponding to the current archive is retrieved, and original files included in the original archive are randomly extracted based on a preset extraction strategy, and the extracted original files are determined as a content verification group.

[0022] For example, in this embodiment, the current operation end can be understood as an operation terminal corresponding to the operator who performs corresponding operations on the current file. A corresponding file label is set on each current file. When any operator expects to perform a corresponding viewing operation on the current file, the operator can use the corresponding current operation end to perform a current scan on the file label corresponding to the current file, so that the server can retrieve the original file corresponding to the current file, and then randomly extract each original file included in the original file according to a preset extraction strategy, so as to determine the extracted original files as a content verification group, so as to perform the integrity comparison of the current file in the following content based on the content verification group; here, the original file can be understood as the original version of the corresponding current file, that is, the file that has never been subjected to the corresponding operation. The original file can exist in the server in electronic form, and the corresponding current file can exist in paper form and / or electronic form.

[0023] Furthermore, the above-mentioned “randomly extracting each original file included in the original archive based on a preset extraction strategy, and determining each extracted original file as a content verification group” further includes the following steps: Determine each original document in the original file that has been pre-checked as a necessary check group, and determine all other original documents in the original file except the necessary check group as a random check group; Performing content evaluation on each original file in the random check group, and summing up the extraction priority values ​​corresponding to each original file to obtain a total extraction value; Calculate the ratio of each extraction priority value to the total extraction value to obtain the extraction probabilities corresponding to the original files; Determine the number of files corresponding to each original file in the content verification group, and calculate the product of the number of files and the preset extraction ratio to obtain the current extraction number; Creating a random drawing wheel, and dividing the random drawing wheel into regions based on the respective drawing probabilities corresponding to the respective original files, to obtain respective sector-shaped regions corresponding to the respective original files; The random drawing turntable is controlled to rotate and draw corresponding to the current drawing quantity, and based on the drawing result, each drawn original file and all original files in the necessary check group are determined as a content check group.

[0024] For example, in this embodiment, the server will first determine each original file in the original file that has been pre-marked for verification as a necessary verification group, and then determine all other original files included in the original file except the necessary verification group as random verification groups; it can be explained that each original file that has been pre-marked for verification can be understood as an original file that needs to be compulsorily verified.

[0025] Then, the server will evaluate the content of each original file in the random verification group to obtain the extraction priority values ​​corresponding to each original file, and then sum up the extraction priority values ​​to obtain the total extraction value.

[0026] Next, the server will calculate the ratio of each extraction priority value to the total extraction value, so as to obtain the extraction probabilities corresponding to each original file, and then determine the number of files of each original file in the content verification group, and multiply the number of files by the preset extraction ratio to obtain the current extraction number.

[0027] Next, the server will create a random extraction wheel and divide the random extraction wheel into regions according to the extraction probabilities corresponding to the original files, thereby obtaining the sector-shaped regions corresponding to the original files, such as Figure 2 In the random drawing wheel shown, if there are three original files A, B, and C with corresponding extraction probabilities of 30%, 50%, and 20% respectively, then three sector-shaped areas with corresponding area proportions of 30%, 50%, and 20% will be generated in the random drawing wheel. Figure 2 The black arrow in the figure is the corresponding extraction rotating needle, that is, when the extraction rotating needle rotates to any sector area, it means that the original file corresponding to the area is extracted.

[0028] Finally, the server will control the random drawing turntable to rotate and draw corresponding to the current number of draws. For example, if the current number of draws is 2, the server will control the random drawing turntable to rotate and draw twice, so as to determine the drawn original files and all the original files in the necessary verification group as the content verification group according to the drawing results, among which the original files in the content verification group are the files that need to be verified accordingly.

[0029] Furthermore, the above-mentioned “performing content evaluation on each original file in the random check group” further includes the following steps: Determine each original file in the random check group based on the file attribute, and determine each original file with sensitive attributes as a sensitive file group, and determine each original file with common attributes as a common file group; Retrieving the first attribute evaluation value to respectively assign scores to the original files in the sensitive file group, and retrieving the second attribute evaluation value to respectively assign scores to the original files in the common file group, wherein the first attribute evaluation value is greater than the second attribute evaluation value; Perform pixel identification on each original file in the random check group to obtain pixel points of each file in the same original file, and obtain the number of pixels corresponding to each file pixel point to obtain the number of pixels corresponding to each original file; Compare the number of pixels with a first number threshold and a second number threshold respectively, wherein the first number threshold is less than the second number threshold; In response to the number of pixels corresponding to any original file being less than the first number threshold, calling the first number evaluation value to perform score configuration on the original file based on the number dimension; In response to the number of pixels corresponding to any original file being greater than or equal to the first number threshold and less than the second number threshold, calling the second number evaluation value to perform score configuration based on the number dimension on the original file; In response to the number of pixels corresponding to any original file being greater than or equal to the second number threshold, calling a third quantity evaluation value to perform score configuration on the original file based on the quantity dimension; The attribute evaluation value and the quantity evaluation value corresponding to the same original file are summed up to obtain the extraction priority values ​​corresponding to each original file.

[0030] For example, in this embodiment, since some original files may contain more important information, that is, have sensitive attributes, and some original files may contain more conventional information, that is, have ordinary attributes, the server will determine each original file in the random verification group based on the file attributes, thereby determining each original file with sensitive attributes as a sensitive file group and determining each original file with ordinary attributes as a ordinary file group.

[0031] Then, the server will retrieve the first attribute evaluation value to assign scores to each original file in the sensitive file group, and then retrieve the second attribute evaluation value to assign scores to each original file in the ordinary file group. Since the original files with sensitive attributes are more important, the first attribute evaluation value is greater than the second attribute evaluation value.

[0032] Next, the server will perform pixel identification on each original file in the random verification group to obtain the pixel points of each file in the same original file, and then obtain the number of pixels corresponding to each file pixel point in the same original file, that is, the number of pixels corresponding to each original file.

[0033] Different numbers of pixels represent the number of words in the original file corresponding to different numbers of pixels. The server will compare the number of pixels with the retrieved first number threshold and second number threshold respectively. Here, the first number threshold is smaller than the second number threshold.

[0034] When the number of pixels in any original file is less than the first quantity threshold, the server will call the first quantity evaluation value to configure the score of the original file based on the quantity dimension; when the number of pixels in any original file is greater than or equal to the first quantity threshold and less than the second quantity threshold, the server will call the second quantity evaluation value to configure the score of the original file based on the quantity dimension; when the number of pixels in any original file is greater than or equal to the second quantity threshold, the server will call the third quantity evaluation value to configure the score of the original file based on the quantity dimension.

[0035] Then, the server will sum up the attribute evaluation value and the quantity evaluation value corresponding to the same original file, so as to obtain the extraction priority value corresponding to each original file.

[0036] Furthermore, the above-mentioned “determining each original file in the random check group based on the file attribute, and determining each original file with sensitive attributes as a sensitive file group, and determining each original file with common attributes as a common file group” also includes the following steps: Obtaining the page numbers of the files respectively corresponding to the original files, and determining the original files having the same page number as the retrieved preset directory page number as the directory files in the original files; Extracting the directory file to obtain the directory titles and the page number intervals corresponding to the directory titles; Classifying each directory title hierarchically to obtain a hierarchical structure tree including nodes of different hierarchical levels, wherein each hierarchical node corresponds to a different directory title; Retrieving a preset sensitive record table, wherein the preset sensitive record table includes various sensitive characters; Determining the directory titles corresponding to the level nodes based on the level order of the level structure tree, and in response to at least a portion of the directory title corresponding to any level node being located in the preset sensitive record table, determining the level node and all level nodes connected downward to the level node as having sensitive attributes; Based on the page number intervals corresponding to each hierarchical node with sensitive attributes, all original files whose corresponding file page numbers are located in the page number intervals in the random verification group are determined to have sensitive attributes, and all remaining original files in the random verification group are determined to have ordinary attributes.

[0037] For example, in this embodiment, the server will first obtain the file page number of each original file, and then determine the original file with the same page number as the retrieved preset directory page number in the original archive as the directory file. Because under normal circumstances, the page number where the directory is located is mostly the second page, that is, the directory page number of the corresponding directory file is 2, the preset directory page number pre-set in the server can be 2, so that the server can determine the corresponding directory file according to the preset directory page number.

[0038] Then, the server will extract the directory file to obtain each directory title and each page number interval corresponding to each directory title. Since in a directory file, there may be directory titles corresponding to different hierarchical levels, such as first-level directory titles, second-level directory titles, etc. Therefore, the server will classify each directory title hierarchically to obtain a hierarchical structure tree including nodes of different hierarchical levels, and each hierarchical node in the hierarchical structure tree corresponds to a different directory title.

[0039] Next, the server will retrieve a preset sensitive record table, which includes various sensitive characters. For example, a sensitive character may be a disciplinary record.

[0040] At this time, the server will determine the directory titles corresponding to each level node according to the hierarchical order of the hierarchical structure tree. When at least part of the content in the directory title corresponding to any level node is located in the preset sensitive record table, the server will determine the level node and all level nodes connected to the level node as having sensitive attributes.

[0041] Then, the server will determine all original files whose corresponding file page numbers are in the page number range as having sensitive attributes in the random verification group according to the page number range corresponding to each level node with sensitive attributes, and then determine all remaining original files in the random verification group as having ordinary attributes.

[0042] It can be explained that this embodiment can quickly determine the file attributes of each original file based on the comparison between the hierarchical structure tree and the preset sensitive record table, thereby improving the corresponding data processing efficiency.

[0043] Furthermore, the above-mentioned “extracting the directory file to obtain each directory title and each page number interval corresponding to each directory title” further includes the following steps: Acquire each file line in the directory file that is arranged in a vertical order from top to bottom, and determine each file line with a leading character as each directory title; Perform character recognition based on a large language model on each directory title to obtain characters of each file in the same directory title; Grouping the file characters in the same directory title by category to obtain a text grouping of the file characters corresponding to the text category and a digital grouping of the file characters corresponding to the digital category; Determine the title page number corresponding to the directory title based on each file character in the digital group, and obtain each title page number corresponding to each directory title; Determine the size of any file character in the text group respectively, and obtain the size of each title corresponding to each text group; Arrange the titles of each directory in a hierarchical manner from large to small according to the corresponding title size, and obtain the hierarchical levels corresponding to each directory title; Based on the vertical order, two catalog titles at adjacent positions and corresponding to the same hierarchical level are respectively determined as adjacent title groups, and the corresponding first catalog title in the same adjacent title group is determined as the first title, and the corresponding last catalog title is determined as the last title; Determine the title page number corresponding to the first title and the title page numbers corresponding to all the directory titles between the first title and the last title as an interval determination group, and determine the maximum page number with the largest corresponding value and the minimum page number with the smallest corresponding value based on all the title page numbers in the interval determination group; The page number interval consisting of the minimum page number and the maximum page number is determined to correspond to the first title, and each page number interval corresponding to each directory title is obtained.

[0044] For example, in this embodiment, the server will obtain each file line located in the directory file and arranged in a vertical order from top to bottom. Because under normal circumstances, the directory title and the corresponding directory page number are connected by a leading character, the server can determine all file lines with a leading character as directory titles.

[0045] Then, the server will perform character recognition based on the large language model on each directory title to obtain the file characters in the same directory title, and then group the file characters in the same directory title into categories to obtain text groups including file characters of corresponding text categories and digital groups including file characters of corresponding digital categories.

[0046] Next, the server will determine the title page number of the corresponding directory title based on each file character in the digital group, thereby obtaining the title page numbers corresponding to each directory title. Since under normal circumstances, the sizes of each file character in the same directory title will be consistent, and the title sizes corresponding to directory titles of different hierarchical levels are also different, the hierarchical levels corresponding to each directory title can be determined based on identifying the corresponding title size. Here, generally speaking, the larger the title size, the larger the corresponding hierarchical level should be; for example, when there are three different directory titles corresponding to the third chapter, the first section of the third chapter, and the second section of the third chapter, the title size of the directory title corresponding to the third chapter should be larger, while the titles corresponding to the first section of the third chapter and the second section of the third chapter should be smaller and the same; further, in order to reduce a certain amount of data processing on the server, the server will determine the size of any file character in the text group, thereby obtaining the title sizes corresponding to each text group.

[0047] Next, the server will arrange the directory titles in a descending order according to the title size, so as to obtain the hierarchical levels corresponding to the directory titles. In general directory files, the directory titles are arranged in vertical order, and the directory titles with smaller hierarchical levels are generally located behind the directory titles with larger hierarchical levels. Therefore, based on this feature, two directory titles in adjacent positions and corresponding to the same hierarchical levels can be determined as adjacent title groups according to the vertical order, and the directory title at the first position in the same adjacent title group can be determined as the first title, and the directory title at the last position can be determined as the last title.

[0048] Then, the server will determine the title page number corresponding to the first title and the title page numbers corresponding to all directory titles between the first title and the last title as an interval determination group, and determine the maximum page number with the largest corresponding value and the minimum page number with the smallest corresponding value based on the interval determination group.

[0049] Finally, the server determines the page number interval consisting of the minimum page number and the maximum page number as corresponding to the first title, thereby obtaining each page number interval corresponding to each directory title.

[0050] It can be explained that, in this embodiment, the page number interval corresponding to each directory title can be quickly acquired based on the arrangement characteristics of each directory title in the directory file, thereby improving the corresponding acquisition accuracy and data processing efficiency.

[0051] Here, the large language model (LLM) mentioned above refers to a deep learning model trained using a large amount of text data. This model can understand and generate natural language text. Since the large language model is an existing technology, this embodiment will not go into details about it.

[0052] Furthermore, the above method further comprises the following steps: Performing handwriting recognition on each original document in the random check group based on a handwriting recognition model, and in response to determining that a handwriting area exists in any original document based on the recognition result, determining that the original document has a handwriting attribute; Determine a handwriting extension direction corresponding to the handwriting area, compare the handwriting extension direction with a horizontal direction corresponding to the original document, and determine a first handwriting weight of a corresponding direction dimension based on the obtained handwriting extension direction; Determine the area ratio of the handwriting area to the original document, and determine a second handwriting weight corresponding to the ratio dimension based on the area ratio; The first handwriting weight and the second handwriting weight are summed to obtain a handwriting added weight of the original file with handwriting attributes; The extraction priority value corresponding to the original file is increased and updated based on the handwriting increase weight to obtain an updated extraction priority value.

[0053] For example, in actual application scenarios, original files generally include various file characters printed by corresponding machines. In some cases, managers with management authority over original files may add a corresponding handwriting area to at least one original file of the original file in order to record important contents in real time based on the original archives. The handwriting area may include file characters formed by handwriting, etc. Based on the occurrence of this situation, the extraction priority value of the original file with the handwriting area should be increased accordingly to improve the extraction probability of the corresponding original file. Therefore, it can be implemented based on the following method steps: First, in order to determine whether there is a corresponding handwriting area in each original file in the random verification group, handwriting recognition can be performed on each original file based on a pre-trained handwriting recognition model, and when it is determined that there is a handwriting area in any original file, the original file is determined to have a handwriting attribute; here, the handwriting recognition model can be established based on a neural network learning model or a machine learning model. Since its establishment process and training process are relatively common, this embodiment will not be described in detail; Next, it can be explained that when the angle between the handwriting extension direction corresponding to the handwriting area and the horizontal direction is smaller, it indicates that the corresponding manager may be in a more focused and concentrated state when writing. Therefore, the importance of the text characters corresponding to the handwriting area may be higher. The handwriting extension direction of the handwriting area can be compared with the horizontal direction of the corresponding original file, and the corresponding handwriting angle can be obtained based on the comparison result. Here, it can be known that when the handwriting angle is larger, the corresponding importance is smaller, and the corresponding handwriting angle is smaller, the corresponding importance is greater. Therefore, the first handwriting weight of the corresponding direction dimension can be obtained based on the following formula:

[0054] in, is the first handwriting weight, is the handwriting angle, The preset direction weight to be retrieved; Next, when the area ratio between the handwritten area and the original file is larger, it means that there are more text characters in the corresponding handwritten area. Therefore, the importance of the text characters in the handwritten area may be higher. The second handwriting weight of the corresponding ratio dimension can be determined by the area ratio of the handwritten area to the original file. Here, it can be known that when the area ratio is larger, the corresponding importance is greater, and when the corresponding area ratio is smaller, the corresponding importance is smaller. Therefore, the second handwriting weight of the corresponding ratio dimension can be obtained based on the following formula:

[0055] in, is the second handwriting weight, is the regional proportion, The preset weight to be retrieved; Finally, after obtaining the corresponding first handwriting weight and second handwriting weight, by summing the two, the extraction priority value corresponding to the original file can be increased and updated (that is, improved accordingly) based on the obtained handwriting increase weight to obtain an updated extraction priority value.

[0056] Furthermore, the above-mentioned “determining the handwriting extension direction corresponding to the handwriting area” further includes the following steps: Establishing a file coordinate system corresponding to the original file, wherein an X-axis of the file coordinate system is parallel to a horizontal direction corresponding to the original file; Performing pixel processing on the handwritten area to obtain handwritten pixel points constituting the handwritten area; Based on the file coordinate system, each handwritten coordinate point corresponding to each handwritten pixel point is obtained, and each handwritten coordinate point corresponding to the maximum horizontal coordinate value, the minimum horizontal coordinate value, the maximum vertical coordinate value, and the minimum vertical coordinate value is determined as the maximum horizontal coordinate point, the minimum horizontal coordinate point, the maximum vertical coordinate point, and the minimum vertical coordinate point; Taking the maximum horizontal coordinate point and the minimum horizontal coordinate point as starting points, respectively, a first longitudinal line segment and a second longitudinal line segment extending in the vertical direction are constructed, and taking the maximum vertical coordinate point and the minimum vertical coordinate point as starting points, respectively, a first horizontal line segment and a second horizontal line segment extending in the horizontal direction are constructed; Controlling the first longitudinal line segment, the second longitudinal line segment, the first transverse line segment and the second transverse line segment to extend to form an initial rectangle, and obtaining the corner points corresponding to the initial rectangle; Determine the center point of the rectangle corresponding to the initial rectangle, and obtain the connecting lines between each corner point and the center point of the rectangle; Taking each corner point as a starting point, establishing each corner extension line perpendicular to each point connection line, and moving each corner extension line along the corresponding point connection line toward the center point of the rectangle; In response to each corner extension line moving to contact different handwritten pixel points, each corner extension line is controlled to stop moving to form a current rectangle; Rectangular frame lines constituting the current rectangle are acquired, and a handwriting extension direction corresponding to the handwriting area is determined based on extension directions of frame lines corresponding to the rectangular frame lines.

[0057] For example, in this embodiment, the handwriting extension direction of the corresponding handwriting area can be obtained in the following manner: First, a file coordinate system corresponding to the original file can be established, and the handwritten area in the original file can be pixelated to obtain the handwritten pixel points constituting the handwritten area. Here, the X axis of the file coordinate system is parallel to the horizontal direction of the corresponding original file. Then, the server can obtain the handwritten coordinate points corresponding to each handwritten pixel point based on the file coordinate system, and then determine the handwritten coordinate points corresponding to the maximum horizontal coordinate value, the minimum horizontal coordinate value, the maximum vertical coordinate value and the minimum vertical coordinate value as the maximum horizontal coordinate point, the minimum horizontal coordinate point, the maximum vertical coordinate point and the minimum vertical coordinate point.

[0058] Next, the server will construct the first longitudinal line segment and the second longitudinal line segment extending along the vertical direction with the maximum transverse coordinate point and the minimum transverse coordinate point as starting points, and construct the first transverse line segment and the second transverse line segment extending along the horizontal direction with the maximum transverse coordinate point and the minimum transverse coordinate point as starting points.

[0059] Next, the server controls the first longitudinal line segment, the second longitudinal line segment, the first transverse line segment, and the second transverse line segment to extend to form an initial rectangle, and obtains each corner point corresponding to the initial rectangular area; Then, the server will first determine the center point of the rectangle corresponding to the initial rectangular area, and then obtain the connection lines between each corner point and the center point of the rectangle. Then, each corner extension line perpendicular to each point connection line is established with each corner point as the starting point; Then, the server will move each corner extension line along the corresponding point connection line toward the center of the rectangle. When any corner extension line touches any handwritten pixel point, the corner extension line will be controlled to stop moving. Finally, when each corner extension line moves to contact different handwriting pixel points, each corner extension line can be controlled to stop moving, thereby forming a corresponding current rectangle, by obtaining the rectangular frame lines that make up the current rectangle and determining the extension direction of the corresponding longest rectangular frame line in each rectangular frame line as the handwriting extension direction corresponding to the handwriting area.

[0060] Furthermore, for example, in this embodiment, the above method steps include updating the corresponding extraction priority value based on the original file with handwritten attributes, and in some cases, in order to further improve the corresponding file management security, when any original file is extracted multiple times in a row, the extraction priority value of the original file can be correspondingly reduced and updated to reduce the extraction probability corresponding to the original file, so as to ensure that other original files can have a higher probability of being extracted and checked, so as to increase the comprehensiveness of the extraction, and the corresponding process can be implemented based on the following method steps: Acquire each historical verification information corresponding to each original file in the random verification group, wherein each historical verification information includes each verification operation arranged in a forward order based on the verification time, and each verification operation has corresponding verification results; Performing reverse extraction based on the forward arrangement for each piece of historical verification information, and determining the verification operations obtained in a continuous arrangement and corresponding to a preset number as a continuous verification group; In response to all verification results corresponding to the same continuous verification group being verified, the preset reduction weight is called to reduce and update the extraction priority value corresponding to the original file to obtain an updated extraction priority value.

[0061] For example, in this embodiment, during random extraction, historical verification information corresponding to each original file in the random verification group can be obtained, wherein each historical verification information includes verification operations arranged in a positive order based on the verification time, and each verification operation has corresponding verification results, that is, when the original file is verified three times, the corresponding verification time also includes the corresponding three time nodes, and each original file has a verification operation corresponding to each verification time, and the verification operation can include verified and unverified. Verified can be understood as the original file being extracted and verified at the corresponding verification time, and unverified can be understood as The original file has not been extracted and verified at the corresponding verification time; by performing reverse extraction based on forward arrangement for each historical verification information, and determining the verification operations based on continuous arrangement and corresponding to a preset number as a continuous verification group, when all verification results in the same continuous verification group are verified, it indicates that the original file has been extracted and verified in the most recent corresponding preset number of verification operations. At this time, the corresponding preset pseudo-Angdi weight can be retrieved from the server to reduce and update the extraction priority value corresponding to the original file accordingly, so as to reduce the extraction probability of the original file in subsequent verification operations to increase the comprehensiveness of the extraction.

[0062] In step S104, the following contents are included: The archive verification plug-in is triggered to perform an integrity-based file verification on the current archive based on the content verification group, and determine a document status corresponding to the current archive based on the verification result, wherein the document status includes a complete status and a tampered status.

[0063] For example, in this embodiment, after the determination of the content verification group is completed, the server will trigger the pre-set archive verification plug-in to perform corresponding verification operations on the integrity of the current document based on the original files in the content verification group, thereby determining the document status corresponding to the current archive according to the verification results, wherein, when the current archive has not been tampered with, the corresponding document status is a complete status, and when the current archive has been tampered with, the corresponding document status is a tampered status.

[0064] It should be noted that, based on the foregoing content, the current file can exist in paper form and / or electronic form. In the present embodiment, the current file generally exists in paper form, but when any current operating terminal performs a corresponding viewing operation on the current file, the current operating terminal needs to perform an integrity scan on the current file, and store the corresponding electronic form of the current file in the server for corresponding archiving, so that when the next current operating terminal needs to perform a viewing operation on the current file, it can first retrieve the current file stored in the server for integrity comparison with the original document to determine whether the current file has been tampered with by the previous current operating terminal, thereby improving the corresponding file management security.

[0065] Furthermore, the above-mentioned “triggering the archive verification plug-in to perform integrity-based file verification on the current archive based on the content verification group, and determining the document status corresponding to the current archive based on the verification result, wherein the document status includes a complete status and a tampered status” further includes the following steps: Based on the archive verification plug-in, each current file having a traceability relationship with each original file in the content verification group is retrieved from the current archive, and each original file and each current file are binarized to obtain each original image and each current image, wherein any of the original images includes each first image pixel point corresponding to the first pixel value, and any of the current images includes each second image pixel point corresponding to the second pixel value; Creating a transparent fusion layer, and sequentially stacking the original file and the current file having the same traceability relationship on the transparent fusion layer; In response to the transparent fusion layer including any first image pixel point and / or any second image pixel point, the file status corresponding to the current archive is determined to be a tampered status.

[0066] For example, in this embodiment, since the current archive can be stored in the server in electronic form, the archive verification plug-in can retrieve each current file that has a traceability relationship with each original file in the content verification group from the current archive, and determine whether the current archive has been tampered with based on a one-to-one file comparison between each current file and each original file. Here, the corresponding file comparison process can be implemented based on the following method steps: First, each original file and each current file may be binarized respectively, so as to obtain corresponding original images and current pressed images, wherein any original image includes first image pixel points corresponding to the first pixel value, and any current image includes second image pixel points corresponding to the second pixel value; Next, a transparent fusion layer can be created, and the original file and the current file with the same traceability relationship can be sequentially superimposed on the transparent fusion layer; Finally, it can be explained that if the current file has not been tampered with, the corresponding transparent fusion layer should only have fused pixel values ​​corresponding to the first pixel value and the second pixel value. For example, when the first pixel value corresponds to yellow and the second pixel value is blue, the corresponding fused pixel value corresponds to green. If the current file has been tampered with, any first image pixel point and / or any second image pixel point will appear in the transparent fusion layer. Therefore, the file status of the current file can be determined as a tampered state.

[0067] In step S106, the following contents are included: In response to the current file being in a tampered state, the tampered content corresponding to the current file is determined, and based on the retrieved operation record table corresponding to the current file, an adjacent operation end for adjacently scanning the current file is determined.

[0068] For example, in this embodiment, when the document status corresponding to the current archive is a tampered state, for example, in this embodiment, based on the aforementioned content, it can be known that after the corresponding current document is viewed by the previous operating terminal, the current document needs to be scanned for integrity to be stored in the server for archiving. Therefore, when the document status corresponding to the current archive is a tampered state, the server will further determine the tampered content corresponding to the current archive, and in order to ensure the accuracy of subsequent responsibility tracing, the server will also determine the adjacent operating terminal that performs an adjacent scan (that is, a viewing operation) on the current archive based on the retrieved operation record table corresponding to the current archive.

[0069] In step S108, the following contents are included: An initial file comparison layer is established based on the current file, the tampered content and the adjacent operation terminal are respectively filled into the content indication area and the terminal indication area in the initial file comparison layer, and the obtained current file comparison layer is sent to the management terminal for display.

[0070] For example, in this embodiment, in order to record the relevant tampered content and the corresponding temporary operation terminal and inform the management end, the server will establish an initial file comparison layer based on the current file, thereby filling the tampered content and the adjacent operation terminal into the content indication area and the terminal indication area in the initial file comparison layer respectively, and send the obtained current file comparison layer to the management end for display.

[0071] It should be noted that the management end can be understood as a terminal device used by the corresponding management personnel.

[0072] Furthermore, the above-mentioned “establishing an initial archive comparison layer based on the current archive, filling the tampered content and the adjacent operation terminal into the content indication area and the terminal indication area located in the initial archive comparison layer respectively, and sending the obtained current archive comparison layer to the management terminal for display” also includes the following steps: Establishing an initial file comparison layer based on the current file, wherein the initial file comparison layer includes a vertically arranged content indication area and a terminal indication area for filling an adjacent operating terminal; Determine a current file in the current archive that contains the tampered content, obtain each file line in the current file that is arranged in a vertical order from top to bottom, and determine all file lines that contain a blank group consisting of two adjacent blank characters as paragraph start lines; Based on the vertical order, two paragraph start lines in adjacent positions are respectively determined as adjacent paragraph groups, and the corresponding first paragraph start line in the same adjacent paragraph group is determined as the first paragraph, and the corresponding last paragraph start line is determined as the last paragraph; Determine the first paragraph and all file lines between the first paragraph and the last paragraph as the same file paragraph, and obtain the file paragraphs in the current file; Determine a file paragraph including the tampered content as a tampered paragraph, and determine an original paragraph corresponding to the tampered paragraph in an original file having the same traceability relationship as the current file; Generating a first bounding box enclosing the tampered paragraph in the current file, and generating a second bounding box enclosing the original paragraph in the original file; Generate a first indication area and a second indication area arranged horizontally in the content indication area, and fill the first indication area with the current file having the first enclosing frame, and fill the second indication area with the original file having the second enclosing frame; The adjacent operating end is filled into the terminal indication area, and the obtained current file comparison layer is sent to the management end for display.

[0073] For example, in this embodiment, when it is determined that the current file is in a tampered state, the server may establish a corresponding initial file comparison layer based on the current file, where the initial file comparison layer may include a content indication area and a terminal indication area arranged vertically; since any current file is generally composed of different file paragraphs, when any current file has corresponding tampered content, the file paragraph in which it is located may be determined based on the tampered content, and the corresponding file paragraph may be further determined as a tampered paragraph, and an original paragraph corresponding to the tampered paragraph may be obtained in the original file; after determining the corresponding tampered paragraph and the original paragraph, a first enclosing frame enclosing the tampered paragraph may be generated in the current file, and a terminal indication area may be generated in the original file; A second enclosing frame enclosing a pair of original paragraphs; after completing the corresponding generation of the first enclosing frame and the second enclosing frame, a first indication area and a second indication area arranged in horizontal lines can be further generated in the content indication area located in the initial file comparison layer, and the current file with the first enclosing frame is filled into the first indication area, and the original file with the second enclosing frame is synchronously filled into the second indication area; finally, the adjacent operating end can be filled into the terminal indication area to complete the update of the initial file comparison layer, and then the obtained current file comparison layer can be sent to the management end for display, so that the management end can simultaneously obtain the corresponding tampered content and the corresponding temporary operating end, thereby improving the corresponding information traceability.

[0074] It should be noted that in order to quickly obtain each file paragraph in the current file, it can be achieved based on the following method steps: First, it can be known that there are file lines in the current file arranged in a vertical order from top to bottom. Based on common writing habits, the beginning of each file paragraph is generally composed of two adjacent blank characters. Therefore, all file lines with corresponding blank groups composed of two adjacent blank characters can be determined as the starting lines of paragraphs. Next, based on the corresponding vertical order, two paragraph start lines that are respectively in adjacent positions may be determined as an adjacent paragraph group, and further, the corresponding first paragraph start line in the same adjacent paragraph group may be determined as the first paragraph, and the corresponding last paragraph start line may be determined as the last paragraph; Finally, the first paragraph and all file lines between the first paragraph and the last paragraph can be determined as the same file paragraph, and then the acquisition of each file paragraph in the current file can be completed quickly.

[0075] According to the scheme of this embodiment, this embodiment can randomly extract original archives in response to the scanning of archive tags by the operating terminal to form a content verification group, which significantly improves the efficiency of archive verification; subsequently, this embodiment can realize efficient file verification based on integrity of the current archive through the archive verification plug-in, and quickly and accurately determine the document status (complete status or tampered status) of the archive, thereby effectively ensuring the originality and authenticity of the archive; and once the archive is detected to be in a tampered state, this embodiment can accurately identify and locate the tampered content, and quickly lock the adjacent operating terminal that is performing an adjacent scan of the current archive based on the operation record table; this double confirmation mechanism not only helps to timely discover and correct false information in the archive, but also provides a basis for subsequent accountability. It provides strong support for any traceability; finally, this embodiment can present key information such as tampered content and adjacent operation terminals in a graphical manner by establishing an initial file comparison layer, so that the management terminal can intuitively understand the true status and potential problems of the files; this visual management method not only makes it convenient for managers to quickly grasp the results of file verification, but also greatly improves the convenience and transparency of file management; this embodiment integrates a variety of technical means such as random sampling, integrity verification, large language model analysis and graphical display, and realizes efficient verification, precise positioning, intuitive display and intelligent management of files, which not only helps to improve the accuracy and efficiency of file management, but also lays a solid foundation for building a safer and more reliable file management system.

[0076] Another embodiment of the present invention provides an intelligent archive management system based on a large language model. Figure 3 As the corresponding system block diagram, the system includes: an extraction module, configured to retrieve an original file corresponding to the current file in response to any current operation terminal performing a current scan on a file tag corresponding to the current file, and randomly extract each original file included in the original file based on a preset extraction strategy, and determine each extracted original file as a content verification group; a verification module configured to trigger the archive verification plug-in to perform an integrity-based file verification on the current archive based on the content verification group, and determine a document status corresponding to the current archive based on the verification result, wherein the document status includes an intact status and a tampered status; a determination module, configured to determine the tampered content corresponding to the current file in response to the current file being in a tampered state, and determine an adjacent operation terminal for adjacent scanning of the current file based on the retrieved operation record table corresponding to the current file; The establishment module is configured to establish an initial archive comparison layer based on the current archive, fill the tampered content and the adjacent operation terminal into the content indication area and the terminal indication area located in the initial archive comparison layer respectively, and send the obtained current archive comparison layer to the management terminal for display.

[0077] In the description provided herein, algorithms and displays are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the examples of the present invention. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to implement the content of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the preferred embodiment of the present invention.

[0078] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0079] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof.

[0080] Those skilled in the art will appreciate that the modules or units or components of the devices in the examples disclosed herein may be arranged in the devices described in the embodiment, or alternatively may be located in one or more devices different from the devices in the examples. The modules in the foregoing examples may be combined into one module or may be divided into multiple submodules.

[0081] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and furthermore may be divided into a plurality of submodules or subunits or subcomponents.

[0082] Furthermore, those skilled in the art will appreciate that although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features from different embodiments is meant to be within the scope of the present invention and to form different embodiments.

[0083] In addition, some of the embodiments are described herein as methods or combinations of method elements that can be implemented by a processor of a computer system or by other devices that perform the functions. Therefore, a processor with necessary instructions for implementing the method or method elements forms a device for implementing the method or method elements. In addition, the elements described herein of the device embodiments are examples of devices for implementing the functions performed by the elements for the purpose of implementing the invention.

[0084] As used herein, unless otherwise specified, the use of ordinal numbers "first," "second," "third," etc. to describe common objects merely indicates that different instances of similar objects are involved, and is not intended to imply that the objects so described must have a given order in time, space, order, or in any other manner.

[0085] Although the present invention has been described according to a limited number of embodiments, it will be apparent to those skilled in the art, with the benefit of the above description, that other embodiments may be envisioned within the scope of the invention thus described. In addition, it should be noted that the language used in this specification is primarily selected for readability and instructional purposes, rather than for explaining or limiting the subject matter of the present invention.

Claims

1. An intelligent archive management method based on a large language model, characterized in that: The following steps are involved: In response to any current operation terminal performing a current scan on a file tag corresponding to a current file, an original file corresponding to the current file is retrieved, and original files included in the original file are randomly extracted based on a preset extraction strategy, and the extracted original files are determined as a content verification group; Triggering the archive verification plug-in to perform integrity-based file verification on the current archive based on the content verification group, and determining a document status corresponding to the current archive based on the verification result, wherein the document status includes a complete status and a tampered status; In response to the current file being in a tampered state, determining the tampered content corresponding to the current file, and determining an adjacent operation terminal for adjacent scanning of the current file based on the retrieved operation record table corresponding to the current file; An initial file comparison layer is established based on the current file, the tampered content and the adjacent operation terminal are respectively filled into the content indication area and the terminal indication area in the initial file comparison layer, and the obtained current file comparison layer is sent to the management terminal for display.

2. The intelligent archive management method based on a large language model according to claim 1 is characterized in that: The original files included in the original archive are randomly extracted based on a preset extraction strategy, and the extracted original files are determined as a content verification group, including: Determine each original document in the original file that has been pre-checked as a necessary check group, and determine all other original documents in the original file except the necessary check group as a random check group; Performing content evaluation on each original file in the random check group, and summing up the extraction priority values ​​corresponding to each original file to obtain a total extraction value; Calculate the ratio of each extraction priority value to the total extraction value to obtain the extraction probabilities corresponding to the original files; Determine the number of files corresponding to each original file in the content verification group, and calculate the product of the number of files and the preset extraction ratio to obtain the current extraction number; Creating a random drawing wheel, and dividing the random drawing wheel into regions based on the respective drawing probabilities corresponding to the respective original files, to obtain respective sector-shaped regions corresponding to the respective original files; The random drawing turntable is controlled to rotate and draw corresponding to the current drawing quantity, and based on the drawing result, each drawn original file and all original files in the necessary check group are determined as a content check group.

3. The intelligent archive management method based on a large language model according to claim 2 is characterized in that: Conducting content evaluation on each original document in the random verification group includes: Determine each original file in the random check group based on the file attribute, and determine each original file with sensitive attributes as a sensitive file group, and determine each original file with common attributes as a common file group; Retrieving the first attribute evaluation value to respectively assign scores to the original files in the sensitive file group, and retrieving the second attribute evaluation value to respectively assign scores to the original files in the common file group, wherein the first attribute evaluation value is greater than the second attribute evaluation value; Perform pixel identification on each original file in the random check group to obtain pixel points of each file in the same original file, and obtain the number of pixels corresponding to each file pixel point to obtain the number of pixels corresponding to each original file; Compare the number of pixels with a first number threshold and a second number threshold respectively, wherein the first number threshold is less than the second number threshold; In response to the number of pixels corresponding to any original file being less than the first number threshold, calling the first number evaluation value to perform score configuration on the original file based on the number dimension; In response to the number of pixels corresponding to any original file being greater than or equal to the first number threshold and less than the second number threshold, calling the second number evaluation value to perform score configuration based on the number dimension on the original file; In response to the number of pixels corresponding to any original file being greater than or equal to the second number threshold, calling a third quantity evaluation value to perform score configuration on the original file based on the quantity dimension; The attribute evaluation value and the quantity evaluation value corresponding to the same original file are summed up to obtain the extraction priority values ​​corresponding to each original file.

4. The intelligent archive management method based on a large language model according to claim 3 is characterized in that: Determining each original file in the random check group based on the file attribute, and determining each original file with sensitive attributes as a sensitive file group, and determining each original file with common attributes as a common file group, includes: Obtaining the page numbers of the files respectively corresponding to the original files, and determining the original files having the same page number as the retrieved preset directory page number as the directory files in the original files; Extracting the directory file to obtain the directory titles and the page number intervals corresponding to the directory titles; Classifying each directory title hierarchically to obtain a hierarchical structure tree including nodes of different hierarchical levels, wherein each hierarchical node corresponds to a different directory title; Retrieving a preset sensitive record table, wherein the preset sensitive record table includes various sensitive characters; Determining the directory titles corresponding to the level nodes based on the level order of the level structure tree, and in response to at least a portion of the directory title corresponding to any level node being located in the preset sensitive record table, determining the level node and all level nodes connected downward to the level node as having sensitive attributes; Based on the page number intervals corresponding to each hierarchical node with sensitive attributes, all original files whose corresponding file page numbers are located in the page number intervals in the random verification group are determined to have sensitive attributes, and all remaining original files in the random verification group are determined to have ordinary attributes.

5. The intelligent archive management method based on a large language model according to claim 4 is characterized in that: Perform directory extraction on the directory file to obtain each directory title and each page number interval corresponding to each directory title, including: Acquire each file line in the directory file that is arranged in a vertical order from top to bottom, and determine each file line with a leading character as each directory title; Perform character recognition based on a large language model on each directory title to obtain characters of each file in the same directory title; Grouping the file characters in the same directory title by category to obtain a text grouping of the file characters corresponding to the text category and a digital grouping of the file characters corresponding to the digital category; Determine the title page number corresponding to the directory title based on each file character in the digital group, and obtain each title page number corresponding to each directory title; Determine the size of any file character in the text group respectively, and obtain the size of each title corresponding to each text group; Arrange the titles of each directory in a hierarchical manner from large to small according to the corresponding title size, and obtain the hierarchical levels corresponding to each directory title; Based on the vertical order, two catalog titles at adjacent positions and corresponding to the same hierarchical level are respectively determined as adjacent title groups, and the corresponding first catalog title in the same adjacent title group is determined as the first title, and the corresponding last catalog title is determined as the last title; Determine the title page number corresponding to the first title and the title page numbers corresponding to all the directory titles between the first title and the last title as an interval determination group, and determine the maximum page number with the largest corresponding value and the minimum page number with the smallest corresponding value based on all the title page numbers in the interval determination group; The page number interval consisting of the minimum page number and the maximum page number is determined to correspond to the first title, and each page number interval corresponding to each directory title is obtained.

6. The intelligent archive management method based on a large language model according to claim 3 is characterized in that: The method further comprises: Performing handwriting recognition on each original document in the random check group based on a handwriting recognition model, and in response to determining that a handwriting area exists in any original document based on the recognition result, determining that the original document has a handwriting attribute; Determine a handwriting extension direction corresponding to the handwriting area, compare the handwriting extension direction with a horizontal direction corresponding to the original document, and determine a first handwriting weight of a corresponding direction dimension based on the obtained handwriting extension direction; Determine the area ratio of the handwriting area to the original document, and determine a second handwriting weight corresponding to the ratio dimension based on the area ratio; The first handwriting weight and the second handwriting weight are summed to obtain a handwriting added weight of the original file with handwriting attributes; The extraction priority value corresponding to the original file is increased and updated based on the handwriting increase weight to obtain an updated extraction priority value.

7. The intelligent archive management method based on a large language model according to claim 3 is characterized in that: The method further comprises: Acquire each historical verification information corresponding to each original file in the random verification group, wherein each historical verification information includes each verification operation arranged in a forward order based on the verification time, and each verification operation has corresponding verification results; Performing reverse extraction based on the forward arrangement for each piece of historical verification information, and determining the verification operations obtained in a continuous arrangement and corresponding to a preset number as a continuous verification group; In response to all verification results corresponding to the same continuous verification group being verified, the preset reduction weight is called to reduce and update the extraction priority value corresponding to the original file to obtain an updated extraction priority value.

8. The intelligent archive management method based on a large language model according to claim 1, characterized in that: The archive verification plug-in is triggered to perform an integrity-based file verification on the current archive based on the content verification group, and determine a document status corresponding to the current archive based on the verification result, wherein the document status includes an intact status and a tampered status, including: Based on the archive verification plug-in, each current file having a traceability relationship with each original file in the content verification group is retrieved from the current archive, and each original file and each current file are binarized to obtain each original image and each current image, wherein any of the original images includes each first image pixel point corresponding to the first pixel value, and any of the current images includes each second image pixel point corresponding to the second pixel value; Creating a transparent fusion layer, and sequentially stacking the original file and the current file having the same traceability relationship on the transparent fusion layer; In response to the transparent fusion layer including any first image pixel point and / or any second image pixel point, the file status corresponding to the current archive is determined to be a tampered status.

9. The intelligent archive management method based on a large language model according to claim 8, characterized in that: The initial file comparison layer is established based on the current file, the tampered content and the adjacent operation terminal are respectively filled into the content indication area and the terminal indication area located in the initial file comparison layer, and the obtained current file comparison layer is sent to the management terminal for display, including: Establishing an initial file comparison layer based on the current file, wherein the initial file comparison layer includes a vertically arranged content indication area and a terminal indication area for filling an adjacent operating terminal; Determine a current file in the current archive that contains the tampered content, obtain each file line in the current file that is arranged in a vertical order from top to bottom, and determine all file lines that contain a blank group consisting of two adjacent blank characters as paragraph start lines; Based on the vertical order, two paragraph start lines in adjacent positions are respectively determined as adjacent paragraph groups, and the corresponding first paragraph start line in the same adjacent paragraph group is determined as the first paragraph, and the corresponding last paragraph start line is determined as the last paragraph; Determine the first paragraph and all file lines between the first paragraph and the last paragraph as the same file paragraph, and obtain the file paragraphs in the current file; Determine a file paragraph including the tampered content as a tampered paragraph, and determine an original paragraph corresponding to the tampered paragraph in an original file having the same traceability relationship as the current file; Generating a first bounding box enclosing the tampered paragraph in the current file, and generating a second bounding box enclosing the original paragraph in the original file; Generate a first indication area and a second indication area arranged horizontally in the content indication area, and fill the first indication area with the current file having the first enclosing frame, and fill the second indication area with the original file having the second enclosing frame; The adjacent operating end is filled into the terminal indication area, and the obtained current file comparison layer is sent to the management end for display.

10. An intelligent archive management system based on a large language model, characterized in that: include: an extraction module, configured to retrieve an original file corresponding to the current file in response to any current operation terminal performing a current scan on a file tag corresponding to the current file, and randomly extract each original file included in the original file based on a preset extraction strategy, and determine each extracted original file as a content verification group; a verification module configured to trigger the archive verification plug-in to perform an integrity-based file verification on the current archive based on the content verification group, and determine a document status corresponding to the current archive based on the verification result, wherein the document status includes an intact status and a tampered status; a determination module, configured to determine the tampered content corresponding to the current file in response to the current file being in a tampered state, and determine an adjacent operation terminal for adjacent scanning of the current file based on the retrieved operation record table corresponding to the current file; The establishment module is configured to establish an initial archive comparison layer based on the current archive, fill the tampered content and the adjacent operation terminal into the content indication area and the terminal indication area located in the initial archive comparison layer respectively, and send the obtained current archive comparison layer to the management terminal for display.

Images