Archive management system and method based on artificial intelligence
Through an archive management system based on artificial intelligence, automatic classification and feature extraction, identification of defect types and dissemination paths, the problems of low efficiency of traditional archive management and insufficient information sharing are solved, and efficient information management and scientific decision-making are achieved.
Patent Information
- Application Number
- CN202411382693.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-30
- Publication Date
- 2025-01-10
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional archive management systems rely on manual operations, are inefficient and susceptible to human factors, lack effective information sharing and in-depth analysis capabilities, resulting in the inability to effectively flow and utilize information, and corporate decision-making relies on subjective judgment, which can easily lead to mistakes.
Design an archive management system based on artificial intelligence, including feature extraction unit, archive diagnosis unit, node matching unit, topology modeling unit and defect evaluation unit. Through automatic classification, feature extraction and topology modeling, we can identify defect types and propagation paths. , generate weighted defect propagation diagram, dynamically mark the level, and provide visualization tools.
It improves the efficiency of archive management, quickly identify and correct defects, systematically displays information flow, identify potential risks, provides scientific decision-making support, and improves the level of enterprise management.
Smart Images
Figure CN119292994A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of archive management, and in particular to an archive management system and method based on artificial intelligence. Background Art
[0002] Electronic records are files, records or data kept in electronic form, usually used for storing and managing information.
[0003] Traditional archive management relies on manual operations, including the organization, classification, storage and search of paper documents. This process is not only time-consuming, but also easily affected by human factors. Duplication or omission of tasks may occur during manual processing, further reducing work efficiency. In traditional systems, information between different departments is often isolated. Due to the lack of an effective sharing mechanism, each department may have its own independent archive management process, which prevents information from flowing smoothly between departments. Traditional systems usually only focus on the storage and management of archives, but lack the ability to deeply analyze data, which makes it difficult for companies to extract valuable information from historical data and cannot identify trends and patterns. The lack of real-time analysis makes companies often rely on subjective judgment when formulating strategies, lacking data support, which may lead to decision-making errors. Summary of the invention
[0004] The purpose of the present invention is to propose an archive management system and method based on artificial intelligence in view of the problems existing in the background technology.
[0005] The technical solution of the present invention is: an artificial intelligence-based archive management system, comprising a feature extraction unit, an archive interface unit, an archive diagnosis unit, a node matching unit, a topology modeling unit, a defect assessment unit and a defect annotation unit, and also comprising:
[0006] A file acquisition unit, which is used to acquire electronic files uploaded by multiple positions in the target enterprise to obtain an electronic file set, and transmit the electronic file set to the file processing unit;
[0007] A file processing unit, which receives the electronic file set transmitted by the file acquisition unit, converts the electronic file set into an electronic file text, performs a preprocessing operation on the electronic file text to obtain a standard electronic file text, and transmits the standard electronic file text to the file classification unit;
[0008] An archive classification unit receives the standard electronic archive text transmitted by the archive processing unit, and classifies the standard electronic archive text into different types of electronic archive collections, wherein the different types of electronic archive collections include personnel archive collections, financial archive collections, contract archive collections and project archive collections, and transmits the different types of electronic archive collections to a feature extraction unit and an archive interface unit.
[0009] Preferably, the feature extraction unit receives the different types of electronic archive collections transmitted by the archive classification unit, and performs feature keyword segmentation on each electronic archive in the different types of electronic archive collections to obtain a feature keyword segmentation sequence corresponding to the electronic archive, obtains the number of word co-occurrences of the feature keyword segmentation sequence based on a word co-occurrence window, constructs a word frequency bigram based on the feature keyword segments in the feature keyword segmentation sequence and the number of word co-occurrences corresponding to the feature keyword segmentations, constructs a word frequency vector sequence of the feature keyword segmentation sequence based on the word frequency bigram, obtains the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections based on the word frequency vector sequence of the feature keyword segmentation sequence, and transmits the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections to the archive diagnosis unit.
[0010] Preferably, the archive diagnosis unit receives the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive sets transmitted by the feature extraction unit, and inputs the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive sets into the trained archive diagnosis model, outputs the defect type corresponding to each electronic archive in the different types of electronic archive sets through the trained defect diagnosis model, and transmits the defect type corresponding to each electronic archive in the different types of electronic archive sets to the node matching unit.
[0011] Preferably, the topology modeling unit is used to perform topology modeling based on the flow of electronic files between multiple positions in the target enterprise to obtain a position topology map corresponding to the multiple positions in the target enterprise, and transmit the position topology map corresponding to the multiple positions in the target enterprise to the node matching unit.
[0012] Preferably, the node matching unit receives the defect types corresponding to the multiple electronic archives in the different types of electronic archive collections transmitted by the archive diagnosis unit and the position topology map corresponding to the multiple positions in the target enterprise transmitted by the topology modeling unit, and performs node matching on the defect types corresponding to the multiple electronic archives in the different types of electronic archive collections and the position topology map corresponding to the multiple positions in the target enterprise to obtain the defect propagation map corresponding to the multiple positions in the target enterprise, and transmits the defect propagation map corresponding to the multiple positions in the target enterprise to the defect assessment unit.
[0013] Preferably, the defect assessment unit receives the defect propagation graph corresponding to multiple positions in the target enterprise transmitted by the node matching unit, and calculates the similarity of each position topology node in the position topology graph corresponding to the multiple positions in the target enterprise, calculates the connection edge weights in the defect propagation graph corresponding to the multiple positions in the target enterprise based on the similarity of each position topology node, calculates the node weights in the defect propagation graph corresponding to the multiple positions in the target enterprise based on the node importance corresponding to the position topology graph corresponding to the multiple positions in the target enterprise, obtains a weighted defect propagation graph based on the connection edge weights and node weights in the defect propagation graph, performs defect assessment on the weighted defect propagation graph to obtain the defect level corresponding to the weighted defect propagation graph, and transmits the defect level corresponding to the weighted defect propagation graph to the defect labeling unit.
[0014] Preferably, the defect labeling unit receives the defect level corresponding to the weighted defect propagation graph transmitted by the defect assessment unit, and labels the defect level of the corresponding area in the position topology graph corresponding to multiple positions in the target enterprise based on the defect level corresponding to the weighted defect propagation graph, so as to obtain a dynamic position topology graph corresponding to multiple positions in the target enterprise, and transmits the dynamic position topology graph corresponding to multiple positions in the target enterprise to the archive interface unit.
[0015] Preferably, the archive interface unit receives the different types of electronic archive collections transmitted by the archive classification unit and the dynamic position topology map corresponding to multiple positions in the target enterprise transmitted by the defect marking unit, and marks the different types of electronic archive collections to the corresponding position topology nodes in the dynamic position topology map corresponding to multiple positions in the target enterprise, and displays the different types of electronic archive collections when the corresponding position topology node in the dynamic position topology map corresponding to multiple positions in the target enterprise is clicked.
[0016] Preferably, node matching is performed on the position topology graph corresponding to the multiple positions in the target enterprise by using the defect types corresponding to the multiple electronic archives in the different types of electronic archive sets to obtain the defect propagation graph corresponding to the multiple positions in the target enterprise, including:
[0017] Matching each electronic file in the different types of electronic file collections with a preset node keyword list to obtain a position topology node corresponding to each electronic file in the different types of electronic file collections;
[0018] Based on the defect type corresponding to each electronic file in the different types of electronic file sets, candidate abnormal node screening is performed on the position topology nodes in the position topology graph corresponding to the multiple positions in the target enterprise to obtain a candidate abnormal node set of the position topology graph corresponding to the multiple positions in the target enterprise;
[0019] Defect propagation graphs of multiple positions in the target enterprise are extracted based on the candidate abnormal node set and the connection edges between the candidate abnormal nodes.
[0020] Preferably, the standard electronic archive text is classified into different types of electronic archive collections, including:
[0021] Segmenting the standard electronic archive text to obtain an archive segment set corresponding to the standard electronic archive text;
[0022] Perform keyword matching on the archive segment set and a preset keyword list, and if an archive segment in the archive segment set matches the preset keyword list, obtain the keyword corresponding to the archive segment in the archive segment set;
[0023] If the archive segment in the archive segment set does not match the preset keyword list, semantically matching the unmatched archive segment in the archive segment set with the preset keyword list based on a word similarity model to obtain keywords corresponding to the unmatched archive segment in the archive segment set;
[0024] Combining the keywords corresponding to the various archive segments in the archive segment set to obtain a keyword set corresponding to the archive segment set;
[0025] The archive segments are classified into different types of log sets based on the keyword set.
[0026] Preferably, the calculation formula of the similarity of the position topology nodes is as follows:
[0027]
[0028] Among them, wvi,vj represents the similarity between any two adjacent nodes vi and vj, α represents the preset weight coefficient, and x vi and x vj represents the feature vector of two adjacent nodes vi and vj, n represents the dimension of the feature vector, x vi,k and x vj,k Represents the kth dimension of the feature vector of two adjacent nodes vi and vj;
[0029] The calculation formula of the node importance corresponding to the position topology node is as follows:
[0030]
[0031] Among them, w vi Indicates the node importance corresponding to the position topology node, P(x vi ) represents the Gaussian kernel probability density estimate of node vi.
[0032] Preferably, obtaining a weighted defect propagation graph based on the connection edge weights and node weights in the defect propagation graph, and performing defect assessment on the weighted defect propagation graph to obtain a defect level corresponding to the weighted defect propagation graph includes:
[0033] The transition probability matrix is calculated based on the edge weights of the weighted defect propagation graph, and the transition probability matrix is as follows:
[0034]
[0035] Among them, P vi,vj represents the probability of node vi transferring to the adjacent node vj;
[0036] The node weights of the iteratively updated post topology nodes of the weighted defect propagation graph are calculated based on the transition probability matrix. The calculation formula of the node weights of the iteratively updated post topology nodes is as follows:
[0037]
[0038] in, represents the node weight of the position topology node after iterative update, q vi,vj Indicates that each iteration is 1-q vi,vj The probability of jumping back to a random position topological node and taking q vi,vj The probability of continues to move along the connecting edges in the weighted defect propagation graph;
[0039] The final node weights of the job topology nodes in the weighted defect propagation graph are summed to obtain the sum of the final node weights of the weighted defect propagation graph, and the final node weights of the weighted defect propagation graph are matched with the preset weight threshold level table to obtain the defect level corresponding to the weighted defect propagation graph.
[0040] The technical solution of the present invention is: an artificial intelligence-based archive management method, which is applicable to the artificial intelligence-based archive management system, comprising:
[0041] Acquire electronic archives uploaded by multiple positions in the target enterprise to obtain an electronic archive set, convert the electronic archive set into an electronic archive text, and perform preprocessing operations on the electronic archive text to obtain a standard electronic archive text;
[0042] Classifying the standard electronic archive text into different types of electronic archive sets, and performing feature extraction on each electronic archive in the different types of electronic archive sets to obtain a word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive sets;
[0043] Inputting the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive sets into the trained archive diagnosis model, and outputting the defect type corresponding to each electronic archive in the different types of electronic archive sets through the trained defect diagnosis model;
[0044] Performing topological modeling based on the electronic archive flow between multiple positions in the target enterprise to obtain a position topological graph corresponding to multiple positions in the target enterprise, performing node matching on the position topological graph corresponding to multiple positions in the target enterprise by using the defect types corresponding to multiple electronic archives in the different types of electronic archive sets to obtain a defect propagation graph corresponding to multiple positions in the target enterprise;
[0045] Performing defect assessment based on the defect propagation graphs corresponding to the multiple positions in the target enterprise to obtain the defect level corresponding to the weighted defect propagation graph, and marking the defect level of the corresponding areas in the position topology graph corresponding to the multiple positions in the target enterprise based on the defect level corresponding to the weighted defect propagation graph to obtain the dynamic position topology graph corresponding to the multiple positions in the target enterprise;
[0046] The electronic archives of different types are set and marked to corresponding position topology nodes in a dynamic position topology map corresponding to a plurality of positions in the target enterprise, and the dynamic position topology map is displayed.
[0047] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:
[0048] 1. The present invention reduces manual processing time and improves the efficiency of file management through automatic classification and feature extraction of electronic files. By analyzing electronic files using trained models, the defect types in the files can be quickly and accurately identified, which helps to take corrective measures in a timely manner. Through topological modeling, the system can intuitively display the relationship between various positions in the enterprise and the flow of files, so that managers can understand the flow of information more clearly, and can identify the propagation path of defects between different positions, provide targeted improvement suggestions for the enterprise, and reduce potential risks.
[0049] 2. By calculating the similarity between positions and the importance of nodes, the system of the present invention can generate a weighted defect propagation diagram and dynamically mark the defect levels, which is convenient for managers to quickly locate and handle problems. The design of the archive interface unit allows users to easily access relevant archives and directly obtain the required information by clicking on nodes, thereby improving user experience. The analysis results and visualization tools provided will help management make more scientific and effective decisions and improve the overall management level of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] Figure 1 A schematic diagram of the system architecture of the overall system in an embodiment of the present invention;
[0051] Figure 2 A schematic flow chart of the steps of an overall method in an embodiment of the present invention.
[0052] Figure numerals: 1. Archive acquisition unit; 2. Archive processing unit; 3. Archive classification unit; 4. Feature extraction unit; 5. Archive interface unit; 6. Archive diagnosis unit; 7. Node matching unit; 8. Topology modeling unit; 9. Defect assessment unit; 10. Defect annotation unit. DETAILED DESCRIPTION
[0053] Embodiment 1, as Figure 1 As shown, the present invention proposes an artificial intelligence-based archive management system, including a feature extraction unit 4, an archive interface unit 5, an archive diagnosis unit 6, a node matching unit 7, a topology modeling unit 8, a defect assessment unit 9 and a defect annotation unit 10, and also includes:
[0054] The file acquisition unit 1 is used to acquire electronic files uploaded by multiple positions in the target enterprise to obtain an electronic file set, and transmit the electronic file set to the file processing unit 2;
[0055] The archive processing unit 2 receives the electronic archive set transmitted by the archive acquisition unit 1, converts the electronic archive set into an electronic archive text, performs preprocessing operations on the electronic archive text to obtain a standard electronic archive text, and transmits the standard electronic archive text to the archive classification unit 3;
[0056] The archive classification unit 3 receives the standard electronic archive text transmitted by the archive processing unit 2, and classifies the standard electronic archive text into different types of electronic archive collections, wherein the different types of electronic archive collections include personnel archive collections, financial archive collections, contract archive collections and project archive collections, and transmits the different types of electronic archive collections to the feature extraction unit 4 and the archive interface unit 5.
[0057] In an optional embodiment, the feature extraction unit 4 receives different types of electronic archive collections transmitted by the archive classification unit 3, and performs feature keyword segmentation on each electronic archive in the different types of electronic archive collections to obtain a feature keyword segmentation sequence corresponding to the electronic archive, obtains the number of word co-occurrences of the feature keyword segmentation sequence based on the word co-occurrence window, constructs a word frequency bigram based on the feature keyword segments in the feature keyword segmentation sequence and the number of word co-occurrences corresponding to the feature keyword segmentations, constructs a word frequency vector sequence of the feature keyword segmentation sequence based on the word frequency bigram, obtains the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections based on the word frequency vector sequence of the feature keyword segmentation sequence, and transmits the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections to the archive diagnosis unit 6.
[0058] It should be noted that feature keyword segmentation refers to processing the document content and breaking down long text into key vocabulary units (keywords) for subsequent analysis, which usually involves natural language processing (NLP) technology; word co-occurrence window represents words that appear simultaneously within a certain range (a specified number of words). It is used to calculate the relationship between them and the frequency of their co-occurrence; word co-occurrence count refers to the number of times two words appear simultaneously within a given word co-occurrence window. This indicator can reflect the degree of association between words; word frequency bigrams are pairs of two elements, used to represent the combination of feature keywords and their frequency of occurrence, and are usually used to construct feature vectors; word frequency vector sequences are the process of converting keywords and their corresponding frequencies in each electronic file into numerical vectors, which is convenient for subsequent machine learning or data analysis.
[0059] In an optional embodiment, the archive diagnosis unit 6 receives the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections transmitted by the feature extraction unit 4, and inputs the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections into the trained archive diagnosis model, outputs the defect type corresponding to each electronic archive in the different types of electronic archive collections through the trained defect diagnosis model, and transmits the defect type corresponding to each electronic archive in the different types of electronic archive collections to the node matching unit 7.
[0060] It should be noted that the archival diagnosis model adopts an improved support vector machine model, and optimizes the kernel function and penalty coefficient of the support vector machine model based on an optimization algorithm to obtain the optimal combination of kernel function and penalty coefficient, and improves the support vector machine model based on the optimal combination of kernel function and penalty coefficient to obtain an improved support vector machine model.
[0061] In an optional embodiment, the topology modeling unit 8 is used to perform topology modeling based on the flow of electronic files between multiple positions in the target enterprise to obtain a position topology map corresponding to multiple positions in the target enterprise, and transmit the position topology map corresponding to multiple positions in the target enterprise to the node matching unit 7.
[0062] It should be noted that flow refers to the transmission path and direction of electronic files between different positions, reflecting how information flows within the enterprise; topological modeling is a mathematical and computer science method used to represent the relationships and connections between elements in a system. Here, it refers to building a model to show the flow structure of electronic files between different positions; the position topology diagram is a graphical representation that shows the various positions, the relationships between them, and the paths of the flow of electronic files. Nodes usually represent positions, while edges represent the direction and intensity of information flow.
[0063] In an optional embodiment, the node matching unit 7 receives the defect types corresponding to multiple electronic files in different types of electronic file collections transmitted by the file diagnosis unit 6 and the position topology map corresponding to multiple positions in the target enterprise transmitted by the topology modeling unit 8, and performs node matching on the defect types corresponding to multiple electronic files in different types of electronic file collections and the position topology map corresponding to multiple positions in the target enterprise to obtain a defect propagation map corresponding to multiple positions in the target enterprise, and transmits the defect propagation map corresponding to multiple positions in the target enterprise to the defect assessment unit 9.
[0064] In an optional embodiment, the defect assessment unit 9 receives the defect propagation graph corresponding to multiple positions in the target enterprise transmitted by the node matching unit 7, and calculates the similarity of each position topology node in the position topology graph corresponding to the multiple positions in the target enterprise, calculates the connection edge weights in the defect propagation graph corresponding to the multiple positions in the target enterprise based on the similarity of each position topology node, calculates the node weights in the defect propagation graph corresponding to the multiple positions in the target enterprise based on the node importance corresponding to the position topology graph corresponding to the multiple positions in the target enterprise, obtains a weighted defect propagation graph based on the connection edge weights and node weights in the defect propagation graph, performs defect assessment on the weighted defect propagation graph to obtain the defect level corresponding to the weighted defect propagation graph, and transmits the defect level corresponding to the weighted defect propagation graph to the defect labeling unit 10.
[0065] In an optional embodiment, the defect labeling unit 10 receives the defect level corresponding to the weighted defect propagation graph transmitted by the defect assessment unit 9, and labels the defect level of the corresponding area in the position topology graph corresponding to multiple positions in the target enterprise based on the defect level corresponding to the weighted defect propagation graph, so as to obtain a dynamic position topology graph corresponding to multiple positions in the target enterprise, and transmits the dynamic position topology graph corresponding to multiple positions in the target enterprise to the archive interface unit 5.
[0066] In an optional embodiment, the archive interface unit 5 receives different types of electronic archive collections transmitted by the archive classification unit 3 and the dynamic position topology map corresponding to multiple positions in the target enterprise transmitted by the defect marking unit 10, and marks the different types of electronic archive collections to the corresponding position topology nodes in the dynamic position topology map corresponding to multiple positions in the target enterprise, and displays different types of electronic archive collections when clicking the corresponding position topology node in the dynamic position topology map corresponding to multiple positions in the target enterprise.
[0067] Embodiment 2, an artificial intelligence-based archive management system proposed by the present invention, compared with embodiment 1, this embodiment further includes: node matching of defect types corresponding to multiple electronic archives in different types of electronic archive sets with position topology graphs corresponding to multiple positions in a target enterprise, so as to obtain defect propagation graphs corresponding to multiple positions in the target enterprise, including:
[0068] Matching each electronic file in different types of electronic file collections with a preset node keyword list to obtain a position topology node corresponding to each electronic file in different types of electronic file collections;
[0069] Based on the defect types corresponding to each electronic file in different types of electronic file sets, candidate abnormal node screening is performed on the position topology nodes in the position topology graph corresponding to multiple positions in the target enterprise to obtain a candidate abnormal node set of the position topology graph corresponding to multiple positions in the target enterprise;
[0070] The defect propagation graphs of multiple positions in the target enterprise are extracted based on the candidate abnormal node set and the connecting edges between the candidate abnormal nodes.
[0071] In an optional embodiment, the standard electronic archive texts are classified into different types of electronic archive collections, including:
[0072] Segmenting the standard electronic archive text to obtain an archive segment set corresponding to the standard electronic archive text;
[0073] Perform keyword matching on the archive segment set and the preset keyword list. If the archive segment in the archive segment set matches the preset keyword list, the keyword corresponding to the archive segment in the archive segment set is obtained.
[0074] If the archive segment in the archive segment set does not match the preset keyword list, the unmatched archive segment in the archive segment set is semantically matched with the preset keyword list based on the word similarity model to obtain the keyword corresponding to the unmatched archive segment in the archive segment set;
[0075] Combining the keywords corresponding to the various archive segments in the archive segment set to obtain a keyword set corresponding to the archive segment set;
[0076] The archive segments are classified into different types of log sets based on keyword sets.
[0077] In an optional embodiment, the calculation formula of the similarity of the position topology nodes is as follows:
[0078]
[0079] Among them, w vi,vj represents the similarity between any two adjacent nodes vi and vj, α represents the preset weight coefficient, and x vi and x vj represents the feature vector of two adjacent nodes vi and vj, n represents the dimension of the feature vector, x vi,k and x vj,k Represents the kth dimension of the feature vector of two adjacent nodes vi and vj;
[0080] The calculation formula for the node importance corresponding to the position topology node is as follows:
[0081]
[0082] Among them, w vi Indicates the node importance corresponding to the position topology node, P(x vi ) represents the Gaussian kernel probability density estimate of node vi.
[0083] In an optional embodiment, a weighted defect propagation graph is obtained based on the connection edge weights and node weights in the defect propagation graph, and a defect assessment is performed on the weighted defect propagation graph to obtain a defect level corresponding to the weighted defect propagation graph, including:
[0084] The transition probability matrix is calculated based on the edge weights of the weighted defect propagation graph. The transition probability matrix is as follows:
[0085]
[0086] Among them, P vi,vj represents the probability of node vi transferring to the adjacent node vj;
[0087] The node weights of the iteratively updated position topology nodes of the weighted defect propagation graph are calculated based on the transition probability matrix. The calculation formula of the node weights of the iteratively updated position topology nodes is as follows:
[0088]
[0089] in, represents the node weight of the position topology node after iterative update, q vi,vj Indicates that each iteration is 1-q vi,vj The probability of jumping back to a random position topological node and taking q vi,vj The probability of continues to move along the connecting edges in the weighted defect propagation graph;
[0090] The final node weights of the job topology nodes in the weighted defect propagation graph are summed to obtain the final node weight sum of the weighted defect propagation graph, and the final node weight sum of the weighted defect propagation graph is matched with the preset weight threshold level table to obtain the defect level corresponding to the weighted defect propagation graph.
[0091] Embodiment three, as Figure 2 As shown, an artificial intelligence-based archive management method, which is applicable to the above-mentioned artificial intelligence-based archive management system, includes:
[0092] S1. Obtain electronic files uploaded by multiple positions in the target enterprise to obtain an electronic file set, convert the electronic file set into an electronic file text, and perform preprocessing operations on the electronic file text to obtain a standard electronic file text;
[0093] S2, classifying the standard electronic archive text into different types of electronic archive collections, and performing feature extraction on each electronic archive in the different types of electronic archive collections to obtain a word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections;
[0094] S3, inputting the word frequency vector sequence corresponding to each electronic file in the different types of electronic file sets into the trained file diagnosis model, and outputting the defect type corresponding to each electronic file in the different types of electronic file sets through the trained defect diagnosis model;
[0095] S4. Perform topological modeling based on the electronic file flow between multiple positions in the target enterprise to obtain a position topology map corresponding to multiple positions in the target enterprise, and perform node matching on the position topology map corresponding to multiple positions in the target enterprise by matching the defect types corresponding to multiple electronic files in different types of electronic file sets to obtain a defect propagation map corresponding to multiple positions in the target enterprise;
[0096] S5. Based on the defect propagation diagrams corresponding to the multiple positions in the target enterprise, defect assessment is performed to obtain defect levels corresponding to the weighted defect propagation diagrams, and based on the defect levels corresponding to the weighted defect propagation diagrams, defect levels are marked on corresponding areas in the position topology diagram corresponding to the multiple positions in the target enterprise to obtain dynamic position topology diagrams corresponding to the multiple positions in the target enterprise;
[0097] S6. Electronic archive collections of different types are marked to corresponding position topology nodes in a dynamic position topology map corresponding to multiple positions in the target enterprise, and the dynamic position topology map is displayed.
[0098] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.
Claims
1. An artificial intelligence-based archive management system, comprising a feature extraction unit (4), an archive interface unit (5), an archive diagnosis unit (6), a node matching unit (7), a topology modeling unit (8), a defect assessment unit (9) and a defect annotation unit (10), characterized in that: A file acquisition unit (1), the file acquisition unit (1) is used to acquire electronic files uploaded by multiple positions in the target enterprise to obtain an electronic file set, and transmit the electronic file set to the file processing unit (2); A file processing unit (2), the file processing unit (2) receives the electronic file set transmitted by the file acquisition unit (1), converts the electronic file set into an electronic file text, performs a preprocessing operation on the electronic file text to obtain a standard electronic file text, and transmits the standard electronic file text to the file classification unit (3); An archive classification unit (3), wherein the archive classification unit (3) receives the standard electronic archive text transmitted by the archive processing unit (2), and classifies the standard electronic archive text into different types of electronic archive collections, wherein the different types of electronic archive collections include personnel archive collections, financial archive collections, contract archive collections, and project archive collections, and transmits the different types of electronic archive collections to a feature extraction unit (4) and an archive interface unit (5).
2. The artificial intelligence-based archive management system according to claim 1, characterized in that: The feature extraction unit (4) receives the different types of electronic archive collections transmitted by the archive classification unit (3), and performs feature keyword segmentation on each electronic archive in the different types of electronic archive collections to obtain a feature keyword segmentation sequence corresponding to the electronic archive, obtains the number of word co-occurrences of the feature keyword segmentation sequence based on a word co-occurrence window, constructs a word frequency binary based on the feature keyword segmentation in the feature keyword segmentation sequence and the number of word co-occurrences corresponding to the feature keyword segmentation, constructs a word frequency vector sequence of the feature keyword segmentation sequence based on the word frequency binary, obtains a word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections based on the word frequency vector sequence of the feature keyword segmentation sequence, and transmits the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections to the archive diagnosis unit (6); The archive diagnosis unit (6) receives the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections transmitted by the feature extraction unit (4), and inputs the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive collections into the trained archive diagnosis model, outputs the defect type corresponding to each electronic archive in the different types of electronic archive collections through the trained defect diagnosis model, and transmits the defect type corresponding to each electronic archive in the different types of electronic archive collections to the node matching unit (7).
3. The artificial intelligence-based archive management system according to claim 2, characterized in that: The topology modeling unit (8) is used to perform topology modeling based on the electronic file flow between multiple positions in the target enterprise to obtain a position topology map corresponding to the multiple positions in the target enterprise, and transmit the position topology map corresponding to the multiple positions in the target enterprise to the node matching unit (7); The node matching unit (7) receives the defect types corresponding to the multiple electronic archives in the different types of electronic archive collections transmitted by the archive diagnosis unit (6) and the position topology map corresponding to the multiple positions in the target enterprise transmitted by the topology modeling unit (8), and performs node matching on the defect types corresponding to the multiple electronic archives in the different types of electronic archive collections and the position topology map corresponding to the multiple positions in the target enterprise to obtain a defect propagation map corresponding to the multiple positions in the target enterprise, and transmits the defect propagation map corresponding to the multiple positions in the target enterprise to the defect assessment unit (9).
4. The artificial intelligence-based archive management system according to claim 3, characterized in that: The defect assessment unit (9) receives the defect propagation graph corresponding to the multiple positions in the target enterprise transmitted by the node matching unit (7), and calculates the similarity of each position topology node in the position topology graph corresponding to the multiple positions in the target enterprise, calculates the connection edge weights in the defect propagation graph corresponding to the multiple positions in the target enterprise based on the similarity of each position topology node, calculates the node weights in the defect propagation graph corresponding to the multiple positions in the target enterprise based on the node importance corresponding to the position topology graph corresponding to the multiple positions in the target enterprise, obtains a weighted defect propagation graph based on the connection edge weights and node weights in the defect propagation graph, performs defect assessment on the weighted defect propagation graph to obtain the defect level corresponding to the weighted defect propagation graph, and transmits the defect level corresponding to the weighted defect propagation graph to the defect annotation unit (10).
5. The artificial intelligence-based archive management system according to claim 4, characterized in that: The defect marking unit (10) receives the defect level corresponding to the weighted defect propagation graph transmitted by the defect assessment unit (9), and marks the defect level of the corresponding area in the position topology graph corresponding to the multiple positions in the target enterprise based on the defect level corresponding to the weighted defect propagation graph, so as to obtain the dynamic position topology graph corresponding to the multiple positions in the target enterprise, and transmits the dynamic position topology graph corresponding to the multiple positions in the target enterprise to the file interface unit (5); The archive interface unit (5) receives the different types of electronic archive collections transmitted by the archive classification unit (3) and the dynamic position topology map corresponding to the multiple positions in the target enterprise transmitted by the defect marking unit (10), and marks the different types of electronic archive collections to the corresponding position topology nodes in the dynamic position topology map corresponding to the multiple positions in the target enterprise, and displays the different types of electronic archive collections when the corresponding position topology nodes in the dynamic position topology map corresponding to the multiple positions in the target enterprise are clicked.
6. The artificial intelligence-based archive management system according to claim 1, characterized in that: The defect types corresponding to the multiple electronic archives in the different types of electronic archive sets are node matched with the position topology graph corresponding to the multiple positions in the target enterprise to obtain the defect propagation graph corresponding to the multiple positions in the target enterprise, including: Matching each electronic file in the different types of electronic file collections with a preset node keyword list to obtain a position topology node corresponding to each electronic file in the different types of electronic file collections; Based on the defect type corresponding to each electronic file in the different types of electronic file sets, candidate abnormal node screening is performed on the position topology nodes in the position topology graph corresponding to the multiple positions in the target enterprise to obtain a candidate abnormal node set of the position topology graph corresponding to the multiple positions in the target enterprise; Defect propagation graphs of multiple positions in the target enterprise are extracted based on the candidate abnormal node set and the connection edges between the candidate abnormal nodes.
7. The artificial intelligence-based archive management system according to claim 1, characterized in that: Classify the standard electronic archive texts into different types of electronic archive collections, including: Segmenting the standard electronic archive text to obtain an archive segment set corresponding to the standard electronic archive text; Perform keyword matching on the archive segment set and a preset keyword list, and if an archive segment in the archive segment set matches the preset keyword list, obtain the keyword corresponding to the archive segment in the archive segment set; If the archive segment in the archive segment set does not match the preset keyword list, semantically matching the unmatched archive segment in the archive segment set with the preset keyword list based on a word similarity model to obtain keywords corresponding to the unmatched archive segment in the archive segment set; Combining the keywords corresponding to the various archive segments in the archive segment set to obtain a keyword set corresponding to the archive segment set; The archive segments are classified into different types of log sets based on the keyword set.
8. The artificial intelligence-based archive management system according to claim 4, characterized in that: The calculation formula of the similarity of the position topology nodes is as follows: Among them, w vi,vj represents the similarity between any two adjacent nodes vi and vj, α represents the preset weight coefficient, and x vi and x vj represents the feature vector of two adjacent nodes vi and vj, n represents the dimension of the feature vector, x vi,k and x vj,k Represents the kth dimension of the feature vector of two adjacent nodes vi and vj; The calculation formula of the node importance corresponding to the position topology node is as follows: Among them, w vi Indicates the node importance corresponding to the position topology node, P(x vi ) represents the Gaussian kernel probability density estimate of node vi.
9. The artificial intelligence-based archive management system according to claim 8, characterized in that: A weighted defect propagation graph is obtained based on the connection edge weights and node weights in the defect propagation graph, and a defect assessment is performed on the weighted defect propagation graph to obtain a defect level corresponding to the weighted defect propagation graph, including: The transition probability matrix is calculated based on the edge weights of the weighted defect propagation graph, and the transition probability matrix is as follows: Among them, P vi,vj represents the probability of node vi transferring to the adjacent node vj; The node weights of the iteratively updated post topology nodes of the weighted defect propagation graph are calculated based on the transition probability matrix. The calculation formula of the node weights of the iteratively updated post topology nodes is as follows: in, represents the node weight of the position topology node after iterative update, q vi,vj Indicates that each iteration is 1-q vi,vj The probability of jumping back to a random position topological node and taking q vi,vj The probability of continues to move along the connecting edges in the weighted defect propagation graph; The final node weights of the job topology nodes in the weighted defect propagation graph are summed to obtain the sum of the final node weights of the weighted defect propagation graph, and the final node weights of the weighted defect propagation graph are matched with the preset weight threshold level table to obtain the defect level corresponding to the weighted defect propagation graph.
10. An artificial intelligence-based archive management method, which is applicable to an artificial intelligence-based archive management system according to any one of claims 1 to 9, characterized in that: include: Acquire electronic archives uploaded by multiple positions in the target enterprise to obtain an electronic archive set, convert the electronic archive set into an electronic archive text, and perform preprocessing operations on the electronic archive text to obtain a standard electronic archive text; Classifying the standard electronic archive text into different types of electronic archive sets, and performing feature extraction on each electronic archive in the different types of electronic archive sets to obtain a word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive sets; Inputting the word frequency vector sequence corresponding to each electronic archive in the different types of electronic archive sets into the trained archive diagnosis model, and outputting the defect type corresponding to each electronic archive in the different types of electronic archive sets through the trained defect diagnosis model; Performing topological modeling based on the electronic archive flow between multiple positions in the target enterprise to obtain a position topological graph corresponding to multiple positions in the target enterprise, performing node matching on the position topological graph corresponding to multiple positions in the target enterprise by using the defect types corresponding to multiple electronic archives in the different types of electronic archive sets to obtain a defect propagation graph corresponding to multiple positions in the target enterprise; Performing defect assessment based on the defect propagation graphs corresponding to the multiple positions in the target enterprise to obtain the defect level corresponding to the weighted defect propagation graph, and marking the defect level of the corresponding areas in the position topology graph corresponding to the multiple positions in the target enterprise based on the defect level corresponding to the weighted defect propagation graph to obtain the dynamic position topology graph corresponding to the multiple positions in the target enterprise; The electronic archives of different types are set and marked to corresponding position topology nodes in a dynamic position topology map corresponding to a plurality of positions in the target enterprise, and the dynamic position topology map is displayed.