Document management method and system based on big data analysis

Abstract

The present invention discloses a document management method and system based on big data analysis, which relates to the field of document management technology, including using weighted linear combination to calculate conflict scores, screening conflict scores greater than a score threshold, and obtaining conflict words; assigning a LIF neuron to each conflict word, using the LIF model to update the membrane potential of each neuron, screening membrane potentials greater than a trigger threshold, marking them as high-conflict words, calculating the turbulence dissipation rate of high-conflict words, calculating the priority and fitness of candidate words, sorting the candidate word set of each high-conflict word in descending order according to fitness, and selecting the highest candidate word for replacement. The present invention encodes words using the BERT model, combines dependency syntactic analysis and the TF-IDF weighted semantic distance method, calculates the semantic heat value and conflict score of each word, and improves the semantic matching and accuracy between the word and the context.

Description

Technical Field

[0001] The present invention relates to the field of document management technology, and in particular to a document management method and system based on big data analysis. Background Art

[0002] With the rapid development of information technology, the surge in data volume has made large-scale document management increasingly complex. Traditional document management methods mainly rely on keyword indexing, manual classification, and rule-based document organization. Although these methods can meet the needs of basic document storage and retrieval to a certain extent, as the amount of information increases, efficiency and accuracy are often difficult to guarantee, especially when processing heterogeneous documents. With the rapid development of big data analysis technology and natural language processing (NLP) technology, document management methods based on semantic analysis have gradually become a research hotspot. In particular, through machine learning and deep learning models, key information can be automatically extracted from large amounts of documents, structured data can be generated, and retrieval and management processes can be optimized.

[0003] Existing document management methods based on big data analysis still face some key challenges. When dealing with conflicting words at the semantic level, existing technologies lack effective algorithms to measure the semantic similarity between words, resulting in the inability to effectively identify and optimize conflicting information in documents, affecting the accurate management of document data. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a document management method and system based on big data analysis to solve the problem that the existing technology lacks an effective algorithm to measure the semantic similarity between words when processing conflicting words at the semantic level, resulting in the inability to effectively identify and optimize conflicting information in documents, affecting the accurate management of document data.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a document management method based on big data analysis, which comprises:

[0008] Document data is collected and preprocessed to generate words. Syntactic coordinates and semantic heat values ​​are mapped into a two-dimensional space to construct a semantic heat field. The gradient modulus of the semantic heat field is calculated using numerical differentiation methods. Each word is defined as a particle, and the displacement modulus of the updated word position is calculated. Based on the gradient modulus and displacement modulus, a weighted linear combination is used to calculate the conflict score. Conflict scores greater than a score threshold are filtered out to obtain conflicting words.

[0009] Assign a LIF neuron to each conflicting word, use the LIF model to update the membrane potential of each neuron, screen the membrane potential greater than the trigger threshold, mark it as a high-conflict word, calculate the turbulence dissipation rate of the high-conflict word, calculate the priority and fitness of the candidate words, sort the candidate word set of each high-conflict word in descending order according to fitness, and select the candidate word with the highest fitness for replacement;

[0010] Build a visual interface to display the replacement results and store the document data generated by collection and analysis.

[0011] As a preferred solution of the document management method based on big data analysis described in the present invention, wherein: the conflict score is calculated using a weighted linear combination, and conflict scores greater than a score threshold are screened to obtain conflict words, including:

[0012] Use the pre-trained BERT model to encode words, generate semantic vectors for each word and perform normalization. Use dependency parsing to parse the text sentences containing the words, generate a dependency syntax tree, assign syntactic coordinates to each word, and obtain the position vector of each word.

[0013] The cosine function is used to calculate cosine similarity. Based on the cosine similarity, the TF-IDF weighted semantic distance method is used to calculate the semantic heat value of each word. The syntactic coordinates and semantic heat value are mapped to a two-dimensional space to construct a semantic heat field. The gradient modulus of the semantic heat field is calculated using the numerical differentiation method.

[0014] Define each word as a particle. Based on the normalized semantic vector, calculate the initial velocity vector of each particle and construct a semantic field tensor. Use PCA to extract the maximum eigenvector of the semantic field tensor and define it as the flow field velocity direction vector. Superimpose the initial velocity vector and the flow field velocity direction vector to obtain the final velocity of the particle. Use the final velocity of the particle to update the word position and calculate the displacement modulus of the updated word position. Based on the gradient modulus and the displacement modulus, use a weighted linear combination to calculate the conflict score.

[0015] The percentile method is used to set the scoring threshold, and conflicting words with scores greater than the scoring threshold are filtered to obtain conflicting words.

[0016] As a preferred solution of the document management method based on big data analysis of the present invention, wherein: the screening of membrane potentials greater than the trigger threshold and marking them as high-conflict words includes:

[0017] Assign a LIF neuron to each conflicting word, initialize the membrane potential to 0, calculate the current of each neuron, use the LIF model to update the membrane potential of each neuron, and use the Euler method to solve it to obtain the membrane potential of each neuron;

[0018] The trigger threshold was set using the statistical distribution method, and membrane potentials greater than the trigger threshold were screened and marked as high-conflict words.

[0019] As a preferred solution of the document management method based on big data analysis of the present invention, wherein: the step of sorting in descending order according to fitness and selecting the highest candidate word for replacement includes:

[0020] Collect candidate words from the industry vocabulary, use the pre-trained Word2Vec model to convert the candidate words into candidate vectors, use the k-selection method to filter them, and perform normalization to obtain a candidate word set;

[0021] Calculate the turbulence dissipation rate of high-conflict words, and calculate the priority and fitness of candidate words based on the candidate word set;

[0022] For each high-conflict word candidate set, sort them in descending order according to fitness, and select the highest candidate word for replacement.

[0023] As a preferred solution of the document management method based on big data analysis of the present invention, the step of constructing a visual interface to display the replacement results includes:

[0024] Build a visual interface to display the replacement results, allowing users who have passed real-name verification to view them.

[0025] As a preferred solution of the document management method based on big data analysis of the present invention, the storing of the document data collected and analyzed includes:

[0026] The collected document data and the replacement results generated by the analysis are stored in the central database, and security access measures are set. The central database will back up the stored data to the cloud and regularly perform integrity checks on the stored data and backup data. After the test is completed, the integrity test record will be generated and stored synchronously in the central database.

[0027] As a preferred solution of the document management method based on big data analysis described in the present invention, the process of collecting document data and preprocessing to generate words includes:

[0028] Use the APL interface to collect document data from cloud storage services, use document parsing to extract text from the document data to obtain text sentences, use natural language processing to segment the text sentences, and perform text cleaning on the segmentation results to generate words.

[0029] In a second aspect, the present invention provides a document management system based on big data analysis, comprising:

[0030] The collection and screening module is used to collect and preprocess document data, generate words, map syntactic coordinates and semantic heat values ​​into a two-dimensional space, construct a semantic heat field, calculate the gradient modulus of the semantic heat field using numerical differentiation methods, define each word as a particle, calculate the displacement modulus to update the word position, calculate the conflict score based on the gradient modulus and displacement modulus using weighted linear combination, and filter out conflict scores greater than the score threshold to obtain conflict words;

[0031] The tag replacement module is used to assign a LIF neuron to each conflicting word, update the membrane potential of each neuron using the LIF model, screen the membrane potentials greater than the trigger threshold, mark them as high-conflict words, calculate the turbulence dissipation rate of high-conflict words, calculate the priority and fitness of candidate words, sort the candidate word set of each high-conflict word in descending order according to fitness, and select the candidate word with the highest fitness for replacement;

[0032] The visual storage module is used to build a visual interface to display the replacement results and store the document data collected and analyzed.

[0033] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the document management method based on big data analysis as described in the first aspect of the present invention is implemented.

[0034] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the document management method based on big data analysis as described in the first aspect of the present invention.

[0035] The beneficial effects of the present invention are as follows: the present invention encodes words by using the BERT model, combines dependency syntactic analysis and the TF-IDF weighted semantic distance method, calculates the semantic heat value and conflict score of each word, thereby improving the semantic matching degree and accuracy between the word and the context, calculates the gradient modulus of the semantic heat field by the numerical differentiation method, and calculates the conflict score by weighted linear combination, thereby optimizing the semantic understanding and recommendation process of the document content, and enhancing the response speed and accuracy of the document management system in complex queries. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0037] Figure 1This is a flowchart of the document management method based on big data analysis in Example 1.

[0038] Figure 2 Schematic diagram of the document management system based on big data analysis in Example 1. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0042] Example 1, with reference to Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a document management method based on big data analysis, including the following steps:

[0043] S1. Collect document data and preprocess it to generate words. Map syntactic coordinates and semantic heat values ​​to a two-dimensional space to construct a semantic heat field. Use numerical differentiation methods to calculate the gradient modulus of the semantic heat field. Define each word as a particle and calculate the displacement modulus of the updated word position. Based on the gradient modulus and displacement modulus, use weighted linear combination to calculate the conflict score. Filter the conflict scores that are greater than the score threshold to obtain the conflict words.

[0044] Specifically, document data is collected and preprocessed to generate words, including:

[0045] Use the APL interface to collect document data from cloud storage services, use document parsing to extract text from the document data to obtain text sentences, use natural language processing to segment the text sentences, and perform text cleaning on the segmentation results (use regular expressions to remove numbers, extra spaces, line breaks, and punctuation in the text, use the stop word list provided by NLTK to remove common meaningless words, and perform standard processing) to generate words.

[0046] By removing redundant information such as unnecessary symbols, numbers, spaces, and stop words, the efficiency and accuracy of subsequent text analysis can be significantly improved. Using pre-trained natural language processing models for standardized processing ensures consistency between documents from different sources, providing an accurate text basis for subsequent semantic analysis.

[0047] Furthermore, a weighted linear combination is used to calculate the conflict score, and conflict scores greater than the score threshold are filtered to obtain conflict words, including:

[0048] Use the pre-trained BERT model to encode words, generate semantic vectors for each word and perform normalization. Use dependency parsing to parse the text sentences containing the words, generate a dependency syntax tree, assign syntactic coordinates to each word, and obtain the position vector of each word.

[0049] The cosine function is used to calculate the cosine similarity. Based on the cosine similarity, the TF-IDF weighted semantic distance method is used to calculate the semantic heat value of each word. The formula is:

[0050] ,

[0051] ,

[0052] ,

[0053] in is the semantic heat value of the i-th word, is the TF-IDF value of the i-th word (calculated using TF-IDF), k is the window size (selected using empirical heuristics), is the average semantic vector of the i-th word context (calculated using average pooling), and are the semantic vectors of the i-th and j-th words after normalization, is the context window of the i-th word (set using the neighboring word selection method), is the cosine similarity between the normalized semantic vector and the average semantic vector;

[0054] Map the syntactic coordinates and semantic heat values ​​into two-dimensional space to construct the semantic heat field. The formula is:

[0055] ,

[0056] ,

[0057] in The semantic thermal field in the two-dimensional plane The value at is the syntactic coordinate of the i-th word, n is the number of words, is the diffusion factor, which controls the diffusion range of the calorific value. is the basic diffusion value (set using empirical estimation), is the diffusion scaling factor (set using grid search);

[0058] The numerical differentiation method is used to calculate the gradient modulus of the semantic thermal field. The formula is:

[0059] ,

[0060] in The semantic thermal field in the two-dimensional plane The gradient modulus at ;

[0061] Define each word as a particle, and calculate the initial velocity vector of each particle based on the normalized semantic vector. The formula is:

[0062] ,

[0063] ,

[0064] in is the initial velocity vector of the pth particle, is the scaling factor (set using dynamic scaling), Regularization constant to avoid division by zero;

[0065] Construct the semantic field tensor, the formula is:

[0066] ,

[0067] ,

[0068] in is the semantic field tensor of the p-th particle, For the outer product operation, generate the interaction matrix between semantics and position, and are the position vectors of the i-th and j-th words respectively;

[0069] PCA is used to extract the maximum eigenvector of the semantic field tensor, which is defined as the flow field velocity direction vector. The initial velocity vector and the flow field velocity direction vector are superimposed to obtain the final velocity of the particle. The formula is:

[0070] ,

[0071] in is the final velocity of the pth particle, is the velocity direction vector of the p-th particle;

[0072] Use the final velocity of the particle to update the word position and calculate the displacement modulus of the updated word position. The formula is:

[0073] ,

[0074] ,

[0075] in is the updated particle position of the i-th particle at time t+1, is the displacement modulus of the i-th particle at time t+1;

[0076] Based on the gradient modulus and displacement modulus, a weighted linear combination is used to calculate the conflict score, which is:

[0077] ,

[0078] in For a two-dimensional plane The conflict score at and are the weights of the weighted linear combination (set using web search);

[0079] The percentile method is used to set the scoring threshold, and conflicting words with scores greater than the scoring threshold are filtered to obtain conflicting words.

[0080] The BERT model can capture the deep semantic information of words, overcoming the limitations of traditional rule-based word segmentation methods. This allows words in documents to not only rely on their surface meaning but also reflect the semantics of the context. Dependency parsing provides a structured representation of the grammatical relationships between words, allowing the system to simultaneously consider the grammatical connections between words when analyzing documents. By combining semantic vectors and syntactic coordinates, it not only analyzes the independent semantics of vocabulary but also reveals the relationships between words and their importance in the document. Particle swarm optimization simulates the laws of particle interaction in nature by dynamically adjusting the positions of words, effectively optimizing the semantic relationships in documents. By calculating the gradient modulus and displacement modulus of the semantic thermal field, the present invention can identify potential conflicting words in documents and effectively process them. The LIF model simulates the discharge process of neurons. Its advantage is that it can make real-time dynamic adjustments based on the semantics and contextual information of conflicting words, avoiding the problem of information misunderstanding or loss caused by conflicting words in traditional document management and improving the accuracy and relevance of documents.

[0081] S2. Assign a LIF neuron to each conflicting word, use the LIF model to update the membrane potential of each neuron, screen the membrane potential greater than the trigger threshold, mark it as a high-conflict word, calculate the turbulence dissipation rate of the high-conflict word, calculate the priority and fitness of the candidate words, sort the candidate word set of each high-conflict word in descending order according to fitness, and select the candidate word with the highest fitness to replace it;

[0082] Specifically, membrane potentials greater than the trigger threshold are screened and marked as high-conflict words, including:

[0083] Assign a LIF neuron to each conflict word, initialize the membrane potential to 0, and calculate the current of each neuron. The formula is:

[0084] ,

[0085] in At time t, The current of a neuron, For the The context window of neurons (based on conflict word statistics), and Respectively and The semantic vector of and Respectively and The position vector of

[0086] The membrane potential of each neuron is updated using the LIF model, as follows:

[0087] ,

[0088] in is the time constant (set using the empirical setting method) that controls the membrane potential decay rate, At time t, The membrane potential of each neuron;

[0089] Use the Euler method to solve and obtain the membrane potential of each neuron. The formula is:

[0090] ,

[0091] in is the time step (set using stability analysis);

[0092] The trigger threshold was set using the statistical distribution method, and membrane potentials greater than the trigger threshold were screened and marked as high-conflict words.

[0093] The LIF neuron model can simulate the biopotential change process of neurons. Its membrane potential gradually changes with the passage of time and the influence of external current. By mapping each conflicting word to an independent neuron, the processing of conflicting words can be controlled more fine-grainedly. The membrane potential update process can reflect the influence that the conflicting words gradually generate in the document. This dynamic process improves the flexibility and accuracy of conflicting word recognition. The calculated results of the membrane potential provide a basis for further conflicting word screening. By real-time updating of the membrane potential of each neuron, the system can dynamically identify high-conflict words in the document and effectively mark them. The trigger threshold of the membrane potential can be dynamically adjusted according to the context of different documents, allowing the system to flexibly respond to different types of document content.

[0094] Furthermore, the candidates are sorted in descending order according to fitness, and the highest candidate words are selected for replacement, including:

[0095] Collect candidate words from the industry vocabulary, use the pre-trained Word2Vec model to convert the candidate words into candidate vectors, use the k-selection method to filter them, and perform normalization to obtain a candidate word set;

[0096] Calculate the turbulence dissipation rate of high-conflict words using the formula:

[0097] ,

[0098] in For the The turbulent dissipation rate of high-conflict words, For the context windows of high-conflict words, based on syntactic coordinate distance, and Respectively and The final speed of high-conflict words, and Respectively and Position vectors of high-conflict words, is the regularization constant, prevent it from being 0;

[0099] Based on the candidate word set, the priority and fitness of the candidate words are calculated using the following formula:

[0100] ,

[0101] ,

[0102] ,

[0103] in For the The context vectors of high-conflict words, is the priority of the qth candidate word, is the normalized candidate vector of the qth candidate word, is the fitness of the qth candidate word;

[0104] For each high-conflict word candidate set, sort them in descending order according to fitness, and select the highest candidate word for replacement.

[0105] By combining the industry vocabulary and deep semantic analysis of the Word2Vec model, multiple suitable alternative words can be provided for each conflicting word, thereby ensuring effective replacement while ensuring semantic consistency, avoiding the context distortion caused by the fixed replacement of a single word. The priority sorting of conflicting words can be further refined through the turbulent dissipation rate, making the conflicting word replacement process more efficient and accurate. The calculation of fitness can take into account multi-dimensional semantic information, including context windows, grammatical structures and position vectors, making the replacement process more refined and intelligent, and the semantic relevance and fluency of the replaced words effectively avoid the semantic deviation caused by improper word replacement.

[0106] S3. Build a visual interface to display the replacement results and store the document data collected and analyzed.

[0107] Specifically, a visual interface is built to display the replacement results, including:

[0108] Build a visual interface to display the replacement results, allowing users who have passed real-name verification to view them.

[0109] The visual interface makes word replacement in documents more intuitive. Users can clearly see which words are replaced and what the new words are, thereby enhancing their trust in the system processing results. Real-name verification ensures that only authorized users can view and modify documents, thereby effectively preventing unauthorized operations and avoiding data leakage or malicious tampering.

[0110] Furthermore, the document data generated by the collection and analysis is stored, including:

[0111] The collected RGB images and the detection results generated by the analysis are stored in the central database, and security access measures are set. The central database will back up the stored data to the cloud and regularly perform integrity checks on the stored data and backup data. After the test is completed, the integrity test record will be generated and stored synchronously in the central database.

[0112] By recording and synchronizing each test result to a central database, data reliability and traceability are further ensured. Through cloud backup, all document data and replacement results can be stored in a distributed cloud environment, greatly reducing the risk of data loss. By regularly performing data integrity checks, the system can automatically detect any damage or tampering with the data, and repair or update it in a timely manner to ensure the consistency of backup data with the original data, thereby enhancing the reliability and security of data management.

[0113] This embodiment also provides a document management system based on big data analysis, including:

[0114] The collection and screening module is used to collect and preprocess document data, generate words, map syntactic coordinates and semantic heat values ​​into a two-dimensional space, construct a semantic heat field, calculate the gradient modulus of the semantic heat field using numerical differentiation methods, define each word as a particle, calculate the displacement modulus to update the word position, calculate the conflict score based on the gradient modulus and displacement modulus using weighted linear combination, and filter out conflict scores greater than the score threshold to obtain conflict words;

[0115] The tag replacement module is used to assign a LIF neuron to each conflicting word, update the membrane potential of each neuron using the LIF model, screen the membrane potentials greater than the trigger threshold, mark them as high-conflict words, calculate the turbulence dissipation rate of high-conflict words, calculate the priority and fitness of candidate words, sort the candidate word set of each high-conflict word in descending order according to fitness, and select the candidate word with the highest fitness for replacement;

[0116] The visual storage module is used to build a visual interface to display the replacement results and store the document data collected and analyzed.

[0117] This embodiment also provides a computer device suitable for the document management method based on big data analysis, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the document management method based on big data analysis proposed in the above embodiment.

[0118] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0119] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the document management method based on big data analysis proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0120] In summary, the present invention uses the BERT model to encode words, combines dependency syntactic analysis and the TF-IDF weighted semantic distance method to calculate the semantic heat value and conflict score of each word, thereby improving the semantic matching and accuracy between the word and the context. The gradient modulus of the semantic heat field is calculated by numerical differentiation method, and the conflict score is calculated using weighted linear combination. This optimizes the semantic understanding and recommendation process of document content and enhances the response speed and accuracy of the document management system in complex queries.

[0121] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A document management method based on big data analysis, characterized by: include, Document data is collected and preprocessed to generate words. Syntactic coordinates and semantic heat values ​​are mapped into a two-dimensional space to construct a semantic heat field. The gradient modulus of the semantic heat field is calculated using numerical differentiation methods. Each word is defined as a particle, and the displacement modulus of the updated word position is calculated. Based on the gradient modulus and displacement modulus, a weighted linear combination is used to calculate the conflict score. Conflict scores greater than a score threshold are filtered out to obtain conflicting words. Assign a LIF neuron to each conflicting word, use the LIF model to update the membrane potential of each neuron, screen the membrane potential greater than the trigger threshold, mark it as a high-conflict word, calculate the turbulence dissipation rate of the high-conflict word, calculate the priority and fitness of the candidate words, sort the candidate word set of each high-conflict word in descending order according to fitness, and select the candidate word with the highest fitness for replacement; Build a visual interface to display the replacement results and store the document data generated by collection and analysis.

2. The document management method based on big data analysis according to claim 1, characterized in that: The conflict score is calculated using a weighted linear combination, and conflict scores greater than a score threshold are screened to obtain conflict words, including: Use the pre-trained BERT model to encode words, generate semantic vectors for each word and perform normalization. Use dependency parsing to parse the text sentences containing the words, generate a dependency syntax tree, assign syntactic coordinates to each word, and obtain the position vector of each word. The cosine function is used to calculate cosine similarity. Based on the cosine similarity, the TF-IDF weighted semantic distance method is used to calculate the semantic heat value of each word. The syntactic coordinates and semantic heat value are mapped to a two-dimensional space to construct a semantic heat field. The gradient modulus of the semantic heat field is calculated using the numerical differentiation method. Define each word as a particle. Based on the normalized semantic vector, calculate the initial velocity vector of each particle and construct a semantic field tensor. Use PCA to extract the maximum eigenvector of the semantic field tensor and define it as the flow field velocity direction vector. Superimpose the initial velocity vector and the flow field velocity direction vector to obtain the final velocity of the particle. Use the final velocity of the particle to update the word position and calculate the displacement modulus of the updated word position. Based on the gradient modulus and the displacement modulus, use a weighted linear combination to calculate the conflict score. The percentile method is used to set the scoring threshold, and conflicting words with scores greater than the scoring threshold are filtered to obtain conflicting words.

3. The document management method based on big data analysis according to claim 2, characterized in that: The membrane potentials that are greater than the trigger threshold are marked as high-conflict words, including: Assign a LIF neuron to each conflicting word, initialize the membrane potential to 0, calculate the current of each neuron, use the LIF model to update the membrane potential of each neuron, and use the Euler method to solve it to obtain the membrane potential of each neuron; The trigger threshold was set using the statistical distribution method, and membrane potentials greater than the trigger threshold were screened and marked as high-conflict words.

4. The document management method based on big data analysis according to claim 3, characterized in that: The candidate words are sorted in descending order according to fitness and the highest candidate words are selected for replacement, including: Collect candidate words from the industry vocabulary, use the pre-trained Word2Vec model to convert the candidate words into candidate vectors, use the k-selection method to filter them, and perform normalization to obtain a candidate word set; Calculate the turbulence dissipation rate of high-conflict words, and calculate the priority and fitness of candidate words based on the candidate word set; For each high-conflict word candidate set, sort them in descending order according to fitness, and select the highest candidate word for replacement.

5. The document management method based on big data analysis according to claim 4, characterized in that: The construction of a visual interface to display the replacement results includes: Build a visual interface to display the replacement results, allowing users who have passed real-name verification to view them.

6. The document management method based on big data analysis according to claim 5, characterized in that: The document data generated by the collection and analysis is stored, including: The collected document data and the replacement results generated by the analysis are stored in the central database, and security access measures are set. The central database will back up the stored data to the cloud and regularly perform integrity checks on the stored data and backup data. After the test is completed, the integrity test record will be generated and stored synchronously in the central database.

7. The document management method based on big data analysis according to claim 6, characterized in that: The document data is collected and preprocessed to generate words, including: Use the APL interface to collect document data from cloud storage services, use document parsing to extract text from the document data to obtain text sentences, use natural language processing to segment the text sentences, and perform text cleaning on the segmentation results to generate words.

8. A document management system based on big data analysis, based on the document management method based on big data analysis according to any one of claims 1 to 7, characterized in that: include, The collection and screening module is used to collect and preprocess document data, generate words, map syntactic coordinates and semantic heat values ​​into a two-dimensional space, construct a semantic heat field, calculate the gradient modulus of the semantic heat field using numerical differentiation methods, define each word as a particle, calculate the displacement modulus to update the word position, calculate the conflict score based on the gradient modulus and displacement modulus using weighted linear combination, and filter out conflict scores greater than the score threshold to obtain conflict words; The tag replacement module is used to assign a LIF neuron to each conflicting word, update the membrane potential of each neuron using the LIF model, screen the membrane potentials greater than the trigger threshold, mark them as high-conflict words, calculate the turbulence dissipation rate of high-conflict words, calculate the priority and fitness of candidate words, sort the candidate word set of each high-conflict word in descending order according to fitness, and select the candidate word with the highest fitness for replacement; The visual storage module is used to build a visual interface to display the replacement results and store the document data collected and analyzed.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the document management method based on big data analysis according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the document management method based on big data analysis according to any one of claims 1 to 7 are implemented.

Images