Interdisciplinary knowledge graph construction system and method

Abstract

The invention relates to the technical field of information retrieval, in particular to an interdisciplinary knowledge graph construction system and method, and the system comprises a relation matrix construction module which analyzes a node relation based on nodes in a knowledge graph, carries out the statistics according to an association frequency, selects interdisciplinary nodes to construct an association matrix, calculates a node fitting degree, and determines an association node set; and generating an interdisciplinary incidence relation matrix. According to the method, through correlation frequency statistics and fitting degree calculation of interdisciplinary nodes, effective construction of comprehensive correlation between the nodes, decomposition of the feature matrix and elimination of non-key feature items, the sparsity of the matrix is improved, data are more concise and clear, and through screening of low-weight elements, reservation of important node relation and further optimization of matrix feature expression, the method has the advantages of being high in robustness and high in reliability. In the coupling analysis and partitioning process of the nodes, redundant connection of the low-coupling-degree nodes is reduced, the core relation is highlighted, the node structure is recombined, and the accuracy and compactness of the multidisciplinary knowledge graph are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of information retrieval technology, and in particular to a system and method for constructing an interdisciplinary knowledge graph. Background Art

[0002] Information retrieval is an important field in computer science, focusing on how to effectively find the information users need from a large amount of unstructured or semi-structured data. It includes the processing of various types of data such as text, images, audio, and video, with the aim of providing efficient and accurate information query services. Information retrieval technology is widely used in search engines, recommendation systems, document management, data mining and other fields, covering multiple links such as data collection, processing, analysis and display, in order to meet the diverse needs of users in terms of information acquisition.

[0003] Among them, the interdisciplinary knowledge graph construction system refers to a system that uses computer technology to unify data modeling and organization of knowledge from different disciplines to form a comprehensive knowledge graph. Its purpose is to help users integrate, associate and visualize knowledge from different disciplines, improve the accuracy and depth of information query, and is especially suitable for scenarios such as scientific research and education that require the integration of multidisciplinary knowledge, ultimately promoting the in-depth understanding and application of cross-domain knowledge.

[0004] Existing information retrieval systems are deficient in the integration of interdisciplinary knowledge and fail to achieve effective modeling of interdisciplinary node associations, resulting in weak connections between different disciplines, fragmented information, and difficulty meeting user query needs in complex multidisciplinary scenarios. In addition, existing systems lack effective feature optimization and sparse processing, resulting in a lot of redundant information, increasing retrieval difficulty and resource consumption. The associations between nodes have not been effectively classified and optimized, resulting in the complexity of coupling relationships, making it difficult to accurately obtain in-depth information in interdisciplinary fields, affecting query results and knowledge integration capabilities. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings existing in the prior art and propose an interdisciplinary knowledge graph construction system and method.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solution: A system for constructing an interdisciplinary knowledge graph includes: The relationship matrix construction module analyzes the node relationships based on the nodes in the knowledge graph, selects interdisciplinary nodes to construct the relationship matrix according to the association frequency statistics, calculates the node fit, determines the associated node set, and generates the interdisciplinary association relationship matrix; The matrix feature decomposition module decomposes the interdisciplinary association relationship matrix into a potential feature matrix and a relationship coefficient matrix, analyzes the weights of the coefficient matrix elements, removes non-critical feature items based on the weights, merges and adjusts the matrices, and obtains a sparse coefficient matrix; The sparsity constraint optimization module selects elements whose weights are lower than a preset threshold element by element based on the sparse coefficient matrix, retains some connections according to the sparse threshold, and establishes a sparse feature representation matrix; The block structure adaptive optimization module calculates node connections one by one based on the sparse feature representation matrix, marks nodes with strong coupling relationships, classifies them into the same block, analyzes the coupling relationship between blocks, screens connections with weak coupling for sparse processing, retains the remaining structures, and establishes an optimized inter-block coupling matrix; Based on the optimized inter-block coupling matrix, the interdisciplinary knowledge mapping module reconstructs the graph nodes and block structures, analyzes the coupling relationship between nodes, adjusts node connections, removes redundant connections, checks the strength of inter-block relationships, reorganizes nodes and blocks, and generates a reconstructed interdisciplinary relationship matrix for intelligent recommendation, providing researchers with suggestions for interdisciplinary cooperation based on the correlation between disciplines.

[0007] The interdisciplinary association relationship matrix includes association frequency, association node set, and node fitting degree; the sparse coefficient matrix includes potential feature matrix, key feature items, and relationship coefficient matrix; the sparse feature representation matrix includes sparse threshold, partial connection, and low weight elements; the optimized inter-block coupling matrix includes block structure, adaptive block partitioning, and inter-block coupling relationship; the reconstructed interdisciplinary relationship matrix includes node connection and inter-block relationship strength.

[0008] As a further solution of the present invention, the step of obtaining the interdisciplinary association relationship matrix is ​​specifically as follows: Based on the knowledge graph, the interaction frequency between nodes is counted, and interdisciplinary nodes are screened. Nodes with association frequencies higher than the average are selected to generate an interdisciplinary node association frequency matrix. Using the interdisciplinary node association frequency matrix, the total association of multiple nodes is calculated, the centrality of each node is evaluated, and the node fitness vector is obtained; Using the node fitness vector and the interdisciplinary node association frequency matrix, an association matrix is ​​constructed by a fitness weighting method, using the formula: ; Calculate the association strength between node i and node j to obtain the interdisciplinary association matrix ; in, Represents the association frequency between node i and node j, which is a basic indicator to measure the frequency of interaction between two nodes. , Represent the fitness of nodes i and j, respectively, which reflects the criticality or centrality of each node in the interdisciplinary network, is the adjustment factor used to strengthen the associated frequency Influence in computing, is another adjustment factor used to strengthen the average value of node fitness The role of association strength, is the normalization factor used to adjust the values ​​in the denominator, It represents the sum of the absolute values ​​of the associated frequency differences of all other nodes connected to node i and node j.

[0009] As a further solution of the present invention, the step of obtaining the sparse coefficient matrix is ​​specifically as follows: Based on principal component analysis, the interdisciplinary association relationship matrix is ​​decomposed into a potential feature matrix and a relationship coefficient matrix, initial potential features and weight relationships are identified, and preliminary feature and coefficient matrices are generated; Calculate the weight of each element from the preliminary feature and coefficient matrix, and set a weight threshold to remove non-critical feature items with weights lower than the threshold to obtain an adjusted relationship coefficient matrix; The adjusted relationship coefficient matrix is ​​combined with the potential feature matrix, and the matrix reconstruction formula is applied: ; Perform compression, calculate and generate a sparse coefficient matrix; Among them, S represents the sparse coefficient matrix, V represents the potential feature matrix, and D represents the adjusted relationship coefficient matrix. represents the weight parameter that strengthens the impact of key features, represents the adjustment coefficient for matrix sparsification, represents the transposed matrix of the latent feature matrix, Represents the diagonal elements of the relationship coefficient matrix D, Represents the sum of the diagonal elements of the relationship coefficient matrix D.

[0010] As a further solution of the present invention, the step of obtaining the sparse feature representation matrix is ​​specifically: Based on the sparse coefficient matrix, determine whether the weight of each element in the matrix is ​​lower than a preset screening threshold, mark and remove the elements below the threshold, and retain the elements above the threshold to establish a preliminary screening matrix; Using the preliminary screening matrix, iterative screening is performed on the remaining contact weight elements, contact items are called according to the weight components and retained according to the sparse threshold, using the formula: , Calculate the adjusted sparse connection weights to obtain the sparse connection matrix retained after screening; in, represents the contact weight after screening, represents the weight value of the i-th row and j-th column in the sparse coefficient matrix, T represents the preset weight screening threshold, Represents the ratio of the adjustment weight to the threshold, which is used to adjust the impact of the threshold. Represents the balance adjustment coefficient, which corrects the overall offset of the sparse matrix. represents an additional sparse adjustment factor, which is used to enhance the sparseness characteristics in the sparse matrix. Represents the square root form of the additional sparse adjustment term, which is used to nonlinearly correct the weight value; Based on the sparse connection matrix retained after the screening, the remaining connection items are integrated to obtain a sparse feature representation matrix.

[0011] As a further solution of the present invention, the step of obtaining the optimized inter-block coupling matrix is ​​specifically as follows: Based on the sparse feature representation matrix, the connection strength of each node pair is calculated one by one, and by comparing whether the weight of each pair of connections is higher than a set coupling strength threshold, the node pairs that meet the conditions are marked as strongly coupled, and a node coupling relationship matrix is ​​obtained; From the node coupling relationship matrix, the node pairs marked as strongly coupled are classified into the same block, and the coupling relationship of the node pairs in each block is analyzed using the formula: , Calculate the coupling degree between blocks; in, represents the coupling degree between block i and block j, represents the connection strength between node i and node k, represents the connection strength between node k and node j, represents the adjustment coefficient, which is used to avoid the situation where the denominator is zero, k represents the node index, and n represents the total number of nodes; According to the coupling degree between the blocks, connections are iteratively screened based on the coupling degree, and connections with weak coupling degree are sparsely processed, and some connections are retained based on the coupling degree to construct an optimized inter-block coupling matrix.

[0012] As a further solution of the present invention, the steps of obtaining the reconstructed interdisciplinary relationship matrix are specifically as follows: Based on the optimized inter-block coupling matrix, the connection relationship of each node is analyzed one by one, and by calling the coupling data of multiple nodes, the node pairs are identified and marked based on the coupling relationship, and the node coupling adjustment matrix is ​​generated; From the node coupling adjustment matrix, the connection between nodes is analyzed using the formula: , Calculate the coupling degree, remove connections with low coupling degree, optimize the direct interaction between nodes, and obtain the adjusted node connection matrix; in, represents the coupling strength between node i and node j, represents the connection weight between node i and node k, represents the connection weight between node k and node j, represents a constant for adjusting the coupling calculation to avoid the denominator being zero. represents the sum of squared weights from node i to all other related nodes, Represents the sum of squared weights from node j to all other associated nodes; According to the adjusted node connection matrix, the nodes and blocks are reorganized, the node pairs with high coupling strength are grouped into the same block, the inter-block relationship is checked and optimized again, and the reconstructed interdisciplinary relationship matrix is ​​generated.

[0013] A method for constructing an interdisciplinary knowledge graph, which is executed based on the above-mentioned interdisciplinary knowledge graph construction system, and includes the following steps: S1: Conduct in-depth analysis of the nodes in the knowledge graph, select interdisciplinary nodes based on the interaction frequency between nodes, and screen the node set with high correlation by calculating the fit of multiple nodes, so as to construct an interdisciplinary correlation matrix. After the calculation is completed, an interdisciplinary correlation matrix is ​​generated; S2: split the interdisciplinary association matrix into a potential feature matrix and a relationship coefficient matrix, analyze the relationship coefficients item by item, filter out non-critical feature items based on weight components, and obtain a sparse coefficient matrix after merging and adjusting; S3: Based on the sparse feature matrix, screen each element one by one, remove weight elements below a preset threshold, retain key connections, refine feature representation, construct a sparse structure, and obtain an optimized feature matrix; S4: Use the optimized feature matrix to analyze and reconstruct the graph nodes and block structures, mark high-coupling nodes and adjust the connection relationships, remove redundant connections, strengthen the relationship between nodes and blocks, and establish a reconstructed interdisciplinary relationship matrix.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, the comprehensive association between nodes is effectively constructed through the correlation frequency statistics and fit calculation of interdisciplinary nodes, which enhances the connectivity of interdisciplinary knowledge. The decomposition of the feature matrix and the elimination of non-critical feature items improve the sparsity of the matrix, making the data more concise and clear. By screening low-weight elements, important node connections are retained, and the matrix feature expression is further optimized, the computational complexity is reduced and effective information is retained. In the process of node coupling analysis and block segmentation, the redundant connections of low-coupling nodes are reduced, the core relationship is highlighted, the node structure is reorganized, and the accuracy and compactness of the multidisciplinary knowledge graph are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a system flow chart of the present invention; Figure 2 A flow chart of the steps for obtaining the interdisciplinary association relationship matrix of the present invention; Figure 3 A flow chart of the steps for obtaining a sparse coefficient matrix of the present invention; Figure 4 A flow chart of the steps of obtaining a sparse feature representation matrix according to the present invention; Figure 5 A flow chart of the steps for obtaining the inter-block coupling matrix optimized for the present invention; Figure 6 Flow chart of the steps for obtaining the interdisciplinary relationship matrix reconstructed by the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] Embodiment 1:

[0019] See also Figure 1 , the present invention provides a technical solution: an interdisciplinary knowledge graph construction system comprising: The relationship matrix construction module analyzes the node relationships based on the nodes in the knowledge graph, selects interdisciplinary nodes to construct the relationship matrix according to the association frequency statistics, calculates the node fit, determines the associated node set, and generates the interdisciplinary association relationship matrix; The matrix feature decomposition module decomposes the interdisciplinary relationship matrix into a potential feature matrix and a relationship coefficient matrix, analyzes the weights of the coefficient matrix elements, removes non-critical feature items based on the weights, merges and adjusts the matrix, and obtains a sparse coefficient matrix. The sparsity constraint optimization module is based on the sparse coefficient matrix, and selects elements whose weights are lower than the preset threshold element by element, retains some connections according to the sparse threshold, and establishes a sparse feature representation matrix; The block structure adaptive optimization module calculates node connections one by one based on the sparse feature representation matrix, marks nodes with strong coupling relationships, classifies them into the same block, analyzes the coupling relationship between blocks, screens connections with weak coupling for sparse processing, retains the remaining structures, and establishes an optimized inter-block coupling matrix; The interdisciplinary knowledge mapping module reconstructs the graph nodes and block structures based on the optimized inter-block coupling matrix, analyzes the coupling relationship between nodes, adjusts node connections, removes redundant connections, checks the strength of inter-block relationships, reorganizes nodes and blocks, and generates a reconstructed interdisciplinary relationship matrix for intelligent recommendation, providing researchers with suggestions for interdisciplinary cooperation based on the correlation between disciplines.

[0020] The interdisciplinary association relationship matrix includes association frequency, association node set, and node fitting degree. The sparse coefficient matrix includes potential feature matrix, key feature items, and relationship coefficient matrix. The sparse feature representation matrix includes sparse threshold, partial connection, and low weight elements. The optimized inter-block coupling matrix includes block structure, adaptive block partitioning, and inter-block coupling relationship. The reconstructed interdisciplinary relationship matrix includes node connection and inter-block relationship strength.

[0021] See also Figure 2 , the specific steps for obtaining the interdisciplinary association relationship matrix are: Based on the knowledge graph, the interaction frequency between nodes is counted, and interdisciplinary nodes are screened. Nodes with association frequencies higher than the average are selected to generate an interdisciplinary node association frequency matrix. Based on the node data of the knowledge graph, a statistical analysis of the interaction frequency between nodes was performed. First, the interaction data of each node was extracted and its total frequency was calculated. By comparing with the average association frequency, those nodes with frequencies higher than the average were selected as key interaction nodes. Next, these nodes were used to construct a preliminary association frequency matrix, which preliminarily reflects the interaction intensity between interdisciplinary nodes. This process ensures that only the most representative nodes are selected for further analysis. In this way, an accurate matrix reflecting the nodes with higher actual associations is obtained, which will be used for the next step of node fit calculation.

[0022] Using the interdisciplinary node association frequency matrix, the total association of multiple nodes is calculated, the centrality of each node is evaluated, and the node fitness vector is obtained; Using the interdisciplinary node association frequency matrix, the association degree of each node is further calculated. The total association degree of each node is achieved by adding all the association frequencies of the node and dividing it by the total number of nodes. This process not only provides the relative importance of each node in the entire network, but also helps to understand which nodes may be key information dissemination points. The obtained node fitness vector is obtained by standardizing these total association data, ensuring the consistency and comparability of the data. This vector provides the necessary mathematical basis for constructing an interdisciplinary association matrix.

[0023] Using the node fitness vector and the interdisciplinary node association frequency matrix, the association matrix is ​​constructed by the fitness weighting method, using the formula: ; Calculate the association strength between node i and node j to obtain the interdisciplinary association matrix ; in, Represents the association frequency between node i and node j, which is a basic indicator to measure the frequency of interaction between two nodes. , Represent the fitness of nodes i and j, respectively, which reflects the criticality or centrality of each node in the interdisciplinary network, is the adjustment factor used to strengthen the associated frequency Influence in computing, is another adjustment factor used to strengthen the average value of node fitness The role of association strength, is the normalization factor used to adjust the values ​​in the denominator, It represents the sum of the absolute values ​​of the associated frequency differences of all other nodes connected to node i and node j.

[0024] formula: , The benefit of the formula is that it enhances the sensitivity of the model to the importance of nodes by weighting the association frequency and node fitness, as well as an adjustment term, thus improving the accuracy and practicality of the association matrix.

[0025] Detailed explanation of the formula and the process of formula calculation and derivation: set up =5, =3, =4, =0.5, =1.5, =2, n=10, ; The calculation process is as follows: ; The result shows that the correlation strength between node i and node j is 0.775, which reflects that the connection strength between the two nodes is higher than the average level. This result helps to identify and strengthen key node connections in the network.

[0026] See also Figure 3 , the specific steps for obtaining the sparse coefficient matrix are: Based on principal component analysis, the interdisciplinary correlation matrix is ​​decomposed into a potential feature matrix and a relationship coefficient matrix, the initial potential features and weight relationships are identified, and preliminary feature and coefficient matrices are generated; When performing eigendecomposition on the interdisciplinary association relationship matrix based on principal component analysis, it is first necessary to calculate the covariance matrix of the original data matrix. This process involves calculating the mean of each variable and summing the product of the difference between each pair of variables. Based on these covariance values, the covariance matrix is ​​constructed and the principal components are extracted by the eigenvalue decomposition method. These principal components correspond to the main variation directions in the data. Subsequently, the potential feature matrix and relationship coefficient matrix are formed by the extracted principal components and their corresponding eigenvalues. This method can effectively extract and represent key information and relationships in interdisciplinary fields, ensuring that important information is retained after data compression. The obtained potential feature matrix and relationship coefficient matrix provide basic data and analysis tools for subsequent steps.

[0027] From the preliminary feature and coefficient matrix, calculate the weight of each element, and set the weight threshold to remove non-critical feature items with weights lower than the threshold to obtain the adjusted relationship coefficient matrix; After calculating the element weights of the preliminary feature and coefficient matrix, low-weight non-critical features are eliminated according to the set threshold. In this process, the calculation of weights needs to be evaluated based on the size of the relationship coefficient, which involves statistical analysis and threshold setting. Through specific weight evaluation models, such as using variance analysis to determine which coefficients are statistically significant, we can determine which are the key features. This method of simplifying the matrix can effectively reduce computational complexity and improve processing speed, so that the processed relationship coefficient matrix can more concentratedly reflect the most important relationships and features.

[0028] Combine the adjusted relationship coefficient matrix with the latent feature matrix and apply the matrix reconstruction formula: ; Perform compression, calculate and generate a sparse coefficient matrix; Among them, S represents the sparse coefficient matrix, V represents the potential feature matrix, and D represents the adjusted relationship coefficient matrix. represents the weight parameter that strengthens the impact of key features, represents the adjustment coefficient for matrix sparsification, represents the transposed matrix of the latent feature matrix, Represents the diagonal elements of the relationship coefficient matrix D, Represents the sum of the diagonal elements of the relationship coefficient matrix D.

[0029] formula: ; The benefit of the formula is that it enhances the computational efficiency and data interpretation ability of the model by sparsely processing the matrix. This formula introduces the weight parameter and , not only the original information of the relationship coefficient matrix D is considered, but also the sum of the diagonal elements of the matrix is ​​added in a weighted manner, which enhances the main feature representation of the data, making the model more efficient when processing large-scale data.

[0030] Detailed explanation of the formula and the process of formula calculation and derivation: Assume that the potential feature matrix V is , the relationship coefficient matrix D is , weight parameter =0.6, = 0.4, the diagonal elements and is 0.8, then The calculation is: , Further substituting into the formula: , The results show that the sparse coefficient matrix S after weight adjustment and feature matrix reconstruction can more accurately reflect the main features and relationships in the original data, and is suitable for efficient processing and analysis of interdisciplinary data.

[0031] See also Figure 4 , the specific steps for obtaining the sparse feature representation matrix are: Based on the sparse coefficient matrix, determine whether the weight of each element in the matrix is ​​lower than the preset screening threshold, mark and remove the elements below the threshold, and retain the elements above the threshold to establish a preliminary screening matrix; Based on the sparse coefficient matrix, we first perform weighted screening of elements. By measuring the strength of association in the actual data set, we apply a preset threshold to each element. This threshold is set based on the average of historical data plus the standard deviation to ensure that only statistically significant connections are retained. Next, elements below this threshold are marked for elimination. Automated screening is performed using Python scripts, and the element position and weight of each operation are recorded to ensure traceability and transparency of operations. In this way, the generated preliminary screening matrix only contains important and representative connections, which improves the efficiency of data processing and the accuracy of the results.

[0032] Using the preliminary screening matrix, the remaining contact weight elements are iteratively screened, the contact items are called according to the weight components and retained according to the sparse threshold, using the formula: , Calculate the adjusted sparse connection weights to obtain the sparse connection matrix retained after screening; in, represents the contact weight after screening, represents the weight value of the i-th row and j-th column in the sparse coefficient matrix, T represents the preset weight screening threshold, Represents the ratio of the adjustment weight to the threshold, which is used to adjust the impact of the threshold. Represents the balance adjustment coefficient, which corrects the overall offset of the sparse matrix. represents an additional sparse adjustment factor, which is used to enhance the sparseness characteristics in the sparse matrix. Represents the square root form of the additional sparse adjustment term, which is used to nonlinearly correct the weight value; formula: , The formula is beneficial in that it can dynamically adjust the threshold and add adjustment factors such as , and , which can flexibly adapt to the variability of data, strengthens the retention of sparse connections and the elimination of non-critical connections, and increases the robustness of the model to abnormal data.

[0033] Detailed explanation of the formula and the process of formula calculation and derivation: Suppose there is an element =0.45, threshold T=0.5, adjustment coefficient =0.1, =0.05, =0.02, substitute into the formula: , This result indicates that the adjusted contact weights The value is 0.41769, which is slightly lower than the original weight, indicating that the connection is considered less important and may be screened out in subsequent processing. Retaining such a weight helps to accurately depict the structure of the sparse feature matrix and better reflect the core connection structure of the data.

[0034] Based on the sparse connection matrix retained after screening, the remaining connection items are integrated to obtain a sparse feature representation matrix.

[0035] Data mining technology is used to perform cluster analysis on these connection items and identify core connection groups. These groups represent the main structures in the data. In order to construct a sparse feature representation, these core groups are used as the basis, and graph theory algorithms are used to further refine the structure within each group to ensure that each step of the calculation is based on actual data rather than assumptions. The sparse feature representation matrix formed can effectively capture and describe the important features of the original data, while removing redundant and irrelevant information, optimizing the data processing process, and improving the quality and efficiency of data analysis.

[0036] See also Figure 5 , the steps to obtain the optimized inter-block coupling matrix are as follows: Based on the sparse feature representation matrix, the connection strength of each node pair is calculated one by one. By comparing whether the weight of each pair of connections is higher than the set coupling strength threshold, the node pairs that meet the conditions are marked as strongly coupled, and the node coupling relationship matrix is ​​obtained; This process not only requires precise data measurement, but also involves dynamic adjustment of weight thresholds. This method ensures that only the most critical node couplings are marked and retained, which has a crucial impact on subsequent network optimization and resource allocation. The node coupling relationship matrix formed will directly affect the structure and functional optimization of the entire network.

[0037] From the node coupling relationship matrix, the node pairs marked as strongly coupled are classified into the same block, and the coupling relationship of the node pairs in each block is analyzed using the formula: , Calculate the coupling degree between blocks; in, represents the coupling degree between block i and block j, represents the connection strength between node i and node k, represents the connection strength between node k and node j, represents the adjustment coefficient, which is used to avoid the situation where the denominator is zero, k represents the node index, and n represents the total number of nodes; formula: , The benefit of the formula is that by introducing the square root of the sum of squares, the formula can more accurately reflect the overall coupling strength between nodes, and by adding a small constant The situation where the denominator is zero is avoided, ensuring the stability and accuracy of the calculation.

[0038] Detailed explanation of the formula and the process of formula calculation and derivation: Assume there are nodes i and j in a network, and the connection strength between nodes i and k is = 0.3, the connection strength between nodes j and k is =0.4, threshold = 0.01, calculate the coupling degree between node i and node j: Molecular computing: , Denominator calculation: , , The results show that in this case the coupling degree between nodes i and j is 0.235, which means that there is a certain connection between the two, but it is not very strong and is suitable for further evaluation to determine whether they need to be considered and retained during the network optimization process.

[0039] According to the coupling degree between blocks, the connections are iteratively screened based on the coupling degree, and the connections with weak coupling are sparsely processed. Some connections are retained based on the coupling degree to construct an optimized inter-block coupling matrix.

[0040] After obtaining the preliminary inter-block coupling matrix, the coupling relationships in the matrix are further screened to remove those connections with weak coupling and retain those with strong coupling. This process involves quantitative analysis and evaluation of the coupling degree, which requires the collection of a large amount of network performance data and the determination of which connections do have a positive impact on the functionality and efficiency of the network through data analysis. The optimized inter-block coupling matrix not only reflects the most important nodes and connections in the network, but also provides reliable data support for further network optimization and management. This method ensures the effective utilization and optimization of network resources.

[0041] See also Figure 6 , the specific steps for obtaining the reconstructed interdisciplinary relationship matrix are: Based on the optimized inter-block coupling matrix, the connection relationship of each node is analyzed one by one. By calling the coupling data of multiple nodes, the node pairs are identified and marked based on the coupling relationship, and the node coupling adjustment matrix is ​​generated; Based on the optimized inter-block coupling matrix, the coupling relationship of each node is analyzed. Through systematic data collection and real-time monitoring, the actual coupling data between each node is obtained. These data are calculated based on the actual interaction frequency and intensity between nodes. Strongly associated nodes will be marked and classified according to the coupling intensity, and then a node coupling adjustment matrix will be formed in the database. This matrix records the dynamic connection status and change trend between nodes in detail, providing data support and decision-making basis for the next step of node connection adjustment.

[0042] From the node coupling adjustment matrix, the connection between nodes is analyzed using the formula: , Calculate the coupling degree, remove connections with low coupling degree, optimize the direct interaction between nodes, and obtain the adjusted node connection matrix; in, represents the coupling strength between node i and node j, represents the connection weight between node i and node k, represents the connection weight between node k and node j, represents a constant for adjusting the coupling calculation to avoid the denominator being zero. represents the sum of squared weights from node i to all other related nodes, Represents the sum of squared weights from node j to all other associated nodes; formula: , The usefulness of the formula is that it can accurately calculate the coupling strength between any two nodes, thereby determining whether to adjust their connection or remove redundant connections, thereby optimizing the structural efficiency of the entire network. If the coupling strength between node i and node j is lower than the set threshold, the connection will be judged as redundant and will be optimized or removed.

[0043] Detailed explanation of the formula and the process of formula calculation and derivation: Assume that the weights of nodes i and k are 0.4, and the weights of nodes k and j are 0.3, to prevent tiny values ​​from dividing by zero is 0.01, , Substituting into the formula, we get the coupling degree, , This result shows that the coupling relationship between node i and node j is of medium strength. If the value is lower than the threshold, the connection between them may need to be adjusted or deleted.

[0044] According to the adjusted node connection matrix, the nodes and blocks are reorganized, the node pairs with high coupling strength are grouped into the same block, the inter-block relationships are checked and optimized again, and the reconstructed interdisciplinary relationship matrix is ​​generated.

[0045] Based on the node connection matrix, the connection quality between nodes is further analyzed, and node connections with insufficient coupling are eliminated. At the same time, the new connection data is used to rebuild the relationship between nodes and blocks. The connection strength between blocks is adjusted through continuous testing and data feedback to ensure that the nodes in each block maintain efficient information exchange. At the same time, redundant connections are eliminated, thereby optimizing the overall network structure so that each node can maintain a low system complexity while ensuring information transmission efficiency.

[0046] A method for constructing an interdisciplinary knowledge graph is provided. The method for constructing an interdisciplinary knowledge graph is performed based on the above-mentioned interdisciplinary knowledge graph construction system and includes the following steps: S1: Conduct in-depth analysis of the nodes in the knowledge graph, select interdisciplinary nodes based on the interaction frequency between nodes, and screen the node set with high correlation by calculating the fit of multiple nodes, so as to construct an interdisciplinary correlation matrix. After the calculation is completed, an interdisciplinary correlation matrix is ​​generated; S2: Split the interdisciplinary association matrix into a potential feature matrix and a relationship coefficient matrix, analyze the relationship coefficients item by item, filter out non-critical feature items based on weight components, and obtain a sparse coefficient matrix after merging and adjusting; S3: Based on the sparse feature matrix, screen each element one by one, remove the weight elements below the preset threshold, retain the key connections, refine the feature representation, build a sparse structure, and obtain the optimized feature matrix; S4: Use the optimized feature matrix to analyze and reconstruct the graph nodes and block structures, mark highly coupled nodes and adjust the connection relationships, remove redundant connections, strengthen the relationship between nodes and blocks, and establish a reconstructed interdisciplinary relationship matrix.

[0047] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A cross-disciplinary knowledge graph construction system, characterized in that: The system comprises: The relationship matrix construction module analyzes the node relationships based on the nodes in the knowledge graph, selects interdisciplinary nodes to construct the relationship matrix according to the association frequency statistics, calculates the node fit, determines the associated node set, and generates the interdisciplinary association relationship matrix; The matrix feature decomposition module decomposes the interdisciplinary association relationship matrix into a potential feature matrix and a relationship coefficient matrix, analyzes the weights of the coefficient matrix elements, removes non-critical feature items based on the weights, merges and adjusts the matrices, and obtains a sparse coefficient matrix; The sparsity constraint optimization module selects elements whose weights are lower than a preset threshold element by element based on the sparse coefficient matrix, retains some connections according to the sparse threshold, and establishes a sparse feature representation matrix; The block structure adaptive optimization module calculates node connections one by one based on the sparse feature representation matrix, marks nodes with strong coupling relationships, classifies them into the same block, analyzes the coupling relationship between blocks, screens connections with weak coupling for sparse processing, retains the remaining structures, and establishes an optimized inter-block coupling matrix; Based on the optimized inter-block coupling matrix, the interdisciplinary knowledge mapping module reconstructs the graph nodes and block structures, analyzes the coupling relationship between nodes, adjusts node connections, removes redundant connections, checks the strength of inter-block relationships, reorganizes nodes and blocks, and generates a reconstructed interdisciplinary relationship matrix for intelligent recommendation, providing researchers with suggestions for interdisciplinary cooperation based on the correlation between disciplines.

2. The interdisciplinary knowledge graph construction system according to claim 1 is characterized in that: The interdisciplinary association relationship matrix includes association frequency, association node set, and node fitting degree; the sparse coefficient matrix includes potential feature matrix, key feature items, and relationship coefficient matrix; the sparse feature representation matrix includes sparse threshold, partial connection, and low weight element; the optimized inter-block coupling matrix includes block structure, adaptive block partitioning, and inter-block coupling relationship; the reconstructed interdisciplinary relationship matrix includes node connection and inter-block relationship strength.

3. The interdisciplinary knowledge graph construction system according to claim 2 is characterized in that: The steps for obtaining the interdisciplinary association relationship matrix are specifically as follows: Based on the knowledge graph, the interaction frequency between nodes is counted, and interdisciplinary nodes are screened. Nodes with association frequencies higher than the average are selected to generate an interdisciplinary node association frequency matrix. Using the interdisciplinary node association frequency matrix, the total association of multiple nodes is calculated, the centrality of each node is evaluated, and the node fitness vector is obtained; Using the node fitness vector and the interdisciplinary node association frequency matrix, an association matrix is ​​constructed by a fitness weighting method, using the formula: ； Calculate the association strength between node i and node j to obtain the interdisciplinary association matrix ; in, Represents the association frequency between node i and node j, which is a basic indicator to measure the frequency of interaction between two nodes. , Represent the fitness of nodes i and j, respectively, which reflects the criticality or centrality of each node in the interdisciplinary network, is the adjustment factor used to strengthen the associated frequency Influence in computing, is another adjustment factor used to strengthen the average value of node fitness The role of association strength, is the normalization factor used to adjust the values ​​in the denominator, It represents the sum of the absolute values ​​of the associated frequency differences of all other nodes connected to node i and node j.

4. The interdisciplinary knowledge graph construction system according to claim 3 is characterized in that: The steps of obtaining the sparse coefficient matrix are specifically as follows: Based on principal component analysis, the interdisciplinary association relationship matrix is ​​decomposed into a potential feature matrix and a relationship coefficient matrix, initial potential features and weight relationships are identified, and preliminary feature and coefficient matrices are generated; Calculate the weight of each element from the preliminary feature and coefficient matrix, and set a weight threshold to remove non-critical feature items with weights lower than the threshold to obtain an adjusted relationship coefficient matrix; The adjusted relationship coefficient matrix is ​​combined with the potential feature matrix, and the matrix reconstruction formula is applied: ； Perform compression, calculate and generate a sparse coefficient matrix; Among them, S represents the sparse coefficient matrix, V represents the potential feature matrix, and D represents the adjusted relationship coefficient matrix. represents the weight parameter that strengthens the impact of key features, represents the adjustment coefficient for matrix sparsification, represents the transposed matrix of the latent feature matrix, Represents the diagonal elements of the relationship coefficient matrix D, Represents the sum of the diagonal elements of the relationship coefficient matrix D.

5. The interdisciplinary knowledge graph construction system according to claim 4 is characterized in that: The steps of obtaining the sparse feature representation matrix are specifically as follows: Based on the sparse coefficient matrix, determine whether the weight of each element in the matrix is ​​lower than a preset screening threshold, mark and remove the elements below the threshold, and retain the elements above the threshold to establish a preliminary screening matrix; Using the preliminary screening matrix, iterative screening is performed on the remaining contact weight elements, contact items are called according to the weight components and retained according to the sparse threshold, using the formula: ， Calculate the adjusted sparse connection weights to obtain the sparse connection matrix retained after screening; in, represents the contact weight after screening, represents the weight value of the i-th row and j-th column in the sparse coefficient matrix, T represents the preset weight screening threshold, Represents the ratio of the adjustment weight to the threshold, which is used to adjust the impact of the threshold. Represents the balance adjustment coefficient, which corrects the overall offset of the sparse matrix. represents an additional sparse adjustment factor, which is used to enhance the sparseness characteristics in the sparse matrix. Represents the square root form of the additional sparse adjustment term, which is used to nonlinearly correct the weight value; Based on the sparse connection matrix retained after the screening, the remaining connection items are integrated to obtain a sparse feature representation matrix.

6. The interdisciplinary knowledge graph construction system according to claim 5, characterized in that: The steps for obtaining the optimized inter-block coupling matrix are specifically as follows: Based on the sparse feature representation matrix, the connection strength of each node pair is calculated one by one, and by comparing whether the weight of each pair of connections is higher than a set coupling strength threshold, the node pairs that meet the conditions are marked as strongly coupled, and a node coupling relationship matrix is ​​obtained; From the node coupling relationship matrix, the node pairs marked as strongly coupled are classified into the same block, and the coupling relationship of the node pairs in each block is analyzed using the formula: ， Calculate the coupling degree between blocks; in, represents the coupling degree between block i and block j, represents the connection strength between node i and node k, represents the connection strength between node k and node j, represents the adjustment coefficient, which is used to avoid the situation where the denominator is zero, k represents the node index, and n represents the total number of nodes; According to the coupling degree between the blocks, connections are iteratively screened based on the coupling degree, and connections with weak coupling degree are sparsely processed, and some connections are retained based on the coupling degree to construct an optimized inter-block coupling matrix.

7. The interdisciplinary knowledge graph construction system according to claim 6, characterized in that: The steps for obtaining the reconstructed interdisciplinary relationship matrix are specifically as follows: Based on the optimized inter-block coupling matrix, the connection relationship of each node is analyzed one by one, and by calling the coupling data of multiple nodes, the node pairs are identified and marked based on the coupling relationship, and the node coupling adjustment matrix is ​​generated; From the node coupling adjustment matrix, the connection between nodes is analyzed using the formula: ， Calculate the coupling degree, remove connections with low coupling degree, optimize the direct interaction between nodes, and obtain the adjusted node connection matrix; in, represents the coupling strength between node i and node j, represents the connection weight between node i and node k, represents the connection weight between node k and node j, represents a constant for adjusting the coupling calculation to avoid the denominator being zero. represents the sum of squared weights from node i to all other related nodes, Represents the sum of squared weights from node j to all other associated nodes; According to the adjusted node connection matrix, the nodes and blocks are reorganized, the node pairs with high coupling strength are grouped into the same block, the inter-block relationship is checked and optimized again, and the reconstructed interdisciplinary relationship matrix is ​​generated.

8. A method for constructing an interdisciplinary knowledge graph, characterized in that: The interdisciplinary knowledge graph construction system according to any one of claims 1 to 7 is implemented, comprising the following steps: Conduct in-depth analysis of the nodes in the knowledge graph, select interdisciplinary nodes based on the interaction frequency between nodes, and filter out highly correlated node sets by calculating the fit of multiple nodes to construct an interdisciplinary correlation matrix. After the calculation is completed, an interdisciplinary correlation matrix is ​​generated. Splitting the interdisciplinary association matrix into a potential feature matrix and a relationship coefficient matrix, analyzing the relationship coefficients item by item, filtering out non-critical feature items based on weight components, and merging and adjusting to obtain a sparse coefficient matrix; Based on the sparse coefficient matrix, screen each element one by one, remove weight elements below a preset threshold, retain key connections, refine feature representation, construct a sparse structure, and obtain an optimized feature matrix; The optimized feature matrix is ​​used to analyze and reconstruct the graph nodes and block structures, mark high-coupling nodes and adjust the connection relationships, remove redundant connections, strengthen the relationship between nodes and blocks, and establish a reconstructed interdisciplinary relationship matrix.

Images