Knowledge management system and method

Abstract

The present invention provides a university knowledge management system and method, the system includes a display module, a processor, and the program modules that can be run on the processor include a knowledge acquisition module, a knowledge parsing module, an input module and an evaluation module. Fixed sub-blocks are obtained by using matrix theory, a heterogeneous network is constructed by the knowledge parsing module, and a multi-layer attention mechanism from nodes, semantics to sub-graphs on the heterogeneous network is constructed based on the knowledge data of the knowledge acquisition module. Finally, the sub-graph normalization coefficient is used as the basis for block division to obtain dynamic blocks and blockchain knowledge management. A diversified, safe and timely school knowledge management solution is realized. It can reflect the knowledge exchange and research and development status between various units of schools in various places. At the same time, the selection of different types of meta-paths can obtain different dynamic blocks, thereby providing users with a wide range of customized knowledge data blockchain implementation forms.

Description

Technical Field

[0001] The invention relates to a university knowledge management system and method, belonging to the field of intelligent management. Background Art

[0002] The knowledge management involved in universities includes multiple structures, including social relationship knowledge (organizational structure), university subject knowledge, various places of access failure system, scientific research projects, each of which includes more subtle knowledge structures, such as daily work, study arrangements, knowledge management problems arising from the cooperation between various colleges, departments and research groups. The knowledge data processed by these problems is dynamic. Therefore, how to structure such complex knowledge data. Summary of the invention

[0003] Based on the complexity of multi-structure knowledge data processing in the above-mentioned university. The present invention proposes a knowledge management system, including a display module, a processor, and a program module that can be run on the processor, including a knowledge acquisition module, which is used to acquire knowledge from school staff and students and record the spatial location and time of knowledge generation. The knowledge includes knowledge related to equipment related to the knowledge, specifically including the name of the equipment, the school staff and students and / or users to which it belongs, the use and technical principles related to the equipment, as well as scientific and technological resources, social and humanities resources, specifically including documents, pictures, videos, and audios. The knowledge also includes the name of the equipment, the school staff and students and / or users to which it belongs, the use and technical principles related to the equipment, the use records, historical and real-time indicators and environmental parameters, equipment management information, and various school departments, research groups, laboratories, classroom libraries, playgrounds and gymnasiums, network information centers, dormitories (seven types of units) and administrative and asset establishment and management knowledge. The data related to the knowledge forms knowledge data.

[0004] The knowledge parsing module is connected to the knowledge acquisition module and is used to construct fixed blocks and dynamic blocks, as well as to construct heterogeneous graph neural networks (HGNN), wherein the fixed blocks involve departments 、Research Group ,laboratory , classroom library , playground and gymnasium , Network Information Center ,dormitory , all knowledge data behaviors and knowledge management related to these units are carried out in fixed blocks and define the matrix , where the second digit 1 of the element subscript represents the school, and other digits represent other schools, and two action matrices are defined (1), (2), where Defined as the role of the left and right matrices, is the transposition symbol. The interaction of each row and column of the left and right matrices forms a seven-element block matrix in which each seven elements on the right side of the equation are not summed (different from the matrix multiplication rule, which requires summation). Each element represents the result of the interaction between each row and column of the left and right matrices, and represents the association of knowledge data between the elements of each row and column. Each block matrix is ​​represented as a fixed block, and each element in the block matrix is ​​a fixed subblock. At the same time, the unit is defined. Substituting the elements in formula (1) and formula (2) to represent each element, , Number the school. is a natural number, which indicates the unit number of the same type of unit. Different numbers indicate different units of the same type. That is, each element in formula (1) and formula (2) All above and below the standard To further express. This is subdivided into the relationship between specific different units of the same or different types. For example, if they are both classrooms, the relationship can be between the first classroom and the second classroom. If classrooms and dormitories are different types of units, the relationship can be between the first classroom and the first dormitory.

[0005] It should be emphasized that the so-called association, for example, the class arrangement, the students of the first dormitory of this school take the class in the first classroom. In this way, the knowledge data of the course arrangement is associated with the knowledge data of the rules and regulations of the first classroom (such as the precautions for the use of the projector, and the last student to leave the evening self-study consciously turning off the lights, etc.). The system can be connected to various devices (such as classroom lights, classroom and dormitory access control, campus network equipment, etc.). When the students of the first dormitory violate the rules and regulations, such as being the last to leave and forgetting to turn off the lights, it will be discovered on the system. If the degree of association given by the attention mechanism of the meta-path from the first dormitory to the first classroom (see below for details) is found in a period of time, that is, the normalization coefficient is very large and exceeds the threshold; on the other hand, the time when a student in the first dormitory connects to the Internet with the network information center on the meta-path also notices that other students who intersect with each other have a greater degree of association when they leave the classroom last, that is, the student is on the meta-path of the first classroom-first dormitory-network information center, and in the period of attention, it is reflected as a greater degree of association. Therefore, in the analysis of the associated behavior process of the student in the first dormitory forgetting to turn off the lights and returning to the dormitory to surf the Internet, it is found that the gateway and other behaviors are most likely the habits of a specific student in the first dormitory, so that he can be investigated and educated and reminded to pay attention.

[0006] for , for example, in formula (1), if the element represents the classroom of this school, then the first classroom of this school can be expressed as , the second classroom of our school can be expressed as ,by represents the library of our school. The first library of our school can be represented as , and so on. The interaction between the left and right matrix elements Indicates the relationship between the knowledge data of the college 1 and the college 2 of the school. For example, data access, download, upload, dissemination, or organizational adjustment, various affairs notifications, personnel transfers and other knowledge data behaviors have occurred between the colleges of the school. As long as the knowledge data behaviors related to the two colleges and departments are involved, as well as management, they are all in To express, for example It represents the relationship between the departments and research groups of the school. The above two groups of interaction matrices form the relationship between similar units within the school and between different schools, as well as the relationship between similar or different units within the school.

[0007] Optionally, define the interaction between the two elements as commutative, that is, , , define the two action matrices to be merged and deduplicated (3), Defined as the deduplication symbol of the fixed sub-block, the removed sub-block is represented by 0 block, forming the deduplication matrix of the fixed block, and the unit is also defined The elements are further substituted into formula (3) to represent each element.

[0008] The knowledge parsing module is also used to construct an attention mechanism (AM) based on a heterogeneous graph neural network using the knowledge and spatial locations received from school staff and students. After the construction is completed, the knowledge context is retrieved again based on the new input (in most cases, the new input is a search operation, that is, the relevant knowledge input is searched for any relevant knowledge information of interest, or there are other purposes of use, learning, organization, etc.), and the analysis results of the attention mechanism are given based on the knowledge context. Based on the analysis results, the heterogeneous graph neural network is dynamically segmented in fixed blocks, and data management passwords are set for each dynamic block based on the dynamically segmented graph neural network, and HGNN and AM are updated within a predetermined time period, and the dynamic blocks and the data management passwords are continuously updated accordingly, and data management passwords for fixed blocks are set for knowledge management.

[0009] It is understandable that the two action matrices and the deduplication matrix still cannot include other situations, such as the relationship between different types of units in different institutions, so dynamic blocks fill this gap. Of course, dynamic blocks also include any situation in the two action matrices and the deduplication matrix. The dynamic depends on the analysis results of the attention mechanism analyzed later. Dynamic blocks are the dynamic blockization of fixed blocks based on the analysis results.

[0010] The system also includes an input module connected to a computer network, for receiving new knowledge input by a user, inputting the knowledge into the knowledge parsing module through the computer network, and clustering the knowledge into a preset target domain with a domain label, and when the user inputs new knowledge, it is necessary to select a target domain for the new knowledge input, and select a target domain label.

[0011] The evaluation module is used to conduct a comprehensive analysis of the retrieved knowledge context and the analysis results of the attention mechanism, and give a final evaluation that is suitable for the user's true search meaning.

[0012] The display module is used to display the analysis results of the knowledge context, attention mechanism and comprehensive analysis results, wherein:

[0013] The school staff and students regularly input the knowledge generated according to the classification into the knowledge acquisition module and transmit it to the knowledge analysis module for the update.

[0014] The data application model of the blockchain is implemented in each fixed block and dynamic block, and when the dynamic block is updated, the dynamic blockchain is updated.

[0015] Among them, the construction of a heterogeneous graph neural network and the use of the knowledge and spatial locations input by school staff and students to build an attention mechanism based on the heterogeneous graph neural network include:

[0016] S1 establishes a geographical map of the school, marks the spatial locations on the geographical map of the school as geographical nodes of the heterogeneous graph neural network, and constructs a heterogeneous graph network based on real spatial locations;

[0017] S2 uses each fixed sub-block as a block node, or different elements within each fixed sub-block as element nodes, to construct at least one fixed sub-block heterogeneous graph network based on geographical nodes;

[0018] S3 constructs multiple types of meta-paths in the heterogeneous graph network based on real space locations and the at least one fixed sub-block heterogeneous graph network;

[0019] S4 trains and constructs multi-type node attention mechanism, semantic attention mechanism, and subgraph attention mechanism for knowledge data for each type of meta-path in the multi-type meta-path.

[0020] The knowledge context is retrieved based on the newly input knowledge, and the analysis result of the attention mechanism is given based on the knowledge context. The specific method is:

[0021] S5 inputs the knowledge into the trained AM to obtain the predicted knowledge data classification.

[0022] S6 draws the knowledge context according to the corresponding spatial location and generation time of the knowledge in the predicted knowledge data classification, and retrieves the drawn knowledge context.

[0023] S7: marking nodes and / or subgraphs corresponding to the new input knowledge in the drawn knowledge context.

[0024] The comprehensive analysis includes statistical analysis of at least one of the following types of distributions: a first distribution on the geographical map of the school in the predicted knowledge field corresponding to the knowledge and at least one other field, a second distribution in at least one fixed sub-block heterogeneous graph network, and a mixed distribution of the first distribution and the second distribution.

[0025] Based on the analysis results, the heterogeneous graph neural network is dynamically segmented in fixed blocks, and at the same time, the data management password of each dynamic block is set based on the dynamically segmented graph neural network, which is performed after the predicted knowledge data classification is obtained, specifically including:

[0026] S5-1 Normalize the coefficients of the subgraph (calculated by AM) for segment clustering;

[0027] S5-2: in the clustered partitions, for the fixed sub-blocks belonging to the block nodes in the sub-graph, or the elements belonging to the element nodes in the sub-graph, these fixed sub-blocks or elements are classified into a dynamic block to complete the dynamic blockization, and a heterogeneous block graph neural network is constructed based on all the block nodes inside the dynamic block, and a heterogeneous block attention mechanism (HBAM) at the node level, semantic level, and sub-graph level is constructed based on the heterogeneous block graph neural network;

[0028] S5-3 is the block node number inside the heterogeneous block graph neural network. A preset number of block node numbers are randomly selected, and the normalization coefficient corresponding to the segment (calculated by HBAM) are randomly sorted together to form the data management password of the dynamic block.

[0029] The HGNN and AM are updated within a predetermined time period, and the blocks and the data management password are continuously updated accordingly, specifically including:

[0030] Q1 continuously updates the nodes of HGNN and AM within a predetermined time period based on the knowledge and spatial locations input by school staff and students received within the period;

[0031] Q2 updates the normalization coefficient based on the updated AM to predict the newly input knowledge , and updating the segment clustering, thereby updating the block;

[0032] Q3 randomly selects a preset number of node numbers and the normalization coefficients corresponding to the updated segments Randomly sort together and update the data management password of the updated dynamic block.

[0033] Preferably, knowledge management of the dynamic block requires verification using both the data management password of the dynamic block and the data management password of the corresponding fixed block in the dynamic block.

[0034] Another object of the present invention is to provide a knowledge management method, including a mobile smart terminal and a desktop terminal with the display module held by students and school staff respectively, and an application app and an application system with a program module are installed on the mobile smart terminal and the desktop terminal respectively to realize the input of knowledge, realize the new input, and obtain the analysis results, comprehensive analysis results and knowledge context; realize the construction of various heterogeneous neural networks and various attention mechanisms on the application system, and complete the update, and the setting and management of passwords.

[0035] Therefore, we use matrix theory to deduplicate the seven types of units in the school and fix them into blocks, build a heterogeneous neural network based on geographical location and fixed blocks, and use knowledge data for the construction of AM and HBAM to obtain dynamic blocks and update dynamic blocks. At the same time, we use the management passwords of fixed blocks and dynamic blocks for knowledge management, and realize a diversified, safe and timely school knowledge management solution. It can reflect the knowledge exchange and research and development status between various units in schools across the country. According to the selection of different types of meta-paths, different dynamic blocks can be obtained, thus providing users with a wide range of customized knowledge data blockchain implementation forms. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A schematic diagram of the structure of a knowledge management system.

[0037] Figure 2 A schematic diagram of the functions of various program modules that can be completed by the mobile phone held by students in a knowledge management method.

[0038] Figure 3Construct a heterogeneous graph neural network, and use the knowledge and spatial locations input by school staff and students to construct a flowchart of the attention mechanism method based on the heterogeneous graph neural network.

[0039] Figure 4 Schematic diagram of the construction process of fixed blocks.

[0040] Figure 5 yes Figure 4 The third type of closed-loop meta-path virtual concrete diagram of block nodes and element nodes given by the deduplication matrix,

[0041] Figure 6 A flowchart of retrieving the knowledge context based on the newly input knowledge and providing the analysis result of the attention mechanism based on the knowledge context,

[0042] Figure 7 Schematic diagram of two types of dynamic block clusters formed by fixed sub-blocks.

[0043] Figure 8 Schematic diagram of two types of dynamic block clusters formed by elements. DETAILED DESCRIPTION

[0044] Figure 1 The present invention provides a structure of a knowledge management system, specifically a university multi-knowledge management system based on heterogeneous blockchain, including a display module and a processor. The program modules that can be run on the processor include a knowledge acquisition module for acquiring knowledge from school staff and students and recording the spatial location and time of knowledge generation; a knowledge parsing module, connected to the knowledge acquisition module, for constructing fixed blocks and dynamic blocks, and constructing a heterogeneous graph neural network, constructing an attention mechanism based on the heterogeneous graph neural network, retrieving knowledge context, and providing an analysis result of the attention mechanism based on the knowledge context, and dynamically blocking the heterogeneous graph neural network in fixed blocks based on the analysis result, and setting each block based on the dynamically blocked graph neural network. The data management password of the dynamic block is updated within a predetermined time period, and the dynamic block and the data management password are continuously updated accordingly, and the data management password of the fixed block is set for knowledge management; the input module is connected to the computer network, and is used to receive the newly input knowledge of the user, and input it into the knowledge analysis module through the computer network, and the knowledge is clustered in the preset target field with the field label, and when the user inputs new knowledge, it is necessary to select the target field for the newly input knowledge, and select the field label of the target; and the evaluation module is used to comprehensively analyze the analysis results of the retrieved knowledge context and the attention mechanism, and give a final evaluation that is suitable for the real meaning of the user's search. Specifically, it includes comparing the predicted knowledge data classification with the field label, and taking the comparison result as the final evaluation.

[0045] A knowledge management method, comprising a mobile intelligent terminal (a mobile phone, such as a mobile phone) with the display module held by a student and a school staff respectively. Figure 2 as shown) and desktop terminals (including display modules and processors, such as Figure 1 ), an application app and an application system with program modules are installed on the mobile intelligent terminal and the desktop terminal respectively to realize the input of knowledge, realize the new input, and obtain the analysis results, comprehensive analysis results and knowledge context; realize the construction of various heterogeneous neural networks and various attention mechanisms on the application system, and complete the update, and the setting and management of passwords ( Figure 1 ).

[0046] Figure 3 A method for fixed block division is given, involving departments 、Research Group ,laboratory , classroom library , playground and gymnasium , Network Information Center ,dormitory ,All knowledge data behaviors and knowledge management related to these units are carried out in fixed blocks. Figure 3 Define the matrix , get the left and right matrix action form ( represents the left matrix, represents the right matrix) , we get two action matrices (1),

[0047] (2),

[0048] Define the interaction between two elements to have a commutative law, merge the two action matrices to remove duplicates and obtain the following deduplication matrix:

[0049] (4), where the fixed sub-blocks of the deduplication are given as 0 blocks. The elements are further substituted into formula (4) to represent each element.

[0050] Among them, Figure 3 and 4 As shown in the figure, a heterogeneous graph neural network is constructed, and the knowledge and spatial locations input by school staff and students are used to construct an attention mechanism based on a heterogeneous graph neural network, including:

[0051] S1 establishes a geographical map of colleges and universities, marks the spatial locations on the geographical map of colleges and universities as geographical nodes of the heterogeneous graph neural network, and constructs a heterogeneous graph network based on real spatial locations; this step describes the geographical distribution of colleges and universities, and forms a heterogeneous network with colleges and universities as nodes, which is called a heterogeneous graph network based on real spatial locations, reflecting the distribution of knowledge data associations in the spatial locations of colleges and universities, that is, the first distribution.

[0052] S2 takes each fixed sub-block as a block node, or different elements within each fixed sub-block as element nodes, and constructs at least one fixed sub-block heterogeneous graph network based on geographical nodes; it reflects the second distribution of knowledge data association clustering in the knowledge field.

[0053] Figure 4 The block nodes are indicated by long red underlines, and one of the underlined nodes is given as an example to represent a fixed sub-block. The block node of ; the cyan short underline represents the element node, Figure 4 The two elements are given in and The element node. Figure 5 The schematic structure of a heterogeneous graph network with fixed sub-blocks of block nodes and element nodes is given in . In fact, a heterogeneous network is an abstract logical structure. Figure 5 In the text, a virtual image is given for the sake of image representation. In mathematics, it is a graph structure in which nodes are connected by edges.

[0054] S3 constructs three types of meta-paths in the heterogeneous graph network based on real space locations and the at least one fixed sub-block heterogeneous graph network. Figure 5 The associations between different block nodes or element nodes form two types of meta-paths, which are graphically represented as nodes connected to each other by edges. Paths can be open or closed. Figure 5 The figure is an exemplary closed loop. Of course, the meta-path can also be a heterogeneous graph neural network composed of element nodes and block nodes. Figure 5 The thick black straight line in the figure represents the third type of closed-loop meta-path formed between two element nodes and two block nodes, forming a mixed distribution of knowledge domain clusters, reflecting the mixed distribution of knowledge data associations in knowledge domains and the geographical locations of institutions.

[0055] S4 targets three types of closed-loop meta-paths (e.g. Figure 5 Each type of closed-loop meta-path in the meta-path (a meta-path consisting of simple element nodes or block nodes, or a closed-loop meta-path formed between two element nodes and two block nodes) is trained to construct multi-type node attention mechanisms, semantic attention mechanisms, and subgraph attention mechanisms for knowledge data.

[0056] like Figure 6, retrieve the knowledge context based on the newly input knowledge, and give the analysis result of the attention mechanism based on the knowledge context. The specific method is:

[0057] S5 inputs the knowledge into the trained AM to obtain the predicted knowledge data classification, that is, through Figure 1 The input module of the system, students or school staff as system users, input knowledge, and the knowledge parsing module performs parsing in this step to obtain the knowledge classification corresponding to the input;

[0058] In the predicted knowledge data classification, S6 draws the knowledge context according to the corresponding spatial location and generation time of the knowledge, and calls the drawn knowledge context, which includes the context of knowledge in time and space. For example, by inputting the keywords of organic compound synthesis with the help of the input module, the output is the context of the field in time and space, showing the development process of the knowledge field.

[0059] S7 marks the nodes and / or subgraphs corresponding to the new input knowledge in the drawn knowledge context. This step is to mark the nodes and / or subgraphs in the corresponding heterogeneous graph on each node in the context. It is a visualization scheme based on a display module on the college geographic map.

[0060] The comprehensive analysis includes statistical analysis of at least one of the following types of distributions: a first distribution on the geographical map of the school in the predicted knowledge field corresponding to the knowledge and at least one other field, a second distribution in at least one fixed sub-block heterogeneous graph network, and a mixed distribution of the first distribution and the second distribution.

[0061] Based on the analysis results, the heterogeneous graph neural network is dynamically segmented in fixed blocks, and at the same time, the data management password of each dynamic block is set based on the dynamically segmented graph neural network, which is performed after the predicted knowledge data classification is obtained, specifically including:

[0062] S5-1 Normalize the coefficients of the subgraph (calculated by AM) for segment clustering;

[0063] S5-2: in the clustered partitions, for the fixed sub-blocks belonging to the block nodes in the sub-graph, or the elements belonging to the element nodes in the sub-graph, these fixed sub-blocks or elements are classified into a dynamic block to complete the dynamic blockization, and a heterogeneous block graph neural network is constructed based on all the block nodes inside the dynamic block, and a heterogeneous block attention mechanism (HBAM) at the node level, semantic level, and sub-graph level is constructed based on the heterogeneous block graph neural network;

[0064] Figure 4 The cyan rectangle is shown in the figure as an example. , and red rectangle , The fixed sub-blocks of the representatives form representatives of the two dynamic blocks, Figure 7 For two dynamic blocks, Block 1 and Block 2 are clustered according to segments as an example, as for the case of fixed sub-blocks belonging to block nodes in the sub-graph, and as for the case of elements belonging to element nodes in the sub-graph, as thick cyan and red squares. Figure 4 The cyan square is shown as an example. , and red square , The elements represented form the representatives of the two dynamic blocks. Figure 8 For the two dynamic blocks, block 3 and block 4 are clustered according to the segments. Depending on the type of meta-path selected by the user, the dynamic blocks also present different block clustering types, either represented by fixed sub-blocks or represented by elements. Figure 4 The three types of closed-loop meta-paths shown;

[0065] S5-3 is the block node number inside the heterogeneous block graph neural network. A preset number of block node numbers are randomly selected, and the normalization coefficient corresponding to the segment (calculated by HBAM) are randomly sorted together to form the data management password of the dynamic block.

[0066] In the present invention, the heterogeneous graph neural network is dynamically segmented in fixed blocks based on the analysis results, which are often the analysis results obtained by analyzing the new input of users accumulated for a certain period of time and within a preset time period. Therefore, there are dynamic blocks that change over time in the fixed blocks. Figure 4 This is well illustrated. Under the premise that the division definition of the fixed block remains unchanged, the dynamic combination of elements corresponding to the sub-blocks in the fixed block caused by the above time forms a dynamic block.

[0067] The HGNN and AM are updated within a predetermined time period, and the blocks and the data management password are continuously updated accordingly, specifically including:

[0068] Q1 continuously updates the nodes of HGNN and AM within a predetermined time period based on the knowledge and spatial locations input by school staff and students received within the period;

[0069] Q2 updates the normalization coefficient based on the updated AM to predict the newly input knowledge , and updating the segment clustering, thereby updating the block;

[0070] Q3 randomly selects a preset number of node numbers and the normalization coefficients corresponding to the updated segments Randomly sort together and update the data management password of the updated dynamic block.

[0071] Finally, the data application model of the blockchain is implemented in each fixed block and dynamic block, and when the dynamic block is updated, the dynamic blockchain is updated.

Claims

1. A knowledge management system, characterized in that: It includes a display module and a processor. The program modules that can be run on the processor include a knowledge acquisition module, a knowledge analysis module, an input module, and an evaluation module, wherein: The knowledge acquisition module is used to acquire knowledge from school staff and students and record the spatial location and time of knowledge generation. The knowledge includes knowledge related to equipment related to the knowledge, specifically including scientific and technological resources, social and humanities resources, specifically including documents, pictures, videos, and audio. The knowledge also includes the name of the equipment, the school staff and students and / or users to whom it belongs, the purpose related to the equipment, technical principles, usage records, historical and real-time indicators and environmental parameters, equipment management information, and various school departments, research groups, laboratories, classrooms and libraries, playgrounds and gymnasiums, network information centers, dormitories, and seven types of unit rules and regulations, as well as administrative and asset establishment and management knowledge; The knowledge parsing module is connected to the knowledge acquisition module to construct fixed blocks and dynamic blocks, and to construct a heterogeneous graph neural network (HGNN), wherein the fixed blocks involve departments 、Research Group ,laboratory , classroom library , playground and gymnasium , Network Information Center ,dormitory , all knowledge data behaviors and knowledge management related to these units are carried out in fixed blocks and define the matrix , where the second digit 1 of the element subscript represents the school, and other digits represent other schools, and two action matrices are defined (1), (2), where Defined as the role of the left and right matrices, is the transposition symbol. The interaction between each row and column of the left and right matrices forms a seven-element block matrix without summing each seven element on the right side of the equation. Each element represents the result of the interaction between each row and column element of the left and right matrices, and represents the association of knowledge data between each row and column element. Each block matrix is ​​represented as a fixed block, and each element in the block matrix is ​​a fixed sub-block. At the same time, the unit is defined. Substituting the elements in formula (1) and formula (2) to represent each element, , Number the school. It is a natural number, which indicates the unit number of the same type of units. Different numbers indicate different units of the same type. The knowledge parsing module is also used to construct an attention mechanism (AM) based on a heterogeneous graph neural network using the knowledge and spatial locations received from school staff and students, and after the construction is completed, the knowledge context is retrieved again based on the newly input knowledge, and the analysis results of the attention mechanism are given based on the knowledge context, and the heterogeneous graph neural network is dynamically segmented in fixed blocks based on the analysis results, and the data management password of each dynamic block is set based on the dynamically segmented graph neural network, and the HGNN and AM are updated within a predetermined time period, and the dynamic blocks and the data management password are continuously updated accordingly, and the data management password of the fixed block is set to perform knowledge management; The input module is connected to the computer network and is used to receive the newly input knowledge of the user, and input it into the knowledge parsing module through the computer network. The knowledge is clustered in a preset target domain with a domain label. When the user newly inputs knowledge, it is necessary to select a target domain for the newly input knowledge and select the target domain label. The evaluation module is used to conduct a comprehensive analysis of the retrieved knowledge context and the analysis results of the attention mechanism, and give a final evaluation that is suitable for the real meaning of the user's search, including comparing the predicted knowledge data classification and the domain label, and taking the comparison result as the final evaluation, wherein: The display module is used to display the analysis results of the knowledge context, attention mechanism and comprehensive analysis results, wherein: The school staff and students regularly input the knowledge generated according to the classification into the knowledge acquisition module and transmit it to the knowledge analysis module for the update. Implement the data application mode of blockchain in each fixed block and dynamic block, and update the dynamic blockchain after the dynamic block is updated; Construct a heterogeneous graph neural network and use the knowledge and spatial locations input by school staff and students to construct an attention mechanism based on the heterogeneous graph neural network, including: S1 establishes a geographical map of the school, marks the spatial locations on the geographical map of the school as geographical nodes of the heterogeneous graph neural network, and constructs a heterogeneous graph network based on real spatial locations; S2 uses each fixed sub-block as a block node, or different elements within each fixed sub-block as element nodes, to construct at least one fixed sub-block heterogeneous graph network based on geographical nodes; S3 constructs multiple types of meta-paths in the heterogeneous graph network based on real space locations and the at least one fixed sub-block heterogeneous graph network; S4 trains the multi-type node attention mechanism, semantic attention mechanism, and subgraph attention mechanism for constructing knowledge data for each type of meta-path in the multi-type meta-path; The knowledge context is retrieved based on the newly input knowledge, and the analysis result of the attention mechanism is given based on the knowledge context. The specific method is: S5 inputs the knowledge into the trained AM to obtain the predicted knowledge data classification. S6 draws the knowledge context according to the corresponding spatial location and generation time of the knowledge in the predicted knowledge data classification, and retrieves the drawn knowledge context. S7 marks the nodes and / or subgraphs corresponding to the new input knowledge in the drawn knowledge context; Based on the analysis results, the heterogeneous graph neural network is dynamically segmented in fixed blocks, and at the same time, the data management password of each dynamic block is set based on the dynamically segmented graph neural network, which is performed after the predicted knowledge data classification is obtained, specifically including: S5-1 Normalize the coefficients of the subgraph Perform segment clustering; S5-2: in the clustered partitions, for fixed sub-blocks belonging to block nodes in the sub-graph, or elements belonging to element nodes in the sub-graph, these fixed sub-blocks or elements are classified into a dynamic block to complete dynamic blockization, and a heterogeneous block graph neural network is constructed based on all block nodes inside the dynamic block, and a heterogeneous block attention mechanism (HBAM) at the node level, semantic level, and sub-graph level is constructed based on the heterogeneous block graph neural network; S5-3 is the block node number inside the heterogeneous block graph neural network. A preset number of block node numbers are randomly selected, and the normalization coefficient corresponding to the segment They are randomly sorted together to form the data management password of the dynamic block.

2. The system according to claim 1, characterized in that Define the interaction between two elements to be commutative, that is, , , define the two action matrices to be merged and deduplicated (4), Defined as the deduplication symbol of the fixed sub-block, the removed sub-block is represented by 0 block, forming the deduplication matrix of the fixed block, and the unit is also defined The elements are further substituted into formula (4) to represent each element.

3. The system according to claim 1, characterized in that The comprehensive analysis includes statistical analysis of at least one of the following types of distributions: a first distribution on the geographical map of the school in the predicted knowledge field corresponding to the knowledge and at least one other field, a second distribution in at least one fixed sub-block heterogeneous graph network, and a mixed distribution of the first distribution and the second distribution.

4. The system according to claim 1, characterized in that The HGNN and AM are updated within a predetermined time period, and the blocks and the data management password are continuously updated accordingly, specifically including: Q1 continuously updates the nodes of HGNN and AM within a predetermined time period based on the knowledge and spatial locations input by school staff and students received within the period; Q2 updates the normalization coefficient based on the updated AM to predict the newly input knowledge , and updating the segment clustering, thereby updating the block; Q3 randomly selects a preset number of node numbers and the normalization coefficients corresponding to the updated segments Randomly sort together and update the data management password of the updated dynamic block.

5. The system according to claim 4, characterized in that To manage the knowledge of dynamic blocks, it is necessary to use the data management password of the dynamic block and the data management password of the corresponding fixed block in the dynamic block for verification.

6. A knowledge management method, comprising a mobile smart terminal and a desktop terminal with the display module held by students and school staff respectively, and an application app and an application system with the program module in the knowledge management system as described in any one of claims 1 to 5 are installed on the mobile smart terminal and the desktop terminal respectively; the construction of various heterogeneous neural networks and various attention mechanisms are implemented on the application system, as well as the update, and the setting and management of passwords are completed.

Images