Intelligent courseware generation method based on knowledge graph

Abstract

The invention relates to the technical field of courseware generation, in particular to an intelligent courseware generation method based on a knowledge graph, which comprises the following steps: determining a graph selection mode according to a dynamic connection index and a graph interaction degree to generate a target graph; determining a generation state according to the map reference value and the learning characteristic value of the target map, and determining a knowledge point selection mode according to the generation state to obtain selected knowledge points; in node analysis and selection, determining a knowledge point supplement mode according to a node evaluation value; in path analysis and selection, determining a node state according to node complexity and a node distribution coefficient, and determining a screening mode according to the node state; under a selection completion condition, determining an optimization mode according to the pre-replacement coefficient of the optimization range to obtain supplementary knowledge points; filling each selected knowledge point and each supplementary knowledge point to generate courseware; according to the invention, the accuracy of courseware generation can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of courseware generation, and in particular to an intelligent courseware generation method based on a knowledge graph. Background Art

[0002] The traditional courseware production method mainly relies on teachers to manually collect materials, organize knowledge points, and use office software for typesetting design. This process is not only time-consuming and laborious, but also easily restricted by the teacher's personal knowledge reserve and teaching experience, resulting in uneven quality of courseware content. In contrast, knowledge graph technology can connect scattered and isolated knowledge points into an organic whole to form a knowledge network with rich semantic associations. However, it still faces problems such as the integrity of knowledge graph construction and the effective selection of knowledge points. Therefore, how to effectively screen knowledge points in the knowledge graph to improve the accuracy of courseware generation is a technical problem that those skilled in the art urgently need to solve.

[0003] Chinese Patent Publication No. CN116932488A discloses a courseware generation method, device, system and storage medium based on a knowledge graph. The method includes: responding to a courseware creation request initiated by a client, and selecting a corresponding knowledge graph; based on the knowledge graph, obtaining a first mapping relationship between class hours and knowledge points, and sending the first mapping relationship to the client; receiving a second mapping relationship sent by the client, and using the knowledge graph to obtain a third mapping relationship between class hours, knowledge points and courseware resources in the second mapping relationship, and sending the third mapping relationship to the client, where the second mapping relationship is obtained after the client edits the first mapping relationship; receiving a fourth mapping relationship sent by the client to generate a courseware, where the fourth mapping relationship is obtained after the client screens the third mapping relationship. It can be seen that the above technical solution has the following problems: only generating courseware based on the mapping relationship, lacking effective screening of knowledge points, and being unable to dynamically adjust the knowledge points according to the actual situation, resulting in poor accuracy of the generated courseware. Summary of the Invention

[0004] To this end, the present invention provides an intelligent courseware generation method based on a knowledge graph to overcome the problems in the prior art that there is a lack of effective screening of knowledge points, and it is unable to dynamically adjust the knowledge points according to the actual situation, resulting in poor accuracy of the generated courseware.

[0005] To achieve the above object, the present invention provides an intelligent courseware generation method based on a knowledge graph, including: Determining a graph selection method according to the dynamic connection index and the graph interaction degree to generate a target graph, where the graph selection method is to select separately according to the feature evaluation value or to select in association according to the explicit interaction value and the implicit coupling degree; Determine the generation status based on the atlas reference value and learning feature value of the target atlas, and determine the knowledge point selection method as node analysis selection or path analysis selection according to the generation status; In node analysis selection, determine the knowledge point supplementation method as node expansion supplementation or associated enhancement supplementation according to the node evaluation value; In path analysis selection, determine the node status according to the node complexity and node distribution coefficient, and determine the screening method as paragraph analysis selection or node analysis selection according to the node status; Under the condition of selection completion, determine the optimization method according to the pre-replacement coefficient of the optimization range. The optimization method is to supplement knowledge points according to the missing coefficient or replace knowledge points according to the replacement efficiency threshold; Fill in each knowledge point to generate a courseware.

[0006] Furthermore, determine the atlas selection method according to the dynamic connection index and atlas interaction degree, including: If the dynamic connection index is greater than or equal to the preset dynamic connection index and the atlas interaction degree is greater than or equal to the preset atlas interaction degree, the atlas selection method is to select alone according to the feature evaluation value; If the dynamic connection index is less than the preset dynamic connection index or the atlas interaction degree is less than the preset atlas interaction degree, the atlas selection method is to select associated according to the explicit interaction value and implicit coupling degree.

[0007] Furthermore, if the generation status is that the atlas reference value is greater than or equal to the preset atlas reference value or the learning feature value is greater than or equal to the preset learning feature value, the knowledge point selection method is node analysis selection.

[0008] Furthermore, if the generation status is that the atlas reference value is less than the preset atlas reference value and the learning feature value is less than the preset learning feature value, the knowledge point selection method is path analysis selection.

[0009] Furthermore, if the node evaluation value is greater than or equal to the preset node evaluation value, the knowledge point supplementation method is node expansion supplementation; In node expansion supplementation, each associated node and each node within the expansion depth corresponding to each associated node are used as the selected knowledge points; The expansion depth corresponding to a single associated node has a positive correlation with the hierarchical quantization index corresponding to the associated node.

[0010] Furthermore, if the node evaluation value is less than the preset node evaluation value, the knowledge point supplementation method is associated enhancement supplementation; In the associated enhancement supplement, node combinations are determined based on node association degrees and semantic distance coefficients, characteristic paragraphs are determined according to combination influence degrees and feature coupling strengths, and the associated enhancement methods for each characteristic paragraph are determined according to path dependence coefficients to obtain a number of supplementary nodes. The associated nodes and the supplementary nodes corresponding to each characteristic paragraph are used as selected knowledge points; If the path dependence coefficient is greater than or equal to the preset path dependence coefficient, the associated enhancement method is interval node enhancement supplement; If the path dependence coefficient is less than the preset path dependence coefficient, the associated enhancement method is feature path enhancement supplement.

[0011] Furthermore, if the node state is that the node complexity is greater than or equal to the preset node complexity or the node distribution coefficient is greater than or equal to the preset node distribution coefficient, the screening method is paragraph analysis selection; In paragraph analysis selection, associated paragraphs are determined according to the associated reference value, and the node selection quantity is determined according to the paragraph intersection degree and the paragraph feature coefficient; The node selection quantity corresponding to a single associated paragraph has a positive correlation with the paragraph intersection value and the paragraph feature coefficient corresponding to this associated paragraph.

[0012] Furthermore, if the node state is that the node complexity is less than the preset node complexity and the node distribution coefficient is less than the preset node distribution coefficient, the screening method is node analysis selection; In node analysis selection, the to-be-selected nodes with analysis feature values greater than the preset analysis feature value and each associated node are used as selected knowledge points.

[0013] Furthermore, the optimization method is determined according to the pre-replacement coefficient of the optimization range, including: If the pre-replacement coefficient is less than the preset pre-replacement coefficient, the selected optimization method is to supplement knowledge points according to the missing coefficient; If the pre-replacement coefficient is greater than or equal to the preset pre-replacement coefficient, the selected optimization method is to replace knowledge points according to the replacement efficiency threshold.

[0014] Furthermore, the confirmation method of the optimization range includes: The initial optimization range is determined according to the radiation influence coefficient and the courseware deviation threshold, and the optimization range is adjusted to increase based on the comparison range deviation value; The initial optimization range after the increase adjustment is recorded as the optimization range.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows. In the technical solution of the present invention, the content adaptability between the course and each knowledge graph is effectively reflected by the dynamic connection index and the graph interaction degree. Furthermore, different graph selection methods are adaptively selected according to the dynamic connection index and the graph interaction degree, so that the selected graph selection method can accurately adapt to the course requirements, avoid missing key knowledge or logical breaks, and dynamically match the knowledge graph that best meets the course requirements, thereby improving the accuracy of courseware generation.

[0016] Furthermore, in the present invention, the generation state is determined according to the graph reference value and the learning feature value of the target graph. The knowledge density and structural integrity of the target graph are effectively reflected by the graph reference value and the learning feature value. Furthermore, different knowledge point selection methods are adaptively selected according to the generation state. Selecting by node analysis can focus on the core knowledge points and reduce the cognitive load; selecting by path analysis can construct a complete knowledge chain to strengthen logical deduction, effectively avoiding resource waste caused by forced path analysis of low-quality graphs, and at the same time preventing the problem of loose logic caused by only using node analysis for high-value graphs, significantly improving the accuracy and adaptability of courseware generation, and making the knowledge point selection method more suitable for the actual application scenario.

[0017] Furthermore, in the present invention, the actual state of the effective nodes is effectively reflected by the node evaluation value. Furthermore, different knowledge point supplementation methods are adaptively selected according to the node evaluation value, so that the selection of the knowledge point supplementation method is more in line with the actual application scenario, can expand the key knowledge points, and at the same time strengthen the association of weak nodes to ensure the logical coherence of the knowledge graph, thereby improving the content quality and teaching adaptability of courseware generation.

[0018] Furthermore, in the present invention, the node state is determined according to the node complexity and the node distribution coefficient. The complexity and distribution degree of the knowledge points are effectively reflected by the node complexity and the node distribution coefficient. Furthermore, different screening methods are adaptively selected according to the node state. When the node state is complex, selecting by paragraph analysis can ensure that the path covers the core knowledge points and avoid logical breaks. When the node state is simple, selecting by endpoint analysis can strengthen the logical connection between simple nodes, ensuring that the generated courseware covers both the core knowledge points and maintains structural coherence.

[0019] Furthermore, in the present invention, the replaceability of the knowledge points in the optimization range is effectively reflected by the pre-replacement coefficient. Furthermore, different optimization methods are adaptively selected according to the pre-replacement coefficient, so that the selection of the optimization method is more in line with the actual application scenario, selects the optimal solution according to the knowledge point state, improves the optimization efficiency, and thereby improves the accuracy of courseware generation. Brief Description of the Drawings

[0020] Figure 1Schematic diagram of the intelligent courseware generation method based on knowledge graph of the present invention; Figure 2 Flowchart of the present invention for determining the graph selection method according to the dynamic connection index and the graph interaction degree; Figure 3 Flowchart of the present invention for determining the knowledge point selection method according to the generation status; Figure 4 Flowchart of the present invention for determining the knowledge point supplement method according to the node evaluation value. Detailed implementation manners

[0021] In order to make the objectives and advantages of the present invention clearer and more understandable, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0022] The preferred implementation manners of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these implementation manners are only used to explain the technical principle of the present invention and do not limit the protection scope of the present invention.

[0023] It should be noted that in the description of the present invention, the terms indicating directions or positional relationships such as "upper", "lower", "left", "right", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings. This is only for convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0024] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0025] Please refer to Figures 1 to 4 As shown, the present invention provides an intelligent courseware generation method based on a knowledge graph, including: Determining a graph selection method according to the dynamic connection index and the graph interaction degree to generate a target graph, where the graph selection method is to select separately according to the feature evaluation value or to select in association according to the explicit interaction value and the implicit coupling degree; Determining the generation status according to the graph reference value and the learning feature value of the target graph, and determining the knowledge point selection method as node analysis selection or path analysis selection according to the generation status; In node analysis selection, the knowledge point supplement method is determined as node expansion supplement or associated enhancement supplement according to the node evaluation value; In path analysis selection, the node state is determined according to the node complexity and the node distribution coefficient, and the screening method is determined as paragraph analysis selection or node analysis selection according to the node state; Under the condition of selection completion, the optimization method is determined according to the pre-replacement coefficient of the optimization scope. The optimization method is to supplement knowledge points according to the missing coefficient or replace knowledge points according to the replacement efficiency threshold; Fill in each knowledge point to generate a courseware.

[0026] The application scenario of the present invention is to generate a courseware using a knowledge graph. The present invention includes several knowledge graphs to be selected. Each knowledge graph to be selected contains several entity keywords, and each entity keyword represents a node. A single node is connected to several nodes through edges. The edges are used to describe the semantic logic between two nodes. The semantic logic includes but is not limited to belonging to, having, containing, and possessing. This is easy for those skilled in the art to understand and will not be elaborated here; When generating a courseware in the present invention, each courseware corresponds to courseware information, which includes but is not limited to the courseware theme, applicable object, and teaching objective. The courseware theme includes but is not limited to "Python Function Basics", "Quadratic Function Image and Properties", and "Derivation and Application of Arithmetic Progression Summation Formula". The applicable object includes grade and major. The teaching objective is a text elaborating the knowledge, skills, and attitude objectives that students should achieve by learning this courseware. This is easy for those skilled in the art to understand and will not be elaborated here; In the present invention, several historical records are correspondingly set. Any historical record records at least one characteristic interaction value, explicit interaction value, implicit coupling degree, graph reference value, learning characteristic value, etc. in the historical process of generating a courseware using a knowledge graph, and each historical record corresponds to a qualified mark. The qualified mark records whether the process of generating a courseware using a knowledge graph meets the user's requirements. The qualified mark can be manually recorded. It can be understood that the user can determine whether the process of generating a courseware using a knowledge graph meets the requirements according to self-set indicators. The self-set indicators can be but are not limited to the error index, which will not be elaborated here. Among them, the error index is the number of coursewares that do not match the course content generated; The selection completion condition is that the selected knowledge points selected by the knowledge point selection method are completed; Fill in each knowledge point to generate a courseware, including: exporting the nodes corresponding to the selected courseware knowledge points and the edges corresponding to each node from the knowledge graph, exporting them in CSV or JSON format, using a Python script to batch generate a courseware content template, and importing the generated content into the PPT template to generate a courseware. This is easy for those skilled in the art to understand and will not be elaborated here.

[0027] Specifically, the method for determining the graph selection method according to the dynamic connection index and the graph interaction degree includes: If the dynamic connection index is greater than or equal to the preset dynamic connection index and the graph interaction degree is greater than or equal to the preset graph interaction degree, the graph selection method is to select individually according to the feature evaluation value; If the dynamic connection index is less than the preset dynamic connection index or the graph interaction degree is less than the preset graph interaction degree, the graph selection method is to select in association according to the explicit interaction value and the implicit coupling degree.

[0028] Among them, the courseware information corresponding to the currently generated courseware is recorded as the target courseware information, and each keyword in the target courseware information is recorded as the courseware keyword; The dynamic connection index is the maximum value among the sub-connection indexes corresponding to each knowledge graph to be selected. The method for confirming the sub-connection index is as follows: for a single knowledge graph to be selected, this knowledge graph to be selected is recorded as the target knowledge graph to be selected, and the average value of the similarity reference values corresponding to each courseware keyword is recorded as the sub-connection index corresponding to the target knowledge graph to be selected; the similarity reference value corresponding to a single courseware keyword is the maximum value among the similarity coefficients corresponding to this courseware keyword and each entity keyword in the target knowledge graph to be selected; The method for confirming the similarity coefficient is as follows: for any two keywords, the keywords are converted into word vectors, and the cosine value of the included angle between the two word vectors is recorded as the similarity coefficient corresponding to the two keywords. The keywords are converted into word vectors through a vector model, and the vector model includes but is not limited to Word2Vec, GloVe, and BERT. This is easily understood by those skilled in the art and will not be elaborated here; The graph interaction degree is the average value of the sub-graph interaction degrees corresponding to each knowledge graph to be selected. The sub-graph interaction degree corresponding to a single knowledge graph to be selected is the average value of the graph association degrees between this knowledge graph to be selected and other knowledge graphs to be selected. The graph association degree corresponding to any two knowledge graphs to be selected = the number of entity keywords that appear simultaneously in the two knowledge graphs to be selected / the larger value among the entity keyword quantities corresponding to the two knowledge graphs to be selected; The values of the preset dynamic connection index and the preset graph interaction degree can be determined by the user according to the actual application scenario. The smaller the values of the preset dynamic connection index and the preset graph interaction degree, the greater the user's need to select individually according to the feature evaluation value. A set of values for the preset dynamic connection index and the preset graph interaction degree is provided. The preset dynamic connection index is 0.7, and the preset graph interaction degree is 70%; Selecting individually according to the feature evaluation value includes: recording the knowledge graph to be selected with a feature interaction value greater than the preset feature interaction value as the selected graph, and fusing each selected graph to obtain the target graph; Tools that can be used for fusing each selected knowledge graph include, but are not limited to, Falcon-AO and LiMES, which are understandable to those skilled in the art and will not be elaborated herein; Select according to the explicit interaction value and the implicit coupling degree, including: Denote the knowledge graph to be selected with the largest feature interaction value as the target selected graph, denote the other knowledge graphs to be selected except the target selected graph as the reference knowledge graphs to be selected, denote the reference knowledge graphs to be selected with an explicit interaction value greater than the preset explicit interaction value or an implicit coupling degree greater than the preset implicit coupling degree with the target selected graph as the reference selected graphs, and fuse each reference selected graph and the target selected graph to obtain the target graph; Denote the entity keyword with a similarity coefficient greater than 0.8 to any courseware keyword in the knowledge graph to be selected as the associated keyword. It can be understood that each associated keyword corresponds to a courseware keyword, and the courseware keyword corresponding to a single associated keyword is the courseware keyword with a similarity coefficient greater than 0.8 to this associated keyword; Explicit interaction value = the number of different courseware keywords among all the courseware keywords corresponding to each associated keyword included in the two knowledge graphs to be selected / the total number of courseware keywords in the target courseware information; The method for confirming the implicit coupling degree is as follows: For any two knowledge graphs to be selected, denoted as the first graph and the second graph respectively, denote the same associated keywords in the two knowledge graphs to be selected as the same keywords, denote the other entity keywords except the same keywords as the interval keywords, respectively detect the shortest paths that can contain the nodes corresponding to each same keyword in the two knowledge graphs to be selected, denoted as the first path and the second path respectively. The implicit coupling degree corresponding to the two knowledge graphs to be selected = the total number of the same interval keywords in the first path and the second path / the larger value of the path lengths corresponding to the first path and the second path. The path length corresponding to a single path is the total number of nodes included in this path; Feature interaction value = sub-graph interaction degree × sub-connection index; The values of the preset feature interaction value, the preset explicit interaction value, and the preset implicit coupling degree can be determined by the user according to the actual application scenario. The smaller the user's requirement for improving the accuracy of the generated courseware, the larger the values of the preset feature interaction value, the preset explicit interaction value, and the preset implicit coupling degree. Provide a method for determining the values of the preset feature interaction value, the preset explicit interaction value, and the preset implicit coupling degree. Detect the historical records selected by the user alone according to the feature evaluation value, and record the average value of the feature interaction values corresponding to the historical records that can meet the user's requirements as the preset feature interaction value. Detect the historical records selected by the user in association according to the explicit interaction value and the implicit coupling degree, and record the average value of the explicit interaction values corresponding to the historical records that can meet the user's requirements as the preset explicit interaction value, and record the average value of the implicit coupling degrees corresponding to the historical records that can meet the user's requirements as the preset implicit coupling degree.

[0029] Specifically, if the generation status is that the graph reference value is greater than or equal to the preset graph reference value or the learning feature value is greater than or equal to the preset learning feature value, the knowledge point selection method is node analysis selection.

[0030] Among them, the generation status includes a first generation status and a second generation status. The first generation status is that the graph reference value is greater than or equal to the preset graph reference value or the learning feature value is greater than or equal to the preset learning feature value. The second generation status is that the graph reference value is less than the preset graph reference value and the learning feature value is less than the preset learning feature value. The graph reference value is the average value of the node reference values corresponding to each node in the target graph, and the node reference value corresponding to a single node is the number of nodes directly connected to that node. Mark the nodes corresponding to each associated keyword in the target graph as associated nodes. Learning feature value = the sum of the node reference values corresponding to each associated node / (the number of associated nodes + 1). The values of the preset graph reference value and the preset learning feature value can be determined by the user according to the actual application scenario. The smaller the values of the preset graph reference value and the preset learning feature value, the greater the user's need for node analysis selection. Provide a method for determining the values of the preset graph reference value and the preset learning feature value. Detect the historical records of the user's node analysis selection, and record the average value of the graph reference values corresponding to the historical records that can meet the user's requirements as the preset graph reference value, and record the average value of the learning feature values corresponding to the historical records that can meet the user's requirements as the preset learning feature value.

[0031] Specifically, if the generation status is that the graph reference value is less than the preset graph reference value and the learning feature value is less than the preset learning feature value, the knowledge point selection method is path analysis selection.

[0032] Specifically, if the node evaluation value is greater than or equal to the preset node evaluation value, the knowledge point supplement method is node expansion supplement; During node expansion and supplementation, each associated node and each node within the expansion depth corresponding to each associated node are used as selected knowledge points; The expansion depth corresponding to a single associated node has a positive correlation with the hierarchical quantization index corresponding to that associated node.

[0033] Among them, the node evaluation value is the average of the sub-evaluation values corresponding to each associated node, and the sub-evaluation value corresponding to a single associated node is the number of associated nodes directly connected to that associated node; For the value of the preset node evaluation value, the user can determine it according to the actual application scenario. The smaller the value of the preset node evaluation value, the greater the user's need for node expansion and supplementation. Provide a value for the preset node evaluation value, detect the historical records of the user's node expansion and supplementation, and record the average value of the node evaluation values corresponding to the historical records that can meet the user's needs as the preset node evaluation value; The confirmation method of the hierarchical quantization index is as follows: for a single associated node, mark this associated node as the first target associated node, and mark the other nodes in the target graph except the first target associated node as analysis nodes. The hierarchical quantization index corresponding to the first target associated node = the number of analysis nodes directly connected to the first target associated node / the average value of the semantic distance coefficients between each analysis node and the first target associated node; The expansion depth corresponding to the first target associated node is the range of nodes that can be reached by moving along the edge n times starting from the first target associated node. n has a positive correlation with the hierarchical quantization index corresponding to the first target associated node.

[0034] Specifically, if the node evaluation value is less than the preset node evaluation value, the knowledge point supplementation method is association enhancement supplementation; In association enhancement supplementation, determine node combinations based on node association degrees and semantic distance coefficients, determine feature paragraphs according to combination influence degrees and feature coupling strengths, and determine the association enhancement methods for each feature paragraph according to path dependence coefficients to obtain several supplementary nodes. Use each associated node and the supplementary nodes corresponding to each feature paragraph as selected knowledge points; If the path dependence coefficient is greater than or equal to the preset path dependence coefficient, the association enhancement method is interval node enhancement supplementation; If the path dependence coefficient is less than the preset path dependence coefficient, the association enhancement method is feature path enhancement supplementation.

[0035] Among them, determining the node combination based on the node correlation degree and the semantic distance coefficient includes: performing combination analysis on each associated node. When performing combination analysis on a single associated node, the associated node is denoted as the target node, the other associated nodes that are not included in the node combination outside the target node are denoted as reference nodes, and the combination of the reference nodes whose node correlation degree with the target node is greater than the preset node correlation degree and the semantic distance coefficient is less than the preset semantic distance coefficient and the target node is denoted as a node combination, and continue to perform combination analysis on the associated nodes that are not included in the node combination until all associated nodes are included in the node combination, then stop the combination analysis; The confirmation method of the node correlation degree is that for any two nodes, the number of paths with the two nodes as the two endpoints is denoted as the node correlation degree corresponding to the two nodes; The semantic distance coefficient corresponding to any two nodes is the total number of nodes included in the shortest path with the two nodes as the two endpoints; The values of the preset node correlation degree and the preset semantic distance coefficient can be determined by the user according to the actual application scenario. The greater the user's demand for improving the accuracy of courseware generation, the greater the value of the preset node correlation degree and the smaller the value of the preset semantic distance coefficient. Provide a set of values for the preset node correlation degree and the preset semantic distance coefficient, detect the historical records of the user's determination of the node combination based on the node correlation degree and the semantic distance coefficient, and denote the average value of the reference node correlation degrees corresponding to each node combination in the historical records that can meet the user's needs as the preset node correlation degree, and denote the average value of the reference semantic distance coefficients corresponding to each node combination in the historical records that can meet the user's needs as the preset semantic distance coefficient. The reference node correlation degree corresponding to a single node combination is the node correlation degree corresponding to any two nodes in the node combination, and the reference semantic distance coefficient corresponding to a single node combination is the semantic distance coefficient corresponding to any two nodes in the node combination; Determining the feature paragraph according to the combination influence degree and the feature coupling strength includes: denoting the associated node with the largest combination influence degree in each node combination as the feature node, and performing matching analysis on each feature node. When performing matching analysis on a single feature node, the feature node is denoted as the target feature node, the other feature nodes that are not denoted as matching nodes outside the target feature node are denoted as reference feature nodes, and the reference feature node with the largest feature coupling strength with the target feature node is used as the matching node of the target feature node, and continue to perform matching analysis on the feature nodes without matching nodes until the preset condition is reached, then stop the matching analysis. It should be noted that when the number of node combinations is even, the preset condition is that each feature node has a matching node; when the number of node combinations is odd, the preset condition is that only one feature node remains without a matching node; denote the feature nodes with matching nodes as feature reference nodes; Each feature reference node corresponds to a feature paragraph. The feature paragraph corresponding to a single feature reference node is each node included in the shortest path with this feature reference node and the matching node corresponding to this feature reference node as the two endpoints; The confirmation method of the combined influence degree is as follows: for a single associated node in a single node combination, record this associated node as the first target node, and record the other nodes in this node combination except the first target node as the first reference nodes. The combined influence degree corresponding to the first target node = the average value of the semantic distance coefficients corresponding to the first target node and each first reference node × the average value of the node correlation degrees corresponding to the first target node and each first reference node; The confirmation method of the feature coupling strength is as follows: for any two feature nodes, the feature coupling strength corresponding to the two feature nodes = the similarity coefficient corresponding to the entity keywords corresponding to the two feature nodes × the total number of nodes included in the shortest path with the two feature nodes as the two endpoints; The confirmation method of the path dependence coefficient is as follows: for a single feature paragraph, record the two feature nodes corresponding to this feature paragraph as the analysis feature nodes, and record the path with the two analysis feature nodes as the two endpoints as the analysis path. The path dependence coefficient = the average value of the feature reference values corresponding to each analysis path / the standard deviation of the feature reference values corresponding to each analysis path; Regarding the value of the preset path dependence coefficient, the user can determine it according to the actual application scenario. The smaller the value of the preset path dependence coefficient, the greater the user's need for feature path enhancement and supplementation. Provide a value of the preset path dependence coefficient, detect the historical records of the user's feature path enhancement and supplementation, and record the average value of the path dependence coefficients corresponding to the historical records that can meet the user's needs as the preset path dependence coefficient; Interval node enhancement and supplementation include: regarding the analysis paths with feature reference values greater than the preset feature reference value as the paths to be selected, selecting nodes in the paths to be selected at node selection intervals, and taking the selected nodes in each path to be selected as supplementary nodes; The node selection interval corresponding to a single path to be selected has a negative correlation with the feature reference value corresponding to this path to be selected; Feature path enhancement and supplementation include: regarding the analysis path with the largest feature reference value as the feature path, and taking each node in the feature path as a supplementary node; The node selection interval is the number of nodes in the middle interval between two adjacent supplementary nodes selected in a single path to be selected; The method for confirming the characteristic reference value is as follows. For a single analysis path, denote this analysis path as the target analysis path, denote the nodes in the target analysis path as target path nodes, and denote the nodes in other analysis paths outside the target analysis path as reference path nodes. The characteristic reference value corresponding to the target analysis path = the number of reference path nodes directly connected to the target path nodes / the number of target path nodes; For the value of the preset characteristic reference value, the user can determine it according to the actual application scenario. The greater the user's demand for improving the association relationship between knowledge points, the smaller the value of the preset characteristic reference value. Provide a value of the preset characteristic reference value, detect the historical records of the user's interval node enhancement and supplementation, and denote the average value of the characteristic reference values corresponding to the historical records that can meet the user's needs as the preset characteristic reference value.

[0036] Specifically, if the node state is that the node complexity is greater than or equal to the preset node complexity or the node distribution coefficient is greater than or equal to the preset node distribution coefficient, the screening method is paragraph analysis selection; In paragraph analysis selection, determine the associated paragraphs according to the association reference value, and determine the node selection quantity according to the paragraph intersection degree and the paragraph characteristic coefficient; The node selection quantity corresponding to a single associated paragraph has a positive correlation with both the paragraph intersection value and the paragraph characteristic coefficient corresponding to this associated paragraph.

[0037] Among them, the node state includes the first node state and the second node state. The first node state is that the node complexity is greater than or equal to the preset node complexity or the node distribution coefficient is greater than or equal to the preset node distribution coefficient, and the second node state is that the node complexity is less than the preset node complexity and the node distribution coefficient is less than the preset node distribution coefficient; The node complexity is the number of associated nodes in the target graph. The node distribution coefficient is the average value of the node influence degrees corresponding to each associated node. For a single associated node, denote this associated node as the target associated node, denote the other associated nodes outside the target associated node as reference associated nodes, and denote the minimum value of the semantic distance coefficients corresponding to the target associated node and each reference associated node as the node influence degree corresponding to the target associated node; For the values of the preset node complexity and the preset node distribution coefficient, the user can determine them according to the actual application scenario. The smaller the values of the preset node complexity and the preset node distribution coefficient, the greater the user's demand for paragraph analysis selection. Provide the values of the preset node complexity and the preset node distribution coefficient, detect the historical records of the user's paragraph analysis selection, denote the average value of the node complexities corresponding to the historical records that can meet the user's needs as the preset node complexity, and denote the average value of the node distribution coefficients corresponding to the historical records that can meet the user's needs as the preset node distribution coefficient; Determine associated paragraphs based on associated reference values, including: Denote other nodes except the associated nodes that appear in the shortest path capable of containing each associated node as candidate nodes to be selected. Conduct association analysis for each candidate node to be selected. When conducting association analysis for a single candidate node to be selected, denote this candidate node to be selected as the target candidate node to be selected, denote candidate nodes to be selected that are outside the target candidate node to be selected and have not been recorded in the associated paragraph as reference candidate nodes to be selected. Denote the set of reference candidate nodes to be selected whose associated reference values with the target candidate node to be selected are greater than the preset associated reference value and the target candidate node to be selected as an associated paragraph, and continue to conduct association analysis for candidate nodes to be selected that have not been recorded in the associated paragraph until all candidate nodes to be selected are recorded in the associated paragraph, then stop the association analysis; The paragraph intersection degree corresponding to a single associated paragraph is the number of other candidate nodes to be selected directly connected to each candidate node to be selected in this associated paragraph; The paragraph feature coefficient corresponding to a single associated paragraph = the number of candidate nodes to be selected included in this associated paragraph × the number of nodes included in the shortest path capable of containing each candidate node to be selected in this associated paragraph; The node selection quantity corresponding to a single associated paragraph is the number of candidate nodes to be selected selected in this associated paragraph. The node selection quantity corresponding to a single associated paragraph has a positive correlation with the paragraph evaluation value. The paragraph evaluation value = the paragraph intersection value corresponding to this associated paragraph + the paragraph feature coefficient corresponding to this associated paragraph; Take each candidate node to be selected selected in each associated paragraph and each associated node as selected knowledge points. Each candidate node to be selected selected in a single associated paragraph is a node randomly selected according to the node selection quantity corresponding to this associated paragraph; The confirmation method of the associated reference value is as follows: For any two nodes, denote them as the first node and the second node respectively. Denote the nodes directly connected to the first node as the first reference nodes, and denote the nodes directly connected to the second node as the second reference nodes. The associated reference value = the number of identical nodes among the first reference nodes and the second reference nodes / the number of nodes included in the shortest path with the first node and the second node as the two end points; For the value of the preset associated reference value, the user can determine it according to the actual application scenario. The greater the user's demand for improving the accuracy of courseware generation, the greater the value of the preset associated reference value. Provide a value of the preset associated reference value. Detect the historical records of the user determining the associated paragraphs according to the associated reference value, and denote the average value of the reference associated reference values corresponding to each associated paragraph in the historical records that can meet the user's needs as the preset associated reference value. The reference associated reference value corresponding to a single associated paragraph is the associated reference value corresponding to any two nodes in this associated paragraph.

[0038] Specifically, if the node state is that the node complexity is less than the preset node complexity and the node distribution coefficient is less than the preset node distribution coefficient, the screening method is node analysis and selection; In node analysis selection, the to-be-selected nodes with analysis eigenvalues greater than the preset analysis eigenvalue and their associated nodes are used as the selected knowledge points.

[0039] Among them, the confirmation method of the analysis eigenvalue is as follows: for a single to-be-selected node, this to-be-selected node is denoted as the target to-be-selected node, and the analysis eigenvalue corresponding to the target to-be-selected node = the number of other to-be-selected nodes directly connected to the target to-be-selected node / the average value of the semantic distance coefficients corresponding to the target to-be-selected node and its associated nodes; The value of the preset analysis eigenvalue can be determined by the user according to the actual application scenario. The greater the user's demand for improving the accuracy of the courseware, the greater the value of the preset analysis eigenvalue. A value of the preset analysis eigenvalue is provided. The historical records of the user's node analysis selection are detected, and the average value of the analysis eigenvalues of the nodes corresponding to the supplementary knowledge points in the historical records that can meet the user's needs is denoted as the preset analysis eigenvalue.

[0040] Specifically, the optimization method is determined according to the pre-replacement coefficient of the optimization range, including: If the pre-replacement coefficient is less than the preset pre-replacement coefficient, the selected optimization method is to supplement knowledge points according to the missing coefficient; If the pre-replacement coefficient is greater than or equal to the preset pre-replacement coefficient, the selected optimization method is to replace knowledge points according to the replacement efficiency threshold.

[0041] Among them, the nodes corresponding to each selected knowledge point are denoted as selected nodes. The pre-replacement coefficient is the average value of the sub-replacement coefficients corresponding to each selected node. The confirmation method of the sub-replacement coefficient is as follows: for a single selected node, this selected node is denoted as the target selected node, and the other selected nodes except the target selected node are denoted as reference selected nodes. The historical records that can meet the user's needs and contain the target selected node are detected and denoted as reference historical records. The sub-replacement coefficient corresponding to the target selected node is the average value of the number of reference selected nodes appearing in each reference historical record. It should be noted that if there is no reference historical record for the target selected node, the sub-replacement coefficient corresponding to the target selected node is 0; The value of the preset pre-replacement coefficient can be determined by the user according to the actual application scenario. The greater the value of the preset pre-replacement coefficient, the greater the user's demand for supplementing knowledge points according to the missing coefficient. A value of the preset pre-replacement coefficient is provided. The historical records of the user's supplementing knowledge points according to the missing coefficient are detected, and the average value of the pre-replacement coefficients corresponding to the historical records that can meet the user's needs is denoted as the preset pre-replacement coefficient; Supplement knowledge points according to the missing coefficient, including: supplement knowledge points for the selection nodes with missing coefficients greater than the preset missing coefficient. When supplementing knowledge points for a single selection node, the expansion depth corresponding to the selection node is positively correlated with the missing coefficient of the selection node. The expansion depth corresponding to the selection node is the range of nodes that can be reached by moving m times along the edge starting from the selection node, and m is positively correlated with the missing coefficient of the selection node; Take each node within the expansion depth corresponding to the selection nodes with missing coefficients greater than the preset missing coefficient as the supplemented selected knowledge points, and take each supplemented selected knowledge point and each selected knowledge point as the knowledge points of the courseware; Replace knowledge points according to the replacement efficiency threshold, including: replace knowledge points for the selection nodes with sub-replacement coefficients greater than the preset sub-replacement coefficient. When replacing knowledge points for a single selection node, delete the selection node, and take the unselected node with the largest replacement efficiency threshold corresponding to the selection node as the supplemented selected knowledge point, and take each supplemented selected knowledge point and each un-replaced selected knowledge point as the knowledge points of the courseware; The confirmation method of the missing coefficient is that for a single selection node, mark the selection node as the target selection node, and the missing coefficient corresponding to the target selection node = the number of selection nodes directly connected to the target selection node / the total number of nodes connected to the target selection node; The values of the preset missing coefficient and the preset sub-replacement coefficient can be determined by the user according to the actual application scenario. The greater the user's demand for improving the accuracy of the generated courseware, the smaller the values of the preset missing coefficient and the preset sub-replacement coefficient. Provide a value for the preset missing coefficient and the preset sub-replacement coefficient. The preset missing coefficient is 50%. Detect the historical records of the user's supplementing knowledge points according to the missing coefficient, and record the average value of the sub-replacement coefficients of the selection nodes corresponding to the historical records that can meet the user's needs during the supplementing of knowledge points as the preset sub-replacement coefficient; The confirmation method of the replacement efficiency threshold is to mark the other nodes outside the selection nodes in the target graph as unselected nodes, and mark the unselected nodes directly connected to the selection nodes as adjacent unselected nodes. For a single selection node, mark the selection node as the target selection node. For a single adjacent unselected node directly connected to the target selection node, mark the adjacent unselected node as the target adjacent unselected node. The replacement efficiency threshold corresponding to the target adjacent unselected node and the target selection node = the similarity coefficient of the entity keyword corresponding to the target adjacent unselected node and the entity keyword corresponding to the target selection node / the average value of the semantic distance coefficients corresponding to the target adjacent unselected node and each selection node.

[0042] Specifically, the confirmation method of the optimization range includes: Determine the initial optimization range according to the radiation influence coefficient and the courseware deviation threshold, and perform an increase adjustment on the optimization range based on the comparison range deviation value; Record the initial optimization range after the increase adjustment as the optimization range.

[0043] Among them, the confirmation method of the radiation influence coefficient is as follows: Denote the nodes corresponding to each selected knowledge point as selected nodes, and denote the other nodes except the associated nodes among the selected nodes as non-associated nodes. The radiation influence coefficient = the average value of the sub-radiation influence coefficients corresponding to each non-associated node / the standard deviation of the sub-radiation influence coefficients corresponding to each non-associated node. The sub-radiation influence coefficient corresponding to a single non-associated node is the average value of the semantic distance coefficients between this non-associated node and the associated nodes; The confirmation method of the courseware deviation threshold is as follows: For a single non-associated node, denote the maximum value among the similarity coefficients between the entity keyword corresponding to this non-associated node and the entity keywords corresponding to the associated nodes as the courseware deviation threshold corresponding to this non-associated node; When determining the initial optimization range according to the radiation influence coefficient and the courseware deviation threshold, the number of non-associated nodes in the initial optimization range has a positive correlation with the radiation influence coefficient. The initial optimization range is the non-associated nodes selected in the order from large to small of the courseware deviation threshold; The comparison range deviation value = |the comparison range reference value corresponding to the currently generated courseware - the average value of the comparison range reference values corresponding to each historical record that can meet the user's needs|. Denote the non-associated nodes outside the initial optimization range as analyzed non-associated nodes, and denote the associated nodes within the initial optimization range as analyzed associated nodes. The comparison range reference value = the average value of the connection coefficients corresponding to each analyzed non-associated node × the number of analyzed non-associated nodes. The connection coefficient corresponding to a single analyzed non-associated node is the number of analyzed associated nodes directly connected to this analyzed non-associated node; The increase value of the optimization range has a positive correlation with the comparison range deviation value. The increase value of the optimization range = the number of non-associated nodes included in the optimization range - the number of non-associated nodes included in the initial optimization range; Select the analyzed non-associated nodes in the order from large to small of the courseware deviation threshold until the increase value of the optimization range is reached. Take each selected analyzed non-associated node and each analyzed associated node in the initial optimization range as the optimization range.

[0044] So far, the technical solution of the present invention has been described in combination with the preferred embodiments shown in the drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0045] The foregoing are only preferred embodiments of the present invention and are not intended to limit the present invention; for those skilled in the art, various modifications and variations can be made to the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for intelligent generation of courseware based on a knowledge graph, characterized in that, Including: Determine the atlas selection method according to the dynamic connection index and the atlas interaction degree to generate a target atlas, where the atlas selection method is to select independently according to the feature evaluation value or to select in association according to the explicit interaction value and the implicit coupling degree; Determine the generation state according to the atlas reference value and the learning feature value of the target atlas, and determine the knowledge point selection method as node analysis selection or path analysis selection according to the generation state; In node analysis selection, determine the knowledge point supplement method as node expansion supplement or association enhancement supplement according to the node evaluation value; In path analysis selection, determine the node state according to the node complexity and the node distribution coefficient, and determine the screening method as paragraph analysis selection or node analysis selection according to the node state; Under the condition of selection completion, determine the optimization method according to the pre-replacement coefficient of the optimization range, and the optimization method is to supplement knowledge points according to the missing coefficient or replace knowledge points according to the replacement efficiency threshold; Fill in each knowledge point to generate a courseware.

2. The intelligent courseware generation method based on a knowledge graph according to claim 1, wherein Determine the atlas selection method according to the dynamic connection index and the atlas interaction degree, including: If the dynamic connection index is greater than or equal to the preset dynamic connection index and the atlas interaction degree is greater than or equal to the preset atlas interaction degree, the atlas selection method is to select independently according to the feature evaluation value; If the dynamic connection index is less than the preset dynamic connection index or the atlas interaction degree is less than the preset atlas interaction degree, the atlas selection method is to select in association according to the explicit interaction value and the implicit coupling degree.

3. The intelligent courseware generation method based on a knowledge graph according to claim 2, wherein, If the generation state is that the atlas reference value is greater than or equal to the preset atlas reference value or the learning feature value is greater than or equal to the preset learning feature value, the knowledge point selection method is node analysis selection.

4. The method for intelligent generation of courseware based on a knowledge graph according to claim 3, wherein If the generation state is that the atlas reference value is less than the preset atlas reference value and the learning feature value is less than the preset learning feature value, the knowledge point selection method is path analysis selection.

5. The intelligent courseware generation method based on a knowledge graph according to claim 3, characterized in that If the node evaluation value is greater than or equal to the preset node evaluation value, the knowledge point supplement method is node expansion supplement; In node expansion supplement, take each associated node and each node within the expansion depth corresponding to each associated node as the selected knowledge points; The expansion depth corresponding to a single associated node has a positive correlation with the hierarchical quantization index corresponding to the associated node.

6. The method for intelligent generation of courseware based on a knowledge graph according to claim 3, wherein If the node evaluation value is less than the preset node evaluation value, the knowledge point supplement method is association enhancement supplement; In association enhancement supplement, determine the node combination based on the node association degree and the semantic distance coefficient, determine the feature paragraph according to the combination influence degree and the feature coupling strength, and determine the association enhancement method for each feature paragraph according to the path dependence coefficient to obtain several supplementary nodes, and take each associated node and the supplementary nodes corresponding to each feature paragraph as the selected knowledge points; If the path dependence coefficient is greater than or equal to the preset path dependence coefficient, the association enhancement method is interval node enhancement supplement; If the path dependence coefficient is less than the preset path dependence coefficient, the association enhancement method is feature path enhancement supplement.

7. The intelligent courseware generation method based on a knowledge graph according to claim 4, wherein If the node state is that the node complexity is greater than or equal to the preset node complexity or the node distribution coefficient is greater than or equal to the preset node distribution coefficient, the screening method is paragraph analysis selection; In paragraph analysis selection, determine the associated paragraph according to the association reference value, and determine the node selection quantity according to the paragraph intersection degree and the paragraph feature coefficient; The amount of nodes selected for a single associated paragraph has a positive correlation with both the paragraph intersection value and the paragraph feature coefficient corresponding to the associated paragraph.

8. The intelligent courseware generation method based on a knowledge graph according to claim 4, wherein If the node state is that the node complexity is less than the preset node complexity and the node distribution coefficient is less than the preset node distribution coefficient, the screening method is node analysis selection; In node analysis selection, the to-be-selected nodes with analysis eigenvalue greater than the preset analysis eigenvalue and their associated nodes are used as the selected knowledge points.

9. The method for intelligently generating a courseware based on a knowledge graph according to claim 8, wherein, Determine the optimization method according to the pre-replacement coefficient of the optimization range, including: If the pre-replacement coefficient is less than the preset pre-replacement coefficient, the selected optimization method is to supplement knowledge points according to the missing coefficient; If the pre-replacement coefficient is greater than or equal to the preset pre-replacement coefficient, the selected optimization method is to replace knowledge points according to the replacement efficiency threshold.

10. The intelligent teaching material generation method based on a knowledge graph according to claim 9, wherein The confirmation method of the optimization range includes: Determine the initial optimization range according to the radiation influence coefficient and the courseware deviation threshold, and perform an increase adjustment on the optimization range based on the comparison range deviation value; Record the initial optimization range after the increase adjustment as the optimization range.

Images