Knowledge graph-based education resource big data visual analysis method and system

Abstract

This application relates to the field of educational resource data visualization technology, specifically to a method and system for big data visualization and analysis of educational resources based on knowledge graphs. The method includes: acquiring educational resources and extracting knowledge to obtain triples; obtaining word vectors for each entity word in each triple; for chapters within the same subject, obtaining the correlation coefficient between any two entity words in each paragraph of each chapter, and constructing a weighted undirected graph for each chapter; obtaining the strength coefficient of the relationship connection between entity words corresponding to any two nodes in each weighted undirected graph; performing relationship connection processing on entity words within each chapter during the knowledge graph construction process; obtaining the consistency coefficient and alignment confidence between any two entity words in different chapters, and performing entity alignment processing on entity words in different chapters during the knowledge graph construction process. This application aims to improve the completeness and accuracy of knowledge graphs.

Description

Technical Field

[0001] This application relates to the field of educational resource data visualization technology, specifically to a method and system for big data visualization analysis of educational resources based on knowledge graphs. Background Technology

[0002] Against the backdrop of the rapid development of educational informatization in the new era, educational informatization has become a public consensus. Educational resource knowledge graphs can integrate massive and fragmented educational resources into a structured knowledge network, thereby clearly presenting the context and connections of subject knowledge. This structured representation not only provides teachers and students with efficient knowledge navigation and accurate resource retrieval capabilities, but also provides data support for teaching decisions and evaluations.

[0003] However, in subject-based education, knowledge points are often scattered and lack overall coherence and systematicity. Educational resources are broad in content, large in volume, logically rigorous, dense in terminology, and rich in implicit knowledge. These characteristics significantly increase the difficulty of connecting relationships and aligning entities when constructing knowledge graphs, leading to difficulties in knowledge integration and breaks in knowledge connections, thus affecting the completeness and accuracy of knowledge graph construction. Summary of the Invention

[0004] In light of the above, it is necessary to provide a method and system for visualizing and analyzing educational resource big data based on knowledge graphs. Compared with traditional methods for visualizing and analyzing educational resource big data based on knowledge graphs, this method improves the completeness and accuracy of knowledge graphs.

[0005] In a first aspect, embodiments of this application provide a method for visual analysis of educational resource big data based on knowledge graphs, the method comprising the following steps:

[0006] Acquire educational resources and extract knowledge to obtain each triplet, and obtain the word vector of each entity word in each triplet;

[0007] For chapters within the same subject, the association coefficient between any two entity words is obtained by considering the similarity of their word vectors within each paragraph and the occurrence of those two entity words across chapters. This coefficient is used to construct a weighted undirected graph for each chapter, where entity words are nodes and the association coefficient is the weight of each edge. The strength coefficient of the relationship between the entity words corresponding to any two nodes is obtained by considering the weights of all weighted edges on the shortest path between any two nodes in each weighted undirected graph and the importance of all nodes on that shortest path. Finally, the distribution of the strength coefficient between two entity words in all triples across all chapters is used to perform relationship connection processing on the entity words within each chapter during the knowledge graph construction process.

[0008] By measuring the similarity between characters of any two entity words in different chapters and the semantic similarity of various attributes between the two entity words, a consistency coefficient between the two entity words is obtained. In addition, by combining the overlap of the entity words connecting the two entity words in their respective chapters, an alignment confidence score between the two entity words is obtained, so as to perform entity alignment processing on entity words in different chapters during the knowledge graph construction process.

[0009] In one embodiment, the process of obtaining the correlation coefficient is as follows:

[0010] The similarity between the word vectors of any two entity words is denoted as vector similarity.

[0011] The percentage of paragraphs in which each entity word appears in each chapter is recorded as the first percentage among all paragraphs.

[0012] The percentage of paragraphs in which any two entity words appear simultaneously in each chapter is recorded as the second percentage;

[0013] The correlation coefficient is obtained by combining the first proportion, the second proportion, and the vector similarity of any two entity words.

[0014] In one embodiment, the correlation coefficient is calculated as follows:

[0015] Calculate the product of the first proportions of any two entity words, and calculate the ratio of the second proportion to the product;

[0016] The correlation coefficients are positively correlated with the vector similarity and the ratio, respectively.

[0017] In one embodiment, the process of obtaining the strength coefficient is as follows:

[0018] Calculate the average weight of all weighted edges on the shortest path between any two nodes;

[0019] Obtain the importance parameters of each node in each weighted undirected graph, and calculate the average value of the importance parameters of all nodes on the shortest path;

[0020] The intensity coefficient is obtained by fusing the mean and the average.

[0021] In one embodiment, the process of performing relational connection processing on entity words within each chapter during the knowledge graph construction includes:

[0022] Calculate the arithmetic mean of the intensity coefficients between two entity words in all triples of all chapters, and perform relational connection processing on two entity words whose intensity coefficients are greater than or equal to the arithmetic mean.

[0023] In one embodiment, the process of obtaining the consistency coefficient is as follows:

[0024] The similarity between the characters of any two entity words is denoted as character similarity.

[0025] The similarity of the semantic vectors of various attributes between any two entity words is denoted as semantic similarity.

[0026] The consistency coefficient is the sum of the semantic similarity of all attributes between any two entity words and the character similarity.

[0027] In one embodiment, the process of obtaining the alignment confidence is as follows:

[0028] For each of the two entity words, all the entity words connected within their respective chapters are formed into a set of nearest neighbor entities for each of the two entity words, and are denoted as the first set of nearest neighbor entities and the second set of nearest neighbor entities.

[0029] The similarity between the word vectors of each entity word in the first nearest neighbor entity set and each entity word in the second nearest neighbor entity set is denoted as entity similarity.

[0030] The alignment confidence is obtained by combining the entity similarity between all entity words in the first nearest neighbor entity set and all entity words in the second nearest neighbor entity set.

[0031] In one embodiment, the alignment confidence is the mean of the entity similarity between all entity words in the first nearest neighbor entity set and all entity words in the second nearest neighbor entity set.

[0032] In one embodiment, the entity alignment process for entity words within different chapters during knowledge graph construction includes:

[0033] The average alignment confidence among all different entity words in all different chapters is denoted as the entity alignment threshold.

[0034] Perform entity alignment processing on two entity words whose alignment confidence is greater than the entity alignment threshold.

[0035] Secondly, embodiments of this application also provide a knowledge graph-based big data visualization and analysis system for educational resources, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described knowledge graph-based big data visualization and analysis methods for educational resources.

[0036] This application has at least the following beneficial effects:

[0037] This application comprehensively considers the similarity of word vectors and the occurrence of entity words within a chapter to obtain the association coefficient. This can more accurately measure the semantic association between different entity words within the same chapter. Compared with methods that rely solely on single factors such as word vector similarity or co-occurrence frequency, it can more realistically reflect whether there is an actual association between entity words and the strength of that association. When calculating the strength coefficient of the relationship connection between two entity words, it not only considers the direct association between the two entity words but also their relative positions in the knowledge structure. This can more comprehensively reflect the strength of the association between the two entity words in the knowledge system, which helps to construct a knowledge graph that can more accurately reflect the internal logical relationship of knowledge and enhance the integrity and coherence of the knowledge graph.

[0038] Furthermore, by comprehensively considering the character similarity of entity words, the semantic similarity of attributes, and the overlap of connecting entity words, a consistency coefficient and alignment confidence score are obtained. This enables the accurate identification of entity words describing the same concept in different chapters, thereby achieving accurate alignment of entity words within different chapters. This avoids the problem of slight differences in the description of the same entity word in different texts due to the diversity of data sources. It also enables the effective integration of relevant knowledge in different chapters, resulting in a knowledge graph with better consistency and accuracy on a global scale. Attached Figure Description

[0039] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flowchart illustrating the steps of a knowledge graph-based big data visualization and analysis method for educational resources, provided in one embodiment of this application.

[0041] Figure 2 A flowchart illustrating the entity alignment process. Detailed Implementation

[0042] In the description of the embodiments in this application, the words "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary," "or," and "for example" is intended to present the relevant concepts in a specific manner.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. It should be understood that, unless otherwise stated, " / " in this application means "or".

[0044] It should also be noted that the terms "first" and "second" in this application are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0045] The following description, in conjunction with the accompanying drawings, details the specific scheme of the knowledge graph-based big data visualization analysis method and system for educational resources provided in this application.

[0046] Please see Figure 1 The diagram illustrates a flowchart of a knowledge graph-based big data visualization and analysis method for educational resources according to an embodiment of this application. The method includes the following steps:

[0047] Step 1: Obtain educational resources and extract knowledge to obtain each triplet, and obtain the word vector of each entity word in each triplet.

[0048] Currently, educational knowledge graphs mainly include subject-specific knowledge graphs and teacher-specific knowledge graphs. Subject-specific knowledge graphs, based on cognitive development theory and schema theory, aim to uncover the logical relationships between knowledge points and construct a comprehensive semantic network of knowledge. Application scenarios include, but are not limited to, learning path optimization and promoting knowledge integration across disciplines. This application provides an in-depth analysis of the construction of subject-specific knowledge graphs.

[0049] Educational resources are acquired from open government data platforms and open educational resource platforms. These resources primarily consist of a large amount of diverse and heterogeneous course materials, such as textbook content documents in various formats including web pages, PDFs, Word documents, and TXT files. To ensure data accuracy and consistency, data cleaning is required first. Techniques such as string matching, stop word lists, and meaningless word lists are used to remove irrelevant noise information such as headers and footers from the textbook content documents. Accurate knowledge extraction is crucial in the construction of the subject knowledge graph. Knowledge extraction includes named entity recognition, relation extraction, and attribute extraction. The identified entities, relations, and attributes are then grouped into triples. A word vector model is used to obtain the word vectors of each entity word in each triple, as well as the semantic vectors of each attribute of each entity word.

[0050] In this embodiment, DeepKE is used for knowledge extraction. DeepKE is a well-known technology and will not be described in detail here.

[0051] In this embodiment, the BERT model is used to obtain word vectors and semantic vectors. The BERT model is a well-known technology and will not be described in detail here. Based on the ability to obtain word vectors and semantic vectors, the implementer may choose other existing feasible models.

[0052] Step 2: For chapters under the same subject, obtain the correlation coefficient between any two different entity words in each paragraph of each chapter, and obtain the strength coefficient of the relationship connection between any two different entity words in each chapter; perform relationship connection processing on entity words in each chapter during the knowledge graph construction process; obtain the consistency coefficient and alignment confidence between any two different entity words in different chapters, and perform entity alignment processing on entity words in different chapters during the knowledge graph construction process.

[0053] The key to constructing a course knowledge graph lies in accurately identifying the logical relationships between knowledge points. However, the fragmented nature of educational resource data, the density of professional terminology, and the richness of implicit knowledge can easily lead to difficulties in knowledge integration and the break in knowledge connections. This results in the failure to accurately identify, extract, or establish the logical connections between knowledge points that should exist, thereby affecting the completeness and accuracy of the knowledge graph.

[0054] Step 2.1: Obtain the correlation coefficient between any two entity words by using the similarity of word vectors between any two entity words in each paragraph of each chapter, and by using the occurrence of any two entity words in each chapter.

[0055] Triples are the basic data units for constructing knowledge graphs. A clear semantic relationship must exist between the two entities that make up a triple. The stronger the semantic relationship, the more obvious the content similarity features and co-occurrence statistics between the entities. Due to the fragmented nature of teaching content, two entities belonging to the same paragraph are likely to have stronger semantic connections. Therefore, for chapters under the same subject, taking entity 1 and entity 2 in the same paragraph of the i-th chapter as an example, the similarity of the word vectors between entity 1 and entity 2 is denoted as vector similarity, used to reflect the degree of content connection between entity 1 and entity 2. However, in course texts, some entities, although not obviously related, may create the illusion of a strong connection due to frequent random co-occurrence in paragraphs. For example, in a history text introducing the Industrial Revolution, the entities "worker" and "coal" may have a high co-occurrence probability. However, in actual descriptions, these two terms do not have a direct semantic or functional connection; their strong co-occurrence is mainly driven by a shared contextual theme. Conversely, the semantic connection between the entities "coal" and "steam engine" is clearly stronger, even with a high co-occurrence probability. Therefore, the lower the probability of two unrelated entities co-occurring, the more pronounced their co-occurrence statistical characteristics are when they frequently co-occur in multiple paragraphs. To obtain these co-occurrence statistical characteristics, the percentage of paragraphs in the i-th chapter where entity 1 appears is denoted as the first percentage, reflecting the probability of entity 1 appearing in a particular paragraph within the i-th chapter. If there is no real connection between entity word 1 and entity word 2, but the higher the frequency of their occurrence, the greater the probability of random co-occurrence. Therefore, by calculating the probabilities of entity word 1 and entity word 2 occurring together and individually, the true co-occurrence connection characteristics of entity word 1 and entity word 2 can be quantitatively evaluated. This method is more accurate than conventional co-occurrence analysis methods, which typically only consider the co-occurrence characteristics of words within a local scope. The proportion of paragraphs in the i-th chapter where entity word 1 and entity word 2 appear simultaneously among all paragraphs is denoted as the second proportion.

[0056] In this embodiment, the similarity between word vectors is cosine similarity. The calculation of cosine similarity is a well-known technique and will not be described in detail here. As other implementation methods, based on the ability to measure the similarity between word vectors, implementers may use other calculation methods, such as the reciprocal of Euclidean distance. This application does not impose any special restrictions.

[0057] Based on the above analysis, the correlation coefficient between entity word 1 and entity word 2 is obtained by analyzing the similarity of their word vectors and their occurrence within the same chapter. This coefficient is used to characterize the semantic correlation between entity word 1 and entity word 2 within the same paragraph. Specifically:

[0058] Calculate the product of the first proportions of entity word 1 and entity word 2, and calculate the ratio of the second proportion to the product; the correlation coefficient between entity word 1 and entity word 2 is positively correlated with the vector similarity and the ratio, respectively.

[0059] It should be noted that positive correlation means that the variables change in the same direction; when one variable increases, the other variable also increases, and when one variable decreases, the other variable also decreases.

[0060] In this embodiment, the vector similarity is mapped to a positive number, and the correlation coefficient between entity word 1 and entity word 2 is the product of the positive number and the ratio.

[0061] In another embodiment, the vector similarity is mapped to a positive number, and the correlation coefficient between entity word 1 and entity word 2 is the sum of the positive number and the ratio.

[0062] It should be noted that the purpose of mapping the vector similarity to a positive number is to ensure that the calculated result of the correlation coefficient is a positive number. There are many ways to map data to a positive number. Specifically, it can be achieved by treating the data as the exponent of an exponential function with the natural constant as the base, and calculating the sum of the data and a preset value greater than 1. This application does not impose any special restrictions on this. Unless otherwise specified, this application achieves this by treating the data as the exponent of an exponential function with the natural constant as the base.

[0063] It should be noted that analyzing the semantic relevance between entity word 1 and entity word 2 by examining the content similarity and co-occurrence statistical characteristics can help to more accurately measure the degree of interdependence between entity word 1 and entity word 2, thereby reducing the bias caused by random and frequent co-occurrence.

[0064] Calculate the correlation coefficient between any two different entity words in each paragraph within each chapter, following the method used to calculate the correlation coefficient between entity word 1 and entity word 2.

[0065] Step 2.2: Construct a weighted undirected graph for each chapter, where entity words are nodes and the association coefficients are weights of the weighted edges. By using the weights of all weighted edges on the shortest path between any two nodes in each weighted undirected graph, and the importance of all nodes on the shortest path, obtain the strength coefficient of the relationship connection between the entity words corresponding to any two nodes.

[0066] Furthermore, some core entity concepts appear frequently in paragraphs with different themes, playing a pivotal role in connecting various knowledge modules. This allows entity words that originally belonged to different paragraphs to be interconnected through these central entity words, forming significant cross-paragraph relationships. For example, entity word 1 is related to entity words 2 and 3 respectively, while entity words 2 and 3, although not directly related, can still form a triple for relational connection.

[0067] Based on the above analysis, a weighted undirected graph is constructed for each chapter. Entity words are used as nodes, and the correlation coefficient between two entity words is used as the weight of the weighted edge. A higher weight indicates a stronger correlation between the two entity words in the paragraph. The importance parameters of each node in each weighted undirected graph are then obtained to reflect the importance of each node within the graph. For example, if any node is located at the center of the weighted undirected graph and is connected to most other nodes, it indicates high importance. In the context of the course text, this means that the entity word represented by that node carries or is associated with a larger amount of information within the chapter. Taking entity words 4 and 5 in the i-th chapter as an example, there is a weighted edge connection between the node corresponding to entity word 4 and the node corresponding to entity word 5. If the average importance of all nodes on the connection path between the nodes corresponding to entity word 4 and the node corresponding to entity word 5 is higher, and the average weight of all weighted edges on the connection path is also greater, it indicates a stronger semantic correlation between entity words 4 and 5.

[0068] Based on the above analysis, the strength coefficient of the relational connection between entity word 4 and entity word 5 is obtained by using the weights of all weighted edges on the shortest path between the node corresponding to entity word 4 and the node corresponding to entity word 5 in the weighted undirected graph of the i-th chapter, as well as the importance of all nodes on the shortest path. Specifically:

[0069] Calculate the average weight of all weighted edges on the shortest path between the node corresponding to entity word 4 and the node corresponding to entity word 5; obtain the importance parameters of each node in the weighted undirected graph of the i-th chapter, and calculate the average value of the importance parameters of all nodes on the shortest path; the intensity coefficient between entity word 4 and entity word 5 is obtained by fusing the average value with the average value.

[0070] It should be noted that fusion refers to combining multiple independent variables in a way that enhances the overall effect, such as by addition or multiplication. This application does not impose any special restrictions on this.

[0071] In this embodiment, the importance parameter of a node is specifically betweenness centrality. The calculation of betweenness centrality is a well-known technique and will not be described in detail here. As other implementation methods, implementers may use other existing techniques, such as degree centrality, to measure the importance of a node. This application does not impose any special restrictions.

[0072] In this embodiment, the intensity coefficient between entity word 4 and entity word 5 is the product of the mean and the average.

[0073] In another embodiment, the intensity coefficient between entity word 4 and entity word 5 is the sum of the mean and the average.

[0074] It should be noted that the shortest path represents the most direct connection between two nodes, meaning that establishing a connection through the fewest intermediate concepts or entities avoids introducing unnecessary or redundant indirect associations, thus more purely reflecting the core semantic connection between the two entity words. The larger the calculated strength coefficient, the more obvious the association features between entity words 4 and 5 under the connection of the central word. By introducing the node importance feature, the degree of association between entity words is further reflected from the perspective of weighted undirected graph connection relationships. This allows the construction of knowledge graphs to more accurately capture the inherent, structured logical relationships of subject knowledge, and to a certain extent, reduce the problem of knowledge association breaks caused by the fragmentation of educational resource data, thereby improving the completeness and accuracy of knowledge graphs.

[0075] Calculate the intensity coefficient between any two different entity words in each chapter, following the method used to calculate the intensity coefficient between entity words 4 and 5.

[0076] Step 2.3: By analyzing the distribution of the intensity coefficients between two entity words in all triples across all chapters, relational connection processing is performed on entity words within each chapter during the knowledge graph construction process.

[0077] Furthermore, by analyzing the distribution of strength coefficients between two entity words in all triples across all chapters, relational connection processing is performed on entity words within each chapter during the knowledge graph construction process. Specifically:

[0078] Calculate the arithmetic mean of the intensity coefficients between two entity words in all triples of all chapters, and perform relational connection processing on two entity words whose intensity coefficients are greater than or equal to the arithmetic mean.

[0079] Step 2.4: Obtain the consistency coefficient between any two entity words by measuring the similarity between characters of any two entity words in different chapters and the semantic similarity of various attributes between any two entity words. Combine this with the overlap of connecting entity words between any two entity words in their respective chapters to obtain the alignment confidence between any two entity words. This is used to perform entity alignment processing on entity words in different chapters during the knowledge graph construction process.

[0080] Furthermore, due to the diversity of data sources, different texts may contain the same entity words with slightly different descriptions, such as "Newton's First Law" and "Newton's First Law of Motion" referring to the same concept. In the process of constructing the knowledge graph of the entire discipline, it is necessary to align entities with the same concept. Therefore, one approach is to evaluate entity words through their textual descriptions and attributes. The more similar the strings of entity word names are, and the higher the semantic similarity of their corresponding attributes, the more likely the two entity words are to point to the same concept. Specifically, taking entity word 6 in the i-th chapter and entity word 7 in the j-th chapter as examples, the similarity between the characters of entity word 6 and entity word 7 is recorded as character similarity; the similarity of the semantic vectors of various attributes between entity word 6 and entity word 7 is recorded as semantic similarity; and the sum of the semantic similarity of all attributes between entity word 6 and entity word 7 and the character similarity is used as the consistency coefficient between entity word 6 and entity word 7. The consistency coefficient reflects the consistent characteristics of textual descriptions and attributes between entity words in different chapters. Another approach is to evaluate entity words through their connection relationships. The higher the overlap rate of other entity words connected to entity words 6 and 7 from different chapters, the more consistent the concepts of entity words 6 and 7 are. From the i-th chapter, extract all entity words connected to entity word 6 to form the nearest neighbor entity set of entity word 6, denoted as the first nearest neighbor entity set. From the j-th chapter, extract all entity words connected to entity word 7 to form the nearest neighbor entity set of entity word 7, denoted as the second nearest neighbor entity set. The similarity of the word vectors between each entity word in the first nearest neighbor entity set and each entity word in the second nearest neighbor entity set is denoted as the entity similarity.

[0081] The mean of the entity similarity between all entity words in the first nearest neighbor entity set and all entity words in the second nearest neighbor entity set is used as the alignment confidence between entity word 6 and entity word 7.

[0082] Calculate the alignment confidence between any two different entity words in different chapters, following the same method used to calculate the alignment confidence between entity word 6 and entity word 7.

[0083] Furthermore, the average alignment confidence score among all different entity words within all different chapters is recorded as the entity alignment threshold. Two entity words with an alignment confidence score greater than the entity alignment threshold are considered the same entity and undergo entity alignment processing. A flowchart illustrating the entity alignment process is shown below. Figure 2 As shown.

[0084] In this embodiment, the similarity between the characters of two entity words is the Jaccard similarity, and the similarity between the semantic vectors of various attributes of two entity words is the reciprocal of the Euclidean distance. Both the Jaccard similarity and the Euclidean distance are well-known techniques and will not be described in detail here. As other implementation methods, based on the ability to measure the similarity between the characters of two entity words and the similarity between the semantic vectors of various attributes of two entity words, the implementer may adopt other existing feasible techniques, which will not be described in detail here.

[0085] Compared to conventional alignment methods, by analyzing the textual descriptions and attribute information of entity words, and combining the context in which the entity words are located, we can effectively capture subtle differences and semantic equivalence at the textual description level.

[0086] Step 3: Construct a knowledge graph.

[0087] By integrating aligned entities and relationships from various chapters, a globally unified knowledge network is formed, thereby enabling the construction of knowledge graphs for each subject. In the visualization phase, this application employs the Neo4j Bloom tool, using force-directed layout to naturally cluster closely related entities, visually revealing the knowledge community. Different node colors and sizes are used to distinguish the type or importance of entities; for example, core concept nodes are larger and brighter. Furthermore, Neo4j Bloom provides interactive features, such as highlighting related paths after clicking a node, searching for and locating specific entities, and zooming to browse different levels of knowledge details, thus helping users efficiently explore and understand complex subject knowledge systems.

[0088] Based on the same inventive concept as the above methods, this application also provides a knowledge graph-based big data visualization and analysis system for educational resources, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described knowledge graph-based big data visualization and analysis methods for educational resources.

[0089] In summary, this application comprehensively considers the similarity of word vectors and the occurrence of entity words within a chapter to obtain the association coefficient. This approach can more accurately measure the semantic association between different entity words within the same chapter. Compared to methods that rely solely on single factors such as word vector similarity or co-occurrence frequency, it more realistically reflects whether there is an actual association between entity words and the strength of that association. When calculating the strength coefficient of the relationship between two entity words, it considers not only the direct association between the two entity words but also their relative positions in the knowledge structure. This more comprehensively reflects the strength of the association between the two entity words in the knowledge system, helping to construct a knowledge graph that more accurately reflects the internal logical relationships of knowledge and enhancing the completeness and coherence of the knowledge graph.

[0090] Furthermore, by comprehensively considering the character similarity of entity words, the semantic similarity of attributes, and the overlap of connecting entity words, a consistency coefficient and alignment confidence score are obtained. This enables the accurate identification of entity words describing the same concept in different chapters, thereby achieving accurate alignment of entity words within different chapters. This avoids the problem of slight differences in the description of the same entity word in different texts due to the diversity of data sources. It also enables the effective integration of relevant knowledge in different chapters, resulting in a knowledge graph with better consistency and accuracy on a global scale.

[0091] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0092] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from its essential characteristics. Therefore, the embodiments described above should be considered exemplary and non-limiting in all respects.

Claims

1. A knowledge graph-based big data visualization and analysis method for educational resources, characterized in that, The method includes the following steps: Acquire educational resources and extract knowledge to obtain each triplet, and obtain the word vector of each entity word in each triplet; For chapters within the same subject, the association coefficient between any two entity words is obtained by considering the similarity of their word vectors within each paragraph and the occurrence of those two entity words across chapters. This coefficient is used to construct a weighted undirected graph for each chapter, where entity words are nodes and the association coefficient is the weight of each edge. The strength coefficient of the relationship between the entity words corresponding to any two nodes is obtained by considering the weights of all weighted edges on the shortest path between any two nodes in each weighted undirected graph and the importance of all nodes on that shortest path. Finally, the distribution of the strength coefficient between two entity words in all triples across all chapters is used to perform relationship connection processing on the entity words within each chapter during the knowledge graph construction process. By measuring the similarity between characters of any two entity words in different chapters and the semantic similarity of various attributes between the two entity words, a consistency coefficient between the two entity words is obtained. In addition, by combining the overlap of the entity words connecting the two entity words in their respective chapters, an alignment confidence score between the two entity words is obtained, so as to perform entity alignment processing on entity words in different chapters during the knowledge graph construction process.

2. The knowledge graph-based big data visualization and analysis method for educational resources as described in claim 1, characterized in that, The process of obtaining the correlation coefficient is as follows: The similarity between the word vectors of any two entity words is denoted as vector similarity. The percentage of paragraphs in which each entity word appears in each chapter is recorded as the first percentage among all paragraphs. The percentage of paragraphs in which any two entity words appear simultaneously in each chapter is recorded as the second percentage; The correlation coefficient is obtained by combining the first proportion, the second proportion, and the vector similarity of any two entity words.

3. The knowledge graph-based big data visualization and analysis method for educational resources as described in claim 2, characterized in that, The correlation coefficient is calculated as follows: Calculate the product of the first proportions of any two entity words, and calculate the ratio of the second proportion to the product; The correlation coefficients are positively correlated with the vector similarity and the ratio, respectively.

4. The knowledge graph-based big data visualization and analysis method for educational resources as described in claim 1, characterized in that, The process of obtaining the strength coefficient is as follows: Calculate the average weight of all weighted edges on the shortest path between any two nodes; Obtain the importance parameters of each node in each weighted undirected graph, and calculate the average value of the importance parameters of all nodes on the shortest path; The intensity coefficient is obtained by fusing the mean and the average.

5. The knowledge graph-based big data visualization and analysis method for educational resources as described in claim 1, characterized in that, The process of performing relational connection processing on entity words within each chapter during knowledge graph construction includes: Calculate the arithmetic mean of the intensity coefficients between two entity words in all triples of all chapters, and perform relational connection processing on two entity words whose intensity coefficients are greater than or equal to the arithmetic mean.

6. The knowledge graph-based big data visualization and analysis method for educational resources as described in claim 1, characterized in that, The process of obtaining the consistency coefficient is as follows: The similarity between the characters of any two entity words is denoted as character similarity. The similarity of the semantic vectors of various attributes between any two entity words is denoted as semantic similarity. The consistency coefficient is the sum of the semantic similarity of all attributes between any two entity words and the character similarity.

7. The knowledge graph-based big data visualization and analysis method for educational resources as described in claim 1, characterized in that, The process of obtaining the alignment confidence is as follows: For each of the two entity words, all the entity words connected within their respective chapters are formed into a set of nearest neighbor entities for each of the two entity words, and are denoted as the first set of nearest neighbor entities and the second set of nearest neighbor entities. The similarity between the word vectors of each entity word in the first nearest neighbor entity set and each entity word in the second nearest neighbor entity set is denoted as entity similarity. The alignment confidence is obtained by combining the entity similarity between all entity words in the first nearest neighbor entity set and all entity words in the second nearest neighbor entity set.

8. The knowledge graph-based big data visualization and analysis method for educational resources as described in claim 7, characterized in that, The alignment confidence is the average of the entity similarities between all entity words in the first nearest neighbor entity set and all entity words in the second nearest neighbor entity set.

9. The knowledge graph-based big data visualization and analysis method for educational resources as described in claim 1, characterized in that, The entity alignment process for entity words in different chapters during knowledge graph construction includes: The average alignment confidence among all different entity words in all different chapters is denoted as the entity alignment threshold. Perform entity alignment processing on two entity words whose alignment confidence is greater than the entity alignment threshold.

10. A knowledge graph-based big data visualization and analysis system for educational resources, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the knowledge graph-based big data visualization and analysis method for educational resources as described in any one of claims 1-9.

Images