Discipline-crossing course-oriented knowledge atom construction method and system

Abstract

This invention discloses a method and system for constructing knowledge atoms for interdisciplinary courses. The method includes: Step 1: Preliminarily dividing the original interdisciplinary courses to obtain original interdisciplinary course units; Step 2: Calculating the semantic information entropy of the original interdisciplinary course units, and determining whether to deconstruct them based on the semantic information entropy; if necessary, using a clustering algorithm to divide the original interdisciplinary course units into multiple sub-course candidate sets; Step 3: Calculating the principal component proportion of the subject distribution vector of each sub-course candidate set; if the principal component proportion is greater than or equal to the purity termination threshold, it is used as an atomic knowledge point; if the principal component proportion is less than the purity termination threshold, steps 2 and 3 are repeated; Step 4: Calculating the vertical dependency association and horizontal homogeneous association of each atomic knowledge point, constructing the association matrix of each atomic knowledge point and the overall association network. This invention achieves precise knowledge decomposition, efficient management, flexible combination, and dynamic iteration.

Description

Technical Field

[0001] This invention relates to the fields of knowledge management and intelligent teaching technology, and in particular to a method and system for constructing knowledge atoms for interdisciplinary courses. Background Technology

[0002] In the context of the trend of cutting-edge multidisciplinary integration, courses related to the field of intelligence (such as interdisciplinary courses in artificial intelligence, intelligent manufacturing, and intelligent unmanned systems) have become the core vehicle for cultivating interdisciplinary technical talents. The knowledge system of these intelligent multidisciplinary courses exhibits significant cross-domain integration characteristics. Their core knowledge points often span multiple disciplines such as computer science, automation, mechanical engineering, and electronic information, and require support from the fundamental theories, technical methods, and practical applications of multiple disciplines. This results in a fragmented knowledge structure, complex cross-domain connections, and frequent technological iterations. Currently, the teaching and management of knowledge in intelligent multidisciplinary courses still rely on traditional document-based carriers (such as subject-specific textbooks, single-domain lecture notes, and scattered academic materials), which are difficult to adapt to the teaching needs of cross-domain integration courses and cannot solve problems such as knowledge integration, association, and dynamic updating. The following are the core pain points:

[0003] Knowledge retrieval is inefficient and inaccurate: Traditional management methods store documents by subject, requiring users to sift through massive amounts of cross-disciplinary documents to find the target knowledge.

[0004] High knowledge update and maintenance costs: The cross-disciplinary integration technology in the field of intelligence is updated and iterated frequently, and new cross-theoretical and integration methods are constantly emerging. In traditional document management, updating one piece of knowledge requires simultaneous adjustment of related documents of multiple disciplines, resulting in low maintenance efficiency and high update costs.

[0005] Weak ability to adapt personalized knowledge: Different users (such as beginners, R&D personnel, and teachers) have significantly different needs for interdisciplinary integrated knowledge (beginners need content that connects basic concepts across disciplines, R&D personnel need integrated application solutions of multidisciplinary technologies, and teachers need a logical framework for cross-domain knowledge). Traditional large-granular knowledge carriers cannot achieve on-demand combination and are difficult to support personalized learning and cross-domain R&D guidance.

[0006] The knowledge connections are unclear and the reusability is poor: Traditional knowledge is simply linked by a single discipline's catalog structure, which cannot intuitively reflect the cross-relationship between core knowledge points of different disciplines (such as the supporting relationship between basic theories and cross-domain applications, and the complementary relationship between different disciplines' technical methods), resulting in low cross-disciplinary knowledge reuse rate and difficulty in the integration and migration of technologies across multiple fields.

[0007] While existing knowledge management technologies include knowledge graphs and document fragmentation, none of them address the knowledge characteristics of path planning algorithms by implementing the smallest granularity of knowledge decomposition and standardized representation, thus failing to solve the core pain points of path planning knowledge management. Summary of the Invention

[0008] To address the problems of inefficient cross-domain knowledge retrieval, difficult updating and maintenance, insufficient personalization, and ambiguous cross-disciplinary knowledge relationships in existing knowledge management of cutting-edge interdisciplinary courses, this invention provides a method and system for constructing knowledge atoms for interdisciplinary courses. This method decomposes the cross-domain knowledge of the original interdisciplinary course into the smallest granularity of knowledge atoms to obtain atomic knowledge points. Based on these atomic knowledge points, an association matrix is ​​constructed, enabling accurate retrieval, efficient updating, flexible combination, and intelligent application of interdisciplinary knowledge. This reduces the cost of knowledge management for cutting-edge interdisciplinary courses and improves the efficiency of cross-domain knowledge utilization.

[0009] To achieve the above objectives, the technical solution of the present invention is as follows:

[0010] The first aspect of this invention proposes a method for constructing knowledge atoms for interdisciplinary courses, comprising:

[0011] Step 1: Initially divide the original interdisciplinary courses into their own units to obtain the original interdisciplinary course units;

[0012] Step 2: Calculate the semantic information entropy of the original interdisciplinary course unit, and determine whether to deconstruct the original interdisciplinary course unit based on the semantic information entropy; if it is necessary to deconstruct the original interdisciplinary course unit, use a clustering algorithm to divide the original interdisciplinary course unit into multiple sub-course candidate sets, which are used to atomize the original interdisciplinary course unit.

[0013] Step 3: Calculate the principal component proportion of the subject distribution vector of each sub-course candidate set based on all existing courses. Compare the principal component proportion with the purity termination threshold. If the principal component proportion is greater than or equal to the purity termination threshold, the sub-course candidate set is used as an atomic knowledge point. If the principal component proportion is less than the purity termination threshold, repeat steps 2 and 3 to obtain the final atomic knowledge points.

[0014] Step 4: Calculate the vertical dependency and horizontal homogeneity associations for each atomic knowledge point. Construct the association matrix for each atomic knowledge point based on the vertical dependency and horizontal homogeneity associations, and build an association network from all association matrices. This facilitates accurate retrieval, efficient updating, flexible combination, and intelligent application of cross-disciplinary knowledge.

[0015] Furthermore, the step of determining whether to deconstruct the original interdisciplinary course unit based on semantic information entropy specifically includes:

[0016] The semantic information entropy of the original interdisciplinary course unit is compared with a preset threshold. If the semantic information entropy is greater than the preset threshold, the original interdisciplinary course unit is deconstructed. If the semantic information entropy is less than or equal to the preset threshold, it is not deconstructed and the original interdisciplinary course unit is treated as an atomic knowledge point, which facilitates the atomization of the original interdisciplinary course unit.

[0017] Furthermore, the calculation of the principal component proportion of the subject distribution vector of each sub-course candidate set specifically includes:

[0018] All existing courses are divided into different subject corpora according to their disciplines; the subject corpora include subject units divided according to the course itself.

[0019] The subject distribution vector of the sub-course candidate set is obtained based on different subject corpora and probability models; the subject distribution vector includes the weight of the sub-course candidate set in different subject corpora;

[0020] The principal component percentage is obtained by dividing the largest weight in the subject distribution vector by the sum of all weights.

[0021] Furthermore, the vertical dependency association is obtained through the following steps:

[0022] Calculate the prior probability between each atomic knowledge point and each remaining atomic knowledge point. If the prior probability is greater than or equal to the prior threshold, it is determined that there is a vertical dependency relationship between the two. If the prior probability is less than the prior threshold, it is determined that there is no vertical dependency relationship between the two.

[0023] Furthermore, the preceding probability is expressed by the following formula:

[0024]

[0025] in, This indicates the number of times that atomic knowledge point A and atomic knowledge point B appear simultaneously within the same knowledge context, and atomic knowledge point A appears before atomic knowledge point B. The frequency of occurrence of atomic knowledge point B, This represents the vertical dependency relationship between atomic knowledge point B and atomic knowledge point A.

[0026] Furthermore, the lateral homogeneous association is obtained through the following steps:

[0027] All existing courses are categorized by subject and deconstructed into atomic knowledge points, resulting in multiple subject databases based on these atomic knowledge points.

[0028] Calculate the semantic vectors of each atomic knowledge point in the original interdisciplinary course and atomic knowledge points in different subject libraries, and calculate the cosine similarity between the semantic vectors. If the cosine similarity is greater than or equal to the similarity threshold, it is determined that the two have a horizontal homogeneous association. If the cosine similarity is less than the similarity threshold, it is determined that the two have no horizontal homogeneous association.

[0029] Furthermore, the process of constructing an association matrix for each atomic knowledge point based on vertical dependency and horizontal homogeneity, and then constructing an association network from all association matrices, specifically includes:

[0030] Based on vertical dependency and horizontal homogeneity, an association matrix is ​​constructed for each atomic knowledge point of the original interdisciplinary course. An association network is then constructed based on all association matrices. The association matrix includes the atomic knowledge point, the location information corresponding to the atomic knowledge point, and the atomic knowledge points that are vertically dependent and horizontally homogeneously associated with that atomic knowledge point.

[0031] The location information includes subject, difficulty, activity level, and number of interdisciplinary associations; the difficulty is manually assigned, the number of interdisciplinary associations is the total number of atomic knowledge points that are horizontally homogeneously associated with the atomic knowledge point, the activity level is the number of interdisciplinary associations divided by the total number of associations, and the total number of associations is the total number of atomic knowledge points that are vertically dependent and horizontally homogeneously associated with the atomic knowledge point.

[0032] A second aspect of this invention proposes a knowledge atom construction system for interdisciplinary courses, comprising:

[0033] The preliminary division module is used to initially divide the original interdisciplinary courses according to their own units, thus obtaining the original interdisciplinary course units;

[0034] The atomization module is used to calculate the semantic information entropy of the original interdisciplinary course unit and determine whether to deconstruct the original interdisciplinary course unit based on the semantic information entropy. If it is necessary to deconstruct the original interdisciplinary course unit, a clustering algorithm is used to divide the original interdisciplinary course unit into multiple sub-course candidate sets for atomization.

[0035] The judgment module is used to calculate the principal component ratio of the subject distribution vector of each sub-course candidate set. Based on the principal component ratio, it determines whether the sub-course candidate set meets the atomization standard. If it meets the atomization standard, the sub-course candidate set is used as an atomic knowledge point. If it does not meet the atomization standard, the atomization module and the judgment module are re-executed to obtain the final atomic knowledge point.

[0036] The association matrix module is used to calculate the vertical dependency associations and horizontal homogeneous associations of each atomic knowledge point. Based on the vertical dependency associations and horizontal homogeneous associations, an association matrix is ​​constructed for each atomic knowledge point, and all association matrices are used to construct an association network, which facilitates accurate retrieval, efficient updating, flexible combination and intelligent application of cross-disciplinary knowledge.

[0037] A third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method described in the first aspect above.

[0038] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect above.

[0039] The beneficial effects of this invention are:

[0040] This invention achieves standardization and refinement of knowledge granularity: by employing a recursive deconstruction algorithm based on knowledge entropy, it solves the problem of excessively coarse granularity in chapter knowledge in traditional teaching resources. This invention can automatically decompose complex course content into indivisible "knowledge atoms," providing a standardized data foundation for the comparison, reuse, and precise mapping of cross-disciplinary content.

[0041] This invention constructs a dynamic and three-dimensional interdisciplinary logical topology (association matrix and association network): unlike the traditional unidirectional tree-like outline, this invention automatically constructs a network of vertical support (basic and applied knowledge) and horizontal mapping (homogeneous knowledge) through the association characteristics in multi-dimensional feature vectors. This structure can reveal the underlying logical connections between disciplines, solving the pain point of difficulty in quantifying and identifying interdisciplinary intersections.

[0042] This invention possesses a high degree of associative inheritance and logical rigor: by introducing an "associative inheritance mechanism" during the deconstruction process, it ensures that the original logical chain of knowledge points is not lost when they evolve from coarse-grained to fine-grained, thus realizing the complete tracing of knowledge context and effectively avoiding the generation of knowledge silos.

[0043] This invention significantly improves the automation level of interdisciplinary system construction: it greatly reduces the cost of manually compiling interdisciplinary outlines and can automatically adjust the weight and links of knowledge atoms according to the corpus update, adapting to the needs of the rapid development of cutting-edge interdisciplinary directions. Attached Figure Description

[0044] Figure 1 A flowchart illustrating a method for constructing knowledge atoms for interdisciplinary courses, provided as an embodiment of the present invention.

[0045] Figure 2This is an architecture diagram of a knowledge atom construction system for interdisciplinary courses provided in an embodiment of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0047] Example 1

[0048] like Figure 1 As shown, this invention proposes a method for constructing knowledge atoms for interdisciplinary courses, including:

[0049] S101: The original interdisciplinary courses are initially divided into their own units to obtain the original interdisciplinary course units;

[0050] S102: Calculate the semantic information entropy of the original interdisciplinary course unit, and determine whether to deconstruct the original interdisciplinary course unit based on the semantic information entropy; if it is necessary to deconstruct the original interdisciplinary course unit, use a clustering algorithm to divide the original interdisciplinary course unit into multiple sub-course candidate sets.

[0051] S103: Calculate the principal component proportion of the subject distribution vector of each sub-course candidate set based on all existing courses, compare the principal component proportion with the purity termination threshold, if the principal component proportion is greater than or equal to the purity termination threshold, then the sub-course candidate set is used as an atomic knowledge point, if the principal component proportion is less than the purity termination threshold, then S102 and S103 are executed again.

[0052] S104: Calculate the vertical dependency association and horizontal homogeneous association for each atomic knowledge point, construct the association matrix for each atomic knowledge point based on the vertical dependency association and horizontal homogeneous association, and construct the association network from all association matrices.

[0053] This invention first preliminarily divides the original interdisciplinary courses into original interdisciplinary course units. Then, it deconstructs these units using a recursive deconstruction algorithm based on knowledge entropy to obtain atomic knowledge points. Specifically, it calculates the semantic information entropy of each original interdisciplinary course unit and determines whether to deconstruct it based on the semantic information entropy. For original interdisciplinary course units that need to be deconstructed, it uses a clustering algorithm to divide them into multiple sub-course candidate sets. Next, it calculates the principal component proportions of the subject distribution vectors of the sub-course candidate sets to determine whether atomic knowledge points have been completed. Finally, it calculates the vertical dependency associations and horizontal homogeneous associations of each atomic knowledge point, constructs an association matrix for each atomic knowledge point based on these associations, and builds an association network from all association matrices. This invention achieves accurate retrieval, efficient updating, flexible combination, and intelligent application of interdisciplinary knowledge, reduces the management cost of cutting-edge interdisciplinary course knowledge, and improves the efficiency of cross-domain knowledge utilization.

[0054] Example 2

[0055] Based on the above embodiments, this invention proposes a specific process for constructing knowledge atoms for interdisciplinary courses, including:

[0056] S201: The original interdisciplinary courses are initially divided into their own units to obtain the original interdisciplinary course units.

[0057] Specifically, the original interdisciplinary course units are divided according to the original interdisciplinary course units themselves.

[0058] Preferably, as an implementation method, during the initial segmentation process, a pre-trained semantic model (such as BERT) can be used to extract the semantic vector of each course unit, and the similarity between the semantic vector of the course unit and the semantic vector of each course unit in all existing courses in the system can be calculated. The existing course unit with the highest similarity is determined as the direct predecessor dependency of the current course unit, and this similarity value is used as the initial probability link to provide the initialization basis for the association inheritance in the subsequent recursive deconstruction process.

[0059] Specifically, the formula for calculating the initial probability link is:

[0060]

[0061] in, The original interdisciplinary course units currently awaiting processing, For any course unit in the existing courses in the system, and These are the semantic vectors extracted by their respective pre-trained semantic models. Represents the dot product of vectors. This represents the magnitude of the vector. The system iterates through all existing course units. And calculate Take the value corresponding to As The initial pre-dependent unit is determined, and the maximum probability value is recorded.

[0062] S202: Calculate the semantic information entropy of the original interdisciplinary course unit, and determine whether to deconstruct the original interdisciplinary course unit based on the semantic information entropy; if it is necessary to deconstruct the original interdisciplinary course unit, use a clustering algorithm to divide the original interdisciplinary course unit into multiple sub-course candidate sets.

[0063] Specifically, for the interdisciplinary course unit U, calculate the semantic information entropy H(U) within unit U. When H(U) > t au (A preset threshold) indicates that the unit contains multiple heterogeneous knowledge clusters, triggering the deconstruction process. Using a clustering algorithm (such as K-means++), U is divided into several sub-course candidate sets {U1, U2, ..., U...}. n}

[0064] Preferably, as one implementation, after obtaining the candidate sub-course sets, semantic coupling detection can be performed on any two candidate sub-course sets: for any two candidate sub-course sets and The core knowledge elements are extracted from each text. These core knowledge elements refer to key terms, variables, or function names defined in the text. They are identified by extracting words near defining expressions such as "called," "defined as," and "that is," as the definition element set. Extract keywords from the text located in algorithm steps, formula parameter descriptions, and other reference positions as a set of reference elements. .

[0065] Then, the bidirectional encoding representation model (BERT) is used to respectively... , The various elements in it, and , Semantic vectors are extracted from each element, and cosine similarity is calculated between each pair of elements. If there exists a pair of elements whose cosine similarity is greater than or equal to a preset coupling threshold (e.g., 0.85), and these two elements belong to the definition element set of one sub-course candidate set and the reference element set of another sub-course candidate set, respectively, then it is determined that these two sub-course candidate sets have an internal functional coupling relationship, and an internal association link is established from the referenced party to the referrer.

[0066] For example, if the sub-course candidate set The term "heuristic function" is defined in [the document / reference]. Sub-course candidate set The algorithm steps reference " "After BERT extracts the semantic vectors of the two and calculates the cosine similarity, it is 0.93, which is greater than the coupling threshold of 0.85, so it is determined that..." Functionally dependent on Automatically create from point to Internal related links.

[0067] S203: Calculate the principal component proportion of the subject distribution vector of each sub-course candidate set based on all existing courses, and compare the principal component proportion with the purity termination threshold. If the principal component proportion is greater than or equal to the purity termination threshold, the sub-course candidate set is used as an atomic knowledge point. If the principal component proportion is less than the purity termination threshold, S202 and S203 are executed again.

[0068] Specifically, all existing courses are divided into different subject corpora based on their disciplines; these subject corpora include subject units divided according to the course itself. The subject distribution vectors of the sub-course candidate sets are obtained based on the different subject corpora and probabilistic models. Specifically, for each sub-course candidate set, the principal component proportion (i.e., subject purity) of its subject distribution vector is calculated. For example, a probabilistic model (such as the Latent Dirichlet Allocation (LDA) model) is used to calculate the weights of the sub-course candidate set in different subject corpora such as mathematics, computer science, and automation. The principal component proportion is obtained by dividing the largest weight in the subject distribution vector by the sum of all weights.

[0069] If the proportion of the principal component r ≥ t au2 If the purity threshold is 0.8, then the candidate set is determined to have reached the atomization standard, the splitting is stopped, and it is defined as an atomic knowledge point, thus completing knowledge atomization. If the principal component proportion r <t au2 If this indicates that the candidate set of the sub-course is still a mixture of multiple disciplines, then it is used as the new input, and S202 and S203 are executed recursively.

[0070] Preferably, a globally unique identifier is automatically generated when the candidate set of sub-courses is determined to meet the atomization criteria. This identifier is used to uniquely identify the atomic knowledge point in the subsequent association matrix and association network, and its generation method may include generating a globally unique string based on timestamps, system identifiers, and / or random numbers.

[0071] S204: Calculate the vertical dependency association and horizontal homogeneous association for each atomic knowledge point, construct the association matrix for each atomic knowledge point based on the vertical dependency association and horizontal homogeneous association, and construct the association network from all association matrices.

[0072] Specifically, vertical dependency essentially characterizes the prerequisite dependency between knowledge atoms, which can be formalized as a joint measure of conditional probability and semantic implication. Calculating the prerequisite probabilities between atoms represents the relationship between knowledge atoms in terms of prerequisites and successors. If the disciplinary purity of atom A leans towards basic disciplines and its semantics are a necessary component of atom B, then a vertical support line from basic to applied is established. Prerequisite probabilities can be comprehensively estimated based on the prerequisite relationships, co-occurrence order, and semantic inclusion relationships of knowledge atoms in the teaching corpus; its essence is modeling the conditional probability of learning atom B depending on atom A.

[0073] At the implementation level, the preceding probability can be expressed by the following formula:

[0074]

[0075] in, This indicates the number of times that atomic knowledge point A and atomic knowledge point B appear simultaneously within the same knowledge context, and atomic knowledge point A appears before atomic knowledge point B. The frequency of occurrence of atomic knowledge point B, This represents the vertical dependency relationship between atomic knowledge point B and atomic knowledge point A.

[0076] Preferably, the prerequisite probability can be calculated directly in the preliminary prerequisite dependency unit of the atomic knowledge point, which avoids calculating the prerequisite probability of all atomic knowledge points and greatly reduces the computational complexity.

[0077] As one implementation method, in a practical system, this probability can be approximated by analyzing the textbook structure, the syllabus dependencies, and the knowledge graph path statistics.

[0078] If the preceding probability is greater than or equal to the preceding threshold, then the two are determined to have a vertical dependency relationship; if the preceding probability is less than the preceding threshold, then the two are determined not to have a vertical dependency relationship.

[0079] Horizontal Mapping: This method uses cosine similarity to calculate the distance between atomic knowledge points from different courses, representing the parallel relationship between these points. When Distance(AKU) i AKU j When the similarity threshold is ≥ ε, it is determined to be interdisciplinary homogeneous knowledge, and an equivalence mapping line is established. i and AKU j Let be the semantic vectors of the i-th and j-th atomic knowledge points, respectively, and Distance be the cosine similarity.

[0080] Specifically, all existing courses are categorized by subject and deconstructed into atomic knowledge points, resulting in multiple subject databases. Semantic vectors are calculated for each atomic knowledge point in the original interdisciplinary courses and for atomic knowledge points in different subject databases. Cosine similarity is then calculated between these semantic vectors. If the cosine similarity is greater than or equal to a similarity threshold, the two are considered to have a horizontal homogeneous association; otherwise, they are considered not to have a horizontal homogeneous association. A pre-trained model (such as BERT) is used to extract the semantic embeddings of the atomic knowledge points as semantic vectors. Finally, an association matrix is ​​constructed for each atomic knowledge point in the original interdisciplinary courses based on vertical dependency and horizontal homogeneous associations. An association network is then built based on all association matrices. The association matrix includes the atomic knowledge point, its corresponding location information, and the atomic knowledge points that have vertical dependency and horizontal homogeneous associations with that atomic knowledge point.

[0081] The location information includes subject, difficulty, activity level, and number of interdisciplinary associations; the difficulty is manually assigned, the number of interdisciplinary associations is the total number of atomic knowledge points that are horizontally homogeneously associated with the atomic knowledge point, the activity level is the number of interdisciplinary associations divided by the total number of associations, and the total number of associations is the total number of atomic knowledge points that are vertically dependent and horizontally homogeneously associated with the atomic knowledge point.

[0082] Preferably, horizontal homogeneous associations also include internal association links.

[0083] Example 3

[0084] Based on the above embodiments, this invention proposes an application example of a knowledge atom construction method for interdisciplinary courses, including:

[0085] Preferably, a preliminary scan can be performed before initially dividing the original interdisciplinary courses into their own units. The original interdisciplinary courses selected are the course packages of "Intelligent Planning and Collaboration" (including an outline, 50 teaching PPT slides, and 30,000 words of textbook text), which are then divided into original interdisciplinary course units.

[0086] Preferably, as one implementation method, the original interdisciplinary course unit U of "Path Planning Based on A* Algorithm" in "Intelligent Planning and Collaboration" is semantically vectorized using the Bidirectional Encoding Representation Model (BERT) to obtain a semantic vector. Next, the system iterates through all existing course units, calculating the cosine similarity of each unit's semantic vector with U as an initial probability link value. The calculation shows that the semantic vector cosine similarity between the "Graph Search Strategies" unit in the existing course "Graph Theory" and U is 0.72, the highest among all existing course units. Based on this, the "graph search strategy" is determined to be the initial prerequisite dependency unit of U, and the probability value of 0.72 is used as the initialization basis for association inheritance in the subsequent recursive destructuring process.

[0087] Then, a recursive deconstruction algorithm based on knowledge entropy is executed, which ensures that the "line" is refined in sync with the "point" during the decomposition process.

[0088] Specifically, in the first stage, primary deconstruction and associative decomposition: the semantic information entropy H(U) of U is calculated as 0.88. A preset threshold t is set. au1 = 0.5. Since 0.88 > 0.5, the algorithm performs the first split, dividing the unit into U1 (search logic) and U2 (heuristic evaluation function).

[0089] Preferably, based on the semantic vector similarity between U1, U2 and the "graph theory" knowledge cluster, preliminary link relationships are assigned to units with high similarity, that is, the link relationship between U and "graph theory" is assigned to U1. Through text coupling analysis, internal association links are automatically established for U1 and U2 (i.e., the output of U2 is the input parameter of U1), which facilitates the establishment of internal associations and prevents logical fragmentation caused by deconstruction.

[0090] The second stage, secondary recursion and atomized decision-making: further deconstruction of U2 (heuristic evaluation function):

[0091] Calculating the subject distribution vector reveals that it covers "Mathematics (Geometry)" in 55% and "Computer Science (Algorithms)" in 45%, with principal component percentages... = 0.55.

[0092] Set purity termination threshold t au2 = 0.85. Since 0.55 < 0.85, continue to split into sub-units U. 21 (Manhattan distance) and U 22 (Comparison of Euclidean distances).

[0093] Final atomization result: for U 21 The Manhattan distance has been recalculated, and the weight of the "mathematics" dimension in its subject vector distribution has increased to 96%, with the principal component accounting for a larger proportion. =0.96>0.85. Decision: Stop splitting.

[0094] Preferably, the final result is marked as an atomic knowledge point AKU. m It is assigned a globally unique identifier, GUID-701, for easy identification and querying.

[0095] Calculation of Atoms Knowledge Points AKU m The prior probability between each remaining atomic knowledge point and the atomic knowledge point AKU is discovered. mThere is a vertical dependency relationship between the atomic knowledge points and the "1-norm" in Linear Algebra. A vertical dependency chain of [Linear Algebra: 1-norm] → [Intelligent Planning: Manhattan Distance] is automatically established. This means that if a student has a poor foundation in linear algebra, the system will warn of potential learning risks related to path planning.

[0096] A search of the subject database revealed a single knowledge point, AKU, in the "Pixel Neighborhood Calculation" chapter of the "Digital Image Processing" course. img (Chessboard distance measurement).

[0097] Calculate similarity: Sim(V) m ,) = { V m ·V img} / {| V m | | V img |} = 0.94

[0098] Where Sim is the cosine similarity, and V m For atomic knowledge points AKU m semantic vector, V img AKU, a knowledge point about atoms img Semantic vector, || is the magnitude, and · is the inner product.

[0099] The similarity threshold is set to 0.90. Since 0.94 > 0.90, an equivalent mapping line is automatically drawn between the knowledge graphs of the two disciplines.

[0100] Based on vertical dependency and horizontal homogeneous association, construct the association matrix of each atomic knowledge point of the original interdisciplinary course, and construct the association network based on all association matrices.

[0101] Atomic Knowledge Points AKU m The position of the correlation matrix in the 3D network is represented by the following formula:

[0102] Location: [Subject: Robotics, Difficulty: 3, Activity: 0.85, Number of interdisciplinary connections: 12].

[0103] Example 4

[0104] Based on the above embodiments, such as Figure 2 As shown, this invention proposes a knowledge atom construction system for interdisciplinary courses, comprising:

[0105] The preliminary division module is used to initially divide the original interdisciplinary courses according to their own units, thus obtaining the original interdisciplinary course units;

[0106] The atomization module is used to calculate the semantic information entropy of the original interdisciplinary course unit and determine whether to deconstruct the original interdisciplinary course unit based on the semantic information entropy. If it is necessary to deconstruct the original interdisciplinary course unit, a clustering algorithm is used to divide the original interdisciplinary course unit into multiple sub-course candidate sets.

[0107] The judgment module is used to calculate the principal component proportion of the subject distribution vector of each sub-course candidate set. Based on the principal component proportion, it determines whether the sub-course candidate set meets the atomization standard. If it meets the atomization standard, the sub-course candidate set is used as an atomic knowledge point. If it does not meet the atomization standard, the atomization module and the judgment module are re-executed.

[0108] The association matrix module is used to calculate the vertical dependency associations and horizontal homogeneous associations of each atomic knowledge point, construct the association matrix of each atomic knowledge point based on the vertical dependency associations and horizontal homogeneous associations, and construct an association network from all association matrices.

[0109] It should be noted that the knowledge atom construction system for interdisciplinary courses provided in this embodiment of the invention is to implement the above-mentioned knowledge atom construction method for interdisciplinary courses. Its specific functions can be referred to in the above-mentioned method embodiments, and will not be repeated here.

[0110] Example 5

[0111] Based on the above embodiments, the present invention proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements a knowledge atom construction method for interdisciplinary courses as described in Embodiment 1 above.

[0112] Example 6

[0113] Based on the above embodiments, the present invention proposes a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements a knowledge atom construction method for interdisciplinary courses as described in Embodiment 1 above.

[0114] In summary, this invention achieves standardization and refinement of knowledge granularity: by employing a recursive deconstruction algorithm based on knowledge entropy, it solves the problem of excessively coarse granularity in chapter knowledge in traditional teaching resources. This invention can automatically decompose complex course content into indivisible "knowledge atoms," providing a standardized data foundation for the comparison, reuse, and precise mapping of interdisciplinary content. This invention constructs a dynamic and three-dimensional interdisciplinary logical topology (association matrix and association network): unlike traditional unidirectional tree-like outlines, this invention automatically constructs a network of vertical support (basics and applications) and horizontal mapping (homogeneous knowledge) through the association characteristics in multidimensional feature vectors. This structure can reveal the underlying logical connections between disciplines, solving the pain point of difficulty in quantifying and identifying interdisciplinary intersections. This invention possesses high correlation inheritance and logical rigor: by introducing an "association inheritance mechanism" during the deconstruction process, it ensures that the original logical chain of knowledge points is not lost when evolving from coarse to fine granularity, achieving complete tracing of knowledge context and effectively avoiding the generation of knowledge silos. This invention significantly improves the automation level of interdisciplinary system construction: it greatly reduces the cost of manually compiling interdisciplinary outlines and can automatically adjust the weight and links of knowledge atoms according to the corpus update, adapting to the needs of the rapid development of cutting-edge interdisciplinary directions.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing knowledge atoms for interdisciplinary courses, characterized in that, include: Step 1: Initially divide the original interdisciplinary courses into their own units to obtain the original interdisciplinary course units; Step 2: Calculate the semantic information entropy of the original interdisciplinary course unit, and determine whether to deconstruct the original interdisciplinary course unit based on the semantic information entropy; if it is necessary to deconstruct the original interdisciplinary course unit, use a clustering algorithm to divide the original interdisciplinary course unit into multiple sub-course candidate sets. Step 3: Calculate the principal component proportion of the subject distribution vector of each sub-course candidate set based on all existing courses. Compare the principal component proportion with the purity termination threshold. If the principal component proportion is greater than or equal to the purity termination threshold, the sub-course candidate set is used as an atomic knowledge point. If the principal component proportion is less than the purity termination threshold, repeat steps 2 and 3. Step 4: Calculate the vertical dependency and horizontal homogeneity associations for each atomic knowledge point. Construct the association matrix for each atomic knowledge point based on the vertical dependency and horizontal homogeneity associations, and build an association network from all association matrices.

2. The method for constructing knowledge atoms for interdisciplinary courses according to claim 1, characterized in that, The specific steps of determining whether to deconstruct the original interdisciplinary course unit based on semantic information entropy include: The semantic information entropy of the original interdisciplinary course unit is compared with a preset threshold. If the semantic information entropy is greater than the preset threshold, the original interdisciplinary course unit is deconstructed. If the semantic information entropy is less than or equal to the preset threshold, no deconstruction is performed, and the original interdisciplinary course unit is treated as an atomic knowledge point.

3. The method for constructing knowledge atoms for interdisciplinary courses according to claim 1, characterized in that, The calculation of the principal component proportion of the subject distribution vector for each sub-course candidate set specifically includes: All existing courses are divided into different subject corpora according to their disciplines; the subject corpora include subject units divided according to the course itself. The subject distribution vector of the sub-course candidate set is obtained based on different subject corpora and probability models; the subject distribution vector includes the weight of the sub-course candidate set in different subject corpora; The principal component percentage is obtained by dividing the largest weight in the subject distribution vector by the sum of all weights.

4. The method for constructing knowledge atoms for interdisciplinary courses according to claim 1, characterized in that, The vertical dependency association is obtained through the following steps: Calculate the prior probability between each atomic knowledge point and each remaining atomic knowledge point. If the prior probability is greater than or equal to the prior threshold, it is determined that there is a vertical dependency relationship between the two. If the prior probability is less than the prior threshold, it is determined that there is no vertical dependency relationship between the two.

5. The method for constructing knowledge atoms for interdisciplinary courses according to claim 4, characterized in that, The preceding probability is expressed by the following formula: in, This indicates the number of times that atomic knowledge point A and atomic knowledge point B appear simultaneously within the same knowledge context, and atomic knowledge point A appears before atomic knowledge point B. The frequency of occurrence of atomic knowledge point B, This represents the vertical dependency relationship between atomic knowledge point B and atomic knowledge point A.

6. The method for constructing knowledge atoms for interdisciplinary courses according to claim 1, characterized in that, The horizontal homogeneous association is obtained through the following steps: All existing courses are categorized by subject and deconstructed into atomic knowledge points, resulting in multiple subject databases based on these atomic knowledge points. Calculate the semantic vectors of each atomic knowledge point in the original interdisciplinary course and atomic knowledge points in different subject libraries, and calculate the cosine similarity between the semantic vectors. If the cosine similarity is greater than or equal to the similarity threshold, it is determined that the two have a horizontal homogeneous association. If the cosine similarity is less than the similarity threshold, it is determined that the two have no horizontal homogeneous association.

7. The method for constructing knowledge atoms for interdisciplinary courses according to claim 1, characterized in that, The process of constructing an association matrix for each atomic knowledge point based on vertical dependency and horizontal homogeneity, and then constructing an association network from all association matrices, specifically includes: Based on vertical dependency and horizontal homogeneity, an association matrix is ​​constructed for each atomic knowledge point of the original interdisciplinary course. An association network is then constructed based on all association matrices. The association matrix includes the atomic knowledge point, the location information corresponding to the atomic knowledge point, and the atomic knowledge points that are vertically dependent and horizontally homogeneously associated with that atomic knowledge point. The location information includes subject, difficulty, activity level, and number of interdisciplinary associations; the difficulty is manually assigned, the number of interdisciplinary associations is the total number of atomic knowledge points that are horizontally homogeneously associated with the atomic knowledge point, the activity level is the number of interdisciplinary associations divided by the total number of associations, and the total number of associations is the total number of atomic knowledge points that are vertically dependent and horizontally homogeneously associated with the atomic knowledge point.

8. A knowledge atom construction system for interdisciplinary courses, characterized in that, include: The preliminary division module is used to initially divide the original interdisciplinary courses according to their own units, thus obtaining the original interdisciplinary course units; The atomization module is used to calculate the semantic information entropy of the original interdisciplinary course unit and determine whether to deconstruct the original interdisciplinary course unit based on the semantic information entropy. If it is necessary to deconstruct the original interdisciplinary course unit, a clustering algorithm is used to divide the original interdisciplinary course unit into multiple sub-course candidate sets. The judgment module is used to calculate the principal component proportion of the subject distribution vector of each sub-course candidate set. Based on the principal component proportion, it determines whether the sub-course candidate set meets the atomization standard. If it meets the atomization standard, the sub-course candidate set is used as an atomic knowledge point. If it does not meet the atomization standard, the atomization module and the judgment module are re-executed. The association matrix module is used to calculate the vertical dependency associations and horizontal homogeneous associations of each atomic knowledge point, construct the association matrix of each atomic knowledge point based on the vertical dependency associations and horizontal homogeneous associations, and construct an association network from all association matrices.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images