A textbook knowledge point reorganization method based on problem anchor points and a knowledge graph

Abstract

The present application relates to the technical field of teaching material knowledge arrangement, and specifically discloses a teaching material knowledge point reorganization method based on problem anchor points and knowledge graphs, comprising: obtaining teaching materials, and constructing a basic knowledge point graph; constructing problem anchor points; for each problem anchor point, selecting knowledge points related to the problem anchor point from the basic knowledge point graph, generating a candidate knowledge group, and determining the relationship between the problem anchor points; dividing the candidate knowledge group according to the relationship between the problem anchor points, generating a candidate division scheme; selecting the optimal candidate division scheme, determining the final teaching knowledge unit and its teaching order suggestion; generating a reality import question, and obtaining a teaching knowledge unit result organized around the problem anchor points. The present application organizes the scattered or non-continuous knowledge points in the original teaching materials into teaching units suitable for teaching based on the teaching material knowledge structure and the teaching problem as the organization clue, and provides structured support for teaching material rewriting, teaching plan generation and classroom teaching design.

Description

Technical Field

[0001] This invention relates to the field of textbook knowledge arrangement technology, and in particular to a method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs. Background Technology

[0002] Traditional textbooks are typically organized according to the logic of knowledge, using chapters, sections, and subsections as organizational units, and unfolding in the order of concepts, definitions, principles, formulas, examples, and exercises. This approach is conducive to the systematic presentation of subject knowledge, but its organizational center remains the knowledge points themselves. In contrast, problem-based learning, inquiry-based learning, and project-based learning emphasize using problems, tasks, or real-world scenarios as learning starting points, guiding students to analyze, explore, acquire, and apply relevant knowledge around the problem. For these teaching scenarios, simply using a traditional table of contents structure is insufficient to fully support the teaching process of introducing knowledge from problems, breaking down and organizing knowledge from problems, and returning to problem-solving from knowledge application.

[0003] With the development of generative artificial intelligence, large language models, and educational knowledge graph technologies, teaching resources such as textbooks, courseware, lecture notes, lesson plans, exercises, and case studies have the technological foundation to be automatically parsed, extracted, and reorganized. The intelligent processing of educational resources is gradually shifting from simple resource retrieval, content recommendation, and automatic question answering to the automated planning of course structure, teaching paths, and the organization of teaching content.

[0004] In existing technologies, there are mainly four construction directions: (1) The direction of constructing textbook knowledge graphs and educational resource knowledge graphs. For example, the Chinese invention patent with publication number CN110287327A takes the textbook catalog as the object and constructs a knowledge expression hypergraph, a knowledge expression chain graph and a path-adaptive knowledge graph to support the generation of personalized knowledge paths.

[0005] (2) The direction of constructing course knowledge system, course map, problem map and ability map. For example, the Chinese invention patent with publication number CN117196022A obtains knowledge map and course summary from teaching platform, determines course teaching logic, extracts course map and integrates teaching resources to form course knowledge system, and constructs course problem map and ability map.

[0006] (3) Knowledge graph-based lesson plan generation and question chain lesson plan generation. For example, Chinese invention patent with publication number CN121919381A converts unstructured lesson plans into question chain lesson plans, constructs a logical structure diagram of the question chain, and determines the question semantic anchor points of the logical nodes.

[0007] (4) Learning path recommendation and course path planning direction. For example, the Chinese invention patent with publication number CN106503172A creates a user's personal knowledge graph and recommends learning paths to the user based on the user's mastery of knowledge points.

[0008] However, the existing technologies mentioned above still have shortcomings: (1) Existing textbook knowledge graph and curriculum modeling technologies focus on the structured representation of knowledge objects, teaching resources and curriculum relationships. Their organizational logic is still biased towards the knowledge structure itself. They do not provide clear support for the process of transforming the structure representation results into teachable teaching units. It is difficult to complete the determination of teaching unit granularity, knowledge point merging and splitting, cross-chapter knowledge point combination and unit connection relationship determination; (2) Learning path recommendation and resource recommendation technologies are mainly aimed at individual learners. They are difficult to meet the needs of teachers to organize knowledge points into units from the perspective of the textbook as a whole; (3) Existing lesson plan generation and problem chain lesson plan generation technologies focus on the expression of teaching text or teaching process. They generate content based on the teaching scope, class hour requirements or existing teaching structure. They do not provide sufficient support for the unit-level structural constraints used to limit the scope of knowledge points, teaching order and textbook basis before generation. In the scenario of textbook rewriting, they directly enter the content generation stage, which is prone to problems such as incomplete knowledge point coverage, unclear connection between the beginning and end, unstable unit boundaries and difficulty in locating the generation basis. Summary of the Invention

[0009] This invention aims to address the problem that existing technologies cannot fully support the need for reorganizing textbook knowledge points during the transformation of traditional textbooks into problem-driven teaching resources. To this end, this invention provides a method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs. Building upon the textbook's knowledge structure, and using teaching questions as organizational clues, it organizes scattered or discontinuous knowledge points in the original textbook into suitable teaching units. It also clarifies the scope of knowledge points, teaching order, and textbook basis for each unit, providing structured support for textbook rewriting, lesson plan generation, and classroom teaching design.

[0010] This invention provides a method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs. The technical solution adopted is as follows: It includes: S1: Obtain structured teaching materials, analyze the teaching materials to obtain teaching material units, and construct a basic knowledge point map based on the teaching material units; S2: Construct problem anchors based on the textbook; and establish an initial cross-level mapping relationship between the problem anchors and the knowledge points of the textbook units and basic knowledge point maps; S3: For each question anchor point, select knowledge points related to that question anchor point from the basic knowledge point graph to generate candidate knowledge groups; determine the relationship between question anchor points based on the candidate knowledge groups and the initial cross-layer mapping relationship; among which, the relationship between question anchor points includes decomposition relationship, merging relationship and sequential relationship; S4: Determine the number of teaching knowledge units; for each teaching knowledge unit, divide the candidate knowledge groups according to the relationship between the problem anchors and generate candidate division schemes; In S4, candidate knowledge groups are divided. When splitting candidate knowledge groups, larger problem anchors with executable splitting combinations are split into multiple smaller problem anchors. Among them, for larger and smaller problem anchors with splitting relationships, if a larger problem anchor corresponds to multiple smaller problem anchors, and the overall coverage ratio of the multiple smaller problem anchors to the larger problem anchor is greater than or equal to a preset threshold, then the multiple smaller problem anchors constitute an executable splitting combination. S5: Select the optimal candidate partitioning scheme based on the coverage rate of core knowledge points, the overlap rate of core knowledge points, the order violation rate, and the load variance, and determine the final teaching knowledge units and their teaching order suggestions; S6: Generate realistic introductory questions for each final teaching knowledge unit, and combine the final teaching knowledge unit with its teaching sequence suggestions to obtain the teaching knowledge unit results organized around the question anchors.

[0011] Furthermore, in S2, candidate question anchors are generated from the textbook title, learning objectives, examples, cases, textbook introduction, and textual descriptions that reflect the purpose or applicable context of the knowledge points. Question anchors are then selected from the candidate question anchors based on the textbook units and basic knowledge point map. The selection criteria are as follows: if a candidate question anchor is only the name of the knowledge point itself, or lacks the source textbook unit, the initial triggering knowledge point, or the evidence text, it will not be used as a question anchor.

[0012] Furthermore, in S3, the relevance of the question anchor and the knowledge point is determined based on the relevance of evidence, the relevance of the graph relationship, the semantic relevance, and the relevance of the source location. Among them, the relevance of evidence is used to determine whether the knowledge point and the question anchor point have direct textbook evidence; The graph relational degree is used to represent the strength of the relationship between the initial triggering knowledge points of knowledge points and problem anchors in the basic knowledge point graph; Semantic relevance is used to represent the semantic similarity between the name, description, and source evidence text of a knowledge point and the name and evidence text of the question anchor. Source location relevance is used to indicate how close the source location of knowledge points and question anchors is in the textbook.

[0013] Furthermore, in S3, Comparing two problem anchors, if their content is similar, one problem anchor has fewer restrictions and can cover more situations, so it is regarded as a larger problem anchor; the other problem anchor adds specific conditions, specific components or specific processing steps, so it is regarded as a smaller problem anchor. A decomposition relationship is established from the larger problem anchor to the smaller problem anchor when the proportion of core knowledge points in the smaller problem anchor that is covered by the larger problem anchor is greater than or equal to a preset coverage threshold, and the following conditions are met simultaneously: Condition 1: Judging from the problem description, the coverage of the larger problem anchor point is greater than that of the smaller problem anchor point; Condition 2: The larger problem anchor point contains core knowledge points that are not covered by the smaller problem anchor point; Condition 3: Evidence that the two materials share the same title, common learning objectives, the same example, the same case, the same introduction, or are from similar sources; For two problem anchors that do not have a decomposition relationship, determine whether the conditions for merging are met; If there is no obvious scope inclusion relationship between two question anchors, but their coverage is similar and they focus on different parts of the same teaching content, then the two question anchors have a merging relationship. The order of priority is determined based on the prerequisite relationships of knowledge points and the content of the textbook.

[0014] Furthermore, overall coverage ratio The calculation formula is: in, Indicates the problem anchor point The set of core knowledge points corresponding to the candidate knowledge group. This serves as an anchor point for larger issues. This represents all the core knowledge points contained within all the smaller problem anchors. Indicates and The combination of smaller problem anchor points that establish a decomposition relationship. This represents the i-th smaller problem anchor point.

[0015] Furthermore, in S5, the global score of the candidate partitioning scheme is calculated based on the core knowledge point coverage, core knowledge point overlap, order violation rate and load variance, and the candidate partitioning scheme with the highest global score is selected as the final partitioning scheme. Each candidate teaching knowledge unit in the final division scheme is the final teaching knowledge unit; Based on the sequential relationship between problem anchors and the prerequisite relationship between knowledge points, the final teaching knowledge units are ordered to obtain a suggested teaching order for the final teaching knowledge units.

[0016] Furthermore, the formula for calculating the global score is as follows: in, Representing candidate partitioning schemes The overall score This indicates the coverage rate of core knowledge points. This indicates the overlap rate of core knowledge points. Indicates the order violation rate. Indicates the load variance; The preset weights represent the coverage of core knowledge points. The preset weights representing the overlap rate of core knowledge points. The preset weights representing the order violation rate Preset weights representing the load variance; Core knowledge point coverage rate is used to indicate the degree to which candidate partitioning schemes cover the core knowledge points of the entire book. The core knowledge point overlap rate is used to represent the degree of overlap in the core knowledge points among different candidate teaching knowledge units. The order violation rate is used to indicate whether a candidate partitioning scheme violates the prerequisite relationship of knowledge points or the order relationship of problem anchor points. Load variance is used to indicate whether the teaching load is balanced among the candidate teaching knowledge units in the candidate partitioning scheme.

[0017] Furthermore, the teaching load of a candidate knowledge unit is calculated based on the teaching load of its corresponding candidate knowledge group; The calculation process for the teaching load of candidate knowledge groups is as follows: determine the core knowledge points of the candidate knowledge group, calculate the teaching load of all core knowledge points, and sum them to obtain the teaching load of the candidate knowledge group. Among them, if a knowledge point is required by the learning objective, or if its structural salience or material support is greater than or equal to the corresponding threshold, then the knowledge point is a core knowledge point; the teaching load is the weighted sum of the structural salience, knowledge type complexity and material support of the knowledge point.

[0018] Furthermore, in S1, knowledge points, knowledge layer relationships, and source evidence are extracted from the textbook units to construct a basic knowledge point map; The basic knowledge point map includes knowledge point nodes, textbook unit nodes, knowledge layer relationships between knowledge points, and the source relationships between knowledge points and textbook units; Knowledge layer relationships include prerequisite relationships, composition relationships, co-occurrence relationships, and similarity relationships.

[0019] The above-described one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects: (1) This invention constructs a problem anchor layer on top of the basic knowledge point map, and uses the problem anchor as the organizational center of the teaching knowledge unit, so that the textbook content can be transformed from the original chapter and knowledge point organization method to a unit organization method centered around the teaching central problem. Based on the cross-layer mapping relationship between problem anchors, knowledge points, and textbook units, knowledge points scattered in different chapters or sections in the original textbook can be organized around the same teaching central problem, thereby reducing the restriction of the original textbook's continuous catalog structure on the division of teaching units. The resulting teaching knowledge unit can more clearly reflect problem guidance, knowledge support, and unit boundaries, providing a more suitable knowledge organization foundation for problem-driven teaching, textbook rewriting, and lesson plan design.

[0020] (2) During the processing, this invention records the source evidence of knowledge points, the initial triggering relationship of problem anchors, the textbook evidence relationship, and the knowledge support relationship. The final output also includes the source path of knowledge points and the mapping path from problem anchors to knowledge points. Through the above paths, it is clear which chapter, section, example, case, learning objective, or paragraph in the original reference materials the knowledge points in each teaching knowledge unit come from. It is also clear which knowledge points lead to the problem anchors, which textbook contents provide evidence, and which knowledge points jointly support them. This helps with subsequent textbook rewriting, lesson plan generation, and teacher review, and reduces the risk of unclear evidence and difficult-to-verify content in the subsequently generated content.

[0021] (3) This invention preserves the decomposition, merging, and sequential relationships between problem anchors, enabling candidate knowledge groups to be adjusted not only based on the number of knowledge points, workload, or semantic similarity, but also by considering the organizational relationships between problem anchors. This invention can more reasonably determine which candidate knowledge groups are suitable for further subdivision, which candidate knowledge groups are suitable for merging into the same teaching knowledge unit, and how to arrange the sequence between different teaching knowledge units; it helps to reduce situations such as teaching knowledge units being too large or too small, central issues being too scattered, and adjacent units not being smoothly connected, thereby improving the stability and interpretability of the teaching knowledge unit division results.

[0022] (4) Before determining the candidate partitioning scheme as the final teaching knowledge unit, this invention comprehensively evaluates the candidate partitioning scheme from aspects such as coverage of core knowledge points, overlap of core knowledge points, teaching order, and unit load; it can screen and compare among multiple candidate partitioning schemes, avoiding directly determining the final result based solely on the relevance of question anchor points or the number of candidate knowledge groups. This invention can reduce the risks of core knowledge points not being included in the teaching knowledge unit, the same core knowledge point being repeatedly assigned to multiple teaching knowledge units, the teaching order violating the prerequisite relationship or the order of question anchor points, and the content volume of different teaching knowledge units being too different, thereby improving the completeness, boundary clarity, order rationality, and load balance of the final partitioning result.

[0023] (5) This invention supports teachers in specifying the number of teaching units and using it as a priority target to generate and evaluate candidate division schemes. When teachers do not specify the number of teaching units, this invention can also recommend a suitable number of teaching units and their corresponding division schemes based on the knowledge point load of the textbook and the evaluation results of the candidate division schemes. This invention retains the teacher's control over the teaching plan while reducing problems such as insufficient coverage of core knowledge points, unbalanced unit load, or unreasonable teaching sequence caused by unreasonable specified numbers.

[0024] (6) The final output of this invention is a structured teaching knowledge unit that has been divided and ordered. The teaching knowledge unit structure clarifies the organizational center, knowledge point scope, core knowledge points, teaching order suggestions, and corresponding real-world introductory questions for each teaching knowledge unit. It can serve as a unit framework for subsequent textbook rewriting, lesson plan generation, classroom activity design, or learning task design. Based on the above structured results, subsequent processing can be carried out with teaching knowledge units as the basic objects. The textbook content can be organized, classroom activities can be designed, and learning tasks can be generated directly around the established units. This reduces the workload of reorganizing the entire textbook, determining the knowledge scope, and arranging the teaching order, and improves the stability, coherence, and controllability of subsequent teaching resource processing.

[0025] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0027] Figure 1 This is a flowchart of the method provided by the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely 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. The following embodiments are used to illustrate this invention but should not be used to limit the scope of this invention.

[0029] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0030] The following is combined with Figure 1 The present invention will be further described in detail below, which describes a method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs: In this embodiment, as Figure 1 As shown, a method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs is provided. This method is suitable for reorganizing traditional textbooks or structured teaching materials with a catalog structure into teaching knowledge units that can support problem-driven teaching. The method includes the following steps: S1: Acquiring the Basic Knowledge Structure of the Textbook. This involves acquiring a structured textbook, analyzing it to obtain textbook units, and constructing a basic knowledge point map based on these units.

[0031] The textbook foundation knowledge structure acquisition stage is used to obtain structured textbooks (teaching materials) and extract the foundational knowledge structure required for subsequent processing. This stage mainly generates a set of textbook units, a set of knowledge points, a set of knowledge layer relationships, a set of source evidence, and a foundational knowledge point graph. The foundational knowledge point graph provides the underlying knowledge basis for subsequent processes.

[0032] This stage does not focus on proposing new methods for constructing textbook knowledge graphs. The specific construction method for the basic knowledge point graph is not limited; it can be achieved through rule extraction, large language model extraction, importing existing knowledge graphs, importing existing structured knowledge data, or a combination of the above methods. This stage specifically includes the following steps: S1.1: Obtain the structured teaching materials and parse them to obtain the teaching material units.

[0033] Structured teaching materials can be traditional textbooks with a table of contents, or teaching materials with a structure such as headings, main text, examples, exercises, cases, or learning objectives.

[0034] This method parses reference materials into several textbook units based on their heading level, content type, and source location. Each textbook unit is a traceable content unit extracted from the original reference material, and each unit has a corresponding unit identifier, unit type, and source location. The source location can be a chapter number, page number, paragraph number, or other location identifier that can pinpoint the original content. Textbook units include titles, body paragraphs, definitions, formulas, examples, case studies, learning objectives, or summaries.

[0035] S1.2: Extract knowledge points, knowledge layer relationships, and source evidence from the textbook units.

[0036] This method extracts knowledge points, knowledge layer relationships, and source evidence from textbook units.

[0037] A knowledge point is a unit of knowledge in a textbook that can be taught, organized, and retrieved. When extracting knowledge points, this method records at least the knowledge point name, type, description, and source textbook unit.

[0038] Knowledge layer relationships refer to the relationships between knowledge points within a basic knowledge point map, which can include prerequisite relationships, composition relationships, co-occurrence relationships, and similarity relationships. Among them, the prerequisite relationship indicates that learning one knowledge point usually requires mastering another knowledge point first; the composition relationship indicates the composition or inclusion relationship between knowledge points; the co-occurrence relationship indicates that two knowledge points appear or are used together in the same textbook unit or teaching material; and the similarity relationship indicates that two knowledge points are similar in semantics, function, or teaching purpose.

[0039] Source evidence is used to explain the location and manner in which knowledge points appear in the original reference materials. This can include the source of the definition, its use in examples, learning objectives, and textual descriptions reflecting the purpose or applicable context of the knowledge point. This source evidence is used for subsequent output of knowledge point source paths and verification of teaching knowledge unit results.

[0040] S1.3: Construct a basic knowledge point map based on knowledge points, knowledge layer relationships, and source evidence.

[0041] This method standardizes knowledge points, knowledge layer relationships, and source evidence to form a basic knowledge point map. The standardization process includes knowledge point standardization, relationship standardization, and source evidence standardization.

[0042] Knowledge point standardization is used to merge synonymous or near-synonymous knowledge points and to unify knowledge point names and types; relationship standardization is used to confirm the type of knowledge layer relationship, such as avoiding misjudging ordinary co-occurrence relationships as prerequisite relationships or composition relationships; source evidence standardization is used to unify the recording of source relationships between knowledge points and textbook units.

[0043] After standardization, a basic knowledge point map is formed. The basic knowledge point map includes knowledge point nodes, textbook unit nodes, knowledge layer relationships between knowledge points, and source relationships between knowledge points and textbook units.

[0044] S2: Problem Anchor Point and Cross-Level Mapping Establishment Stage. Construct problem anchor points based on the textbook; and establish initial cross-level mapping relationships between problem anchor points and knowledge points in textbook units and basic knowledge point maps.

[0045] The problem anchor and cross-layer mapping establishment phase is used to construct a problem anchor layer on top of the basic knowledge point map and establish the initial cross-layer mapping relationship between problem anchors and the underlying knowledge structure. This phase mainly generates a set of problem anchors, initial trigger relationships, and textbook evidence relationships, providing a basis for subsequently generating candidate knowledge groups around the problem anchors.

[0046] This method sets question anchors at a separate question anchor layer, rather than at the underlying basic knowledge point graph layer. Question anchors are not considered knowledge point nodes in the basic knowledge point graph; they are only connected to knowledge points and textbook units through cross-layer mapping relationships. The specific process of this stage is as follows: This method generates candidate question anchors from the textbook's title, learning objectives, examples, case studies, textbook introduction, and textual descriptions that demonstrate the purpose or applicable context of the knowledge points. Then, based on the textbook units and a map of fundamental knowledge points, the candidate question anchors are selected to obtain the final question anchors.

[0047] A problem anchor should express a problem, task, explanation objective, comparison objective, or application objective, and not just the name of a knowledge point. A problem anchor should also have the potential to organize multiple knowledge points and include a source textbook unit, an initial triggering knowledge point, and supporting text. The source textbook unit indicates which part of the original reference material the problem anchor originates from; the initial triggering knowledge point indicates which knowledge points initially led to the problem anchor; and the supporting text indicates where in the original textbook the problem anchor originated.

[0048] For example, in the previous stage, the knowledge point "for loop" was extracted from a certain textbook unit, and its source evidence was recorded as the textbook statement "for loop is usually used to handle tasks with a known number of repetitions." In this stage, based on this usage description, the question anchor point "how to express a task with a fixed number of repetitions" can be generated, and "for loop" can be identified as the initial trigger knowledge point for this question anchor point, with the aforementioned textbook statement identified as the evidence text.

[0049] Screening criteria: If a candidate question anchor is only the name of the knowledge point itself, or lacks the source textbook unit, the initial triggering knowledge point, or the evidence text, it will not be considered a valid question anchor for the subsequent process.

[0050] After screening, multiple question anchors were obtained, and a question anchor set was constructed. Each valid question anchor in the question anchor set includes at least the question anchor name, the source textbook unit, the initial triggering knowledge point, and the evidence text.

[0051] After generating valid question anchors, this method establishes an initial cross-layer mapping relationship between the question anchors and the underlying knowledge structure. This initial cross-layer mapping relationship includes initial triggering relationships and textbook evidence relationships.

[0052] Initial trigger relationships are used to indicate which initial knowledge points trigger the problem anchor. For example, the problem anchor "how to express a task that repeats a fixed number of times" can be triggered by the knowledge point "a loop statement with a fixed number of times".

[0053] Textbook evidence relationships are used to indicate which textbook units provide evidence for the problem anchor. For example, the textbook's explanation that "fixed-number loop statements are usually used to handle tasks with a known number of repetitions" can serve as supporting evidence for the problem anchor "how to express a fixed-number repetition task".

[0054] This method connects the problem anchor layer with the underlying knowledge point graph layer through an initial cross-layer mapping relationship, enabling subsequent processes to reorganize knowledge points around the problem anchors, avoiding simply following the original textbook's table of contents structure. The problem anchor set, initial trigger relationships, and textbook evidence relationships obtained in this stage can serve as the basis for generating candidate knowledge groups in the next stage.

[0055] S3: Candidate Knowledge Group Generation and Problem Anchor Layer Refinement Stage. For each problem anchor, select knowledge points related to that problem anchor from the basic knowledge point map to generate candidate knowledge groups, and calculate the teaching load of the candidate knowledge groups; determine the relationship between problem anchors based on the candidate knowledge groups and the initial cross-layer mapping relationship.

[0056] The candidate knowledge group generation and problem anchor layer refinement stage is used to select knowledge points that can support each problem anchor from the basic knowledge point map based on the problem anchor set and initial cross-layer mapping relationship obtained in the previous stage, forming candidate knowledge groups; then, the core knowledge points and teaching load of each candidate knowledge group are determined; and the relationship between problem anchors is further improved.

[0057] This stage mainly obtains the relationships between candidate knowledge groups, knowledge support relationships, and problem anchors, as well as the load information of candidate knowledge groups.

[0058] This stage specifically includes the following steps: S3.1: Generate candidate knowledge groups around the problem anchor.

[0059] For each question anchor, this method selects knowledge points related to that question anchor from the basic knowledge point graph to form a candidate knowledge group corresponding to that question anchor. Specifically, based on the knowledge layer relationships and source evidence in step S1, and the initial cross-layer mapping relationships in step S2, the evidence relevance, graph relationship relevance, semantic relevance, and source location relevance between the question anchor and the knowledge point are determined; then, the relevance of the question anchor and the knowledge point is judged based on the evidence relevance, graph relationship relevance, semantic relevance, and source location relevance; if they are related, the knowledge point is added to the candidate knowledge group corresponding to that question anchor.

[0060] set up Indicates knowledge points, Indicates the problem anchor point. Representing knowledge points anchor point of the problem The relevance is then: in, Indicates the relevance of evidence. Indicates the correlation between the graphs. Indicates semantic relevance. Indicates the relevance of the source location; Weights that indicate the relevance of evidence The weights representing the relevance of the graph relationships. Weights representing semantic relevance The weights representing the relevance of the source location satisfy the following conditions: In one implementation, the following can be taken: In practical applications, the weights can be adjusted based on factors such as course type and the clarity of the textbook structure. Generally, the weights for evidence relevance and graph relationship relevance are higher than those for source location relevance to prevent this method from degenerating into continuous segmentation based on the original textbook table of contents.

[0061] Relevance of evidence Used to determine knowledge points anchor point of the problem Whether there is direct teaching material evidence can be determined according to the following rules: If multiple conditions are met simultaneously, the highest value is taken.

[0062] Correlation of Graph Relationships Used to represent knowledge points anchor point of the problem The strength of the relationship between the initial triggering knowledge points in the basic knowledge point graph. Let... Indicates the problem anchor point The initial set of triggering knowledge points, express For any knowledge point in the following, then: in, Representing knowledge points With knowledge points The strength of the relationship can be determined by the following rules: semantic relevance Used to represent knowledge points Name, description, and source evidence text and problem anchors The semantic similarity between the names and evidence texts. This value can be structurally determined through text vector similarity, keyword matching, semantic matching models, or large language models, and then normalized to... to This method does not limit the specific calculation model for semantic relevance.

[0063] Source location relevance Used to represent knowledge points anchor point of the problem The rules for determining the similarity of sources in textbooks can be as follows: Source location relevance is only a supplementary factor. Even if a knowledge point is not in the same chapter as the question anchor, it can still be included in the candidate knowledge group corresponding to the question anchor as long as its evidence relevance, graph relationship relevance, or semantic relevance is high.

[0064] set up Indicates the problem anchor point The corresponding candidate knowledge group Indicates the problem anchor point The initial set of triggering knowledge points, To represent the relevance threshold, then: Among them, the initial set of knowledge points Knowledge points in the above must be included in the candidate knowledge group; other knowledge points, if they meet the following conditions... If it is, then it will be added to the candidate knowledge group. It can be preset in advance or configured according to the course type. In one implementation method, it is preferable to... .

[0065] After candidate knowledge groups are generated, this method uses the question anchor point corresponding to the candidate knowledge group as the initial master question anchor point for that candidate knowledge group; for example, question anchor point Candidate knowledge groups The initial main problem anchor point is established; at the same time, knowledge support relationships are established between the initial main problem anchor point and the knowledge points of the candidate knowledge group.

[0066] Knowledge support relationships indicate that a certain knowledge point can support a certain problem anchor point. This represents a cross-layer mapping relationship between the problem anchor point layer and the underlying knowledge point graph layer, not a knowledge layer relationship between knowledge points. Subsequently, when merging or splitting candidate knowledge groups based on the target number of teaching knowledge units, the problem anchor points and their primary problem anchor points corresponding to the candidate knowledge groups can be updated according to the processing results.

[0067] S3.2: Determine the set of core knowledge points for candidate knowledge groups and calculate the teaching load of candidate knowledge groups.

[0068] This method first determines the set of core knowledge points for the entire book, then determines the core knowledge points in each candidate knowledge group based on the intersection of the candidate knowledge group and the set of core knowledge points for the entire book; calculates the teaching load of all core knowledge points, and sums them to obtain the teaching load of the candidate knowledge group.

[0069] set up This represents the collection of core knowledge points from the entire book. Representing knowledge points Structural salience, Representing knowledge points Material support, Indicates the structural saliency threshold. If we represent the material support threshold, then: in, It is a collection of core knowledge points used for subsequent coverage, overlap, and load calculations throughout the book. and This is a preset threshold. In one implementation, Desirable , Desirable .

[0070] set up Indicates candidate knowledge groups The set of knowledge points contained therein Indicates candidate knowledge groups The core knowledge points in the text are: .

[0071] In other words, the core knowledge points in a candidate knowledge group are determined by the intersection of the set of knowledge points in the candidate knowledge group and the set of core knowledge points in the whole book.

[0072] like It is an empty set, i.e., a candidate knowledge set. If a candidate knowledge group does not contain core knowledge points, it will be deleted and will not be included in subsequent calculations.

[0073] This method calculates the load of candidate knowledge groups based on the knowledge point load. Let... Representing knowledge points The teaching load, Representing knowledge points Structural salience, Representing knowledge points The complexity of knowledge types, Representing knowledge points Given the material support, then: in, Weights representing structural significance Weights representing the complexity of knowledge types. The weight representing the material support. And it satisfies: In one implementation, the following can be taken: .

[0074] Structural salience Used to represent knowledge points The degree of prominence within the textbook structure. Its value can be determined based on the location of the knowledge point, for example: If a knowledge point appears in multiple locations, the highest value is taken.

[0075] Knowledge type complexity This is used to indicate the relative complexity of different types of knowledge points in teaching organization. Its value can be determined based on the type of knowledge point, for example: Material support Used to represent knowledge points The degree to which it is supported by teaching materials such as examples, cases, exercises, experiments, or texts. Representation and knowledge points The number of teaching materials with explicit associations. If the upper limit of the material quantity is indicated, then: in, We can statistically analyze the source evidence between knowledge points and textbook units, for example, by statistically analyzing knowledge points. The number of different textbook units used in example questions, exercises, case studies, or experimental materials. This is used to prevent the workload of a particular knowledge point from increasing indefinitely due to an excessive amount of material. In one implementation, Desirable .

[0076] set up Indicates candidate knowledge groups The teaching load is as follows: .

[0077] This method calculates the instructional load only for the core knowledge points within candidate knowledge groups to avoid amplifying the instructional load due to auxiliary knowledge points or background materials. The instructional load of candidate knowledge groups is used to subsequently determine whether the candidate knowledge groups are too large or too small, and whether they need to be split or merged.

[0078] S3.3: Establish the relationship between problem anchors.

[0079] This method further establishes relationships between problem anchors to improve the problem anchor layer. The types of relationships between problem anchors mainly include decomposition relationships, merging relationships, and sequential relationships.

[0080] When establishing relationships between question anchors, this method compares each anchor point pairwise. During the comparison, information such as the anchor's name, supporting text, source textbook unit, textbook title level, and the relationships between knowledge points within the candidate knowledge group and the basic knowledge point map can be comprehensively utilized. The relationship between each anchor point can record the relationship type and its basis. The basis explains which textbook evidence, knowledge point relationships, or candidate knowledge groups contain, overlap, or connect to establish the relationship.

[0081] The determination of the relationship between problem anchors is carried out in the following order: Deduplicate problem anchors that actually express the same problem and have the same core knowledge points; determine whether there is a decomposition relationship between two problem anchors; for two problem anchors that do not have a decomposition relationship, determine whether they meet the conditions for merging; the priority relationship is determined separately based on the knowledge point prerequisite relationship and the order of the textbook.

[0082] When deduplicating question anchors, retain those with more textbook evidence and summarize their source textbook units, initial triggering knowledge points, evidence texts, and candidate knowledge groups. Question anchors with less textbook evidence are used as question anchors for the summarized candidate knowledge groups; these anchors do not participate in subsequent question anchor relationship establishment.

[0083] S3.31: Establishment of disassembly relationships.

[0084] The decomposition relation is used to represent a problem anchor with a large coverage that can be further broken down into two or more problem anchors with smaller coverage. This relation can be used for subsequent decomposition when it is necessary to increase the number of teaching knowledge units or refine candidate knowledge groups.

[0085] When determining whether there is a decomposition relationship between two problem anchors, first compare the objects discussed, the tasks to be completed, and the additional constraints. If the two problem anchors discuss similar content, and one of them has fewer constraints and covers more situations, then it is considered a larger problem anchor (a problem anchor with a wider coverage); the other problem anchor adds specific conditions, components, or processing steps, then it is considered a smaller problem anchor (a problem anchor with a smaller coverage). For example, the problem anchor "How to programmatically express repetitive tasks?" discusses various repetitive tasks and has a wider coverage, while the other problem anchor "How to express a repetitive task with a fixed number of repetitions?" adds the condition of "fixed number of repetitions" and has a smaller coverage.

[0086] set up This serves as an anchor point for larger issues. For smaller issues, Indicates the problem anchor point The set of core knowledge points corresponding to the candidate knowledge group. Indicates the problem anchor point This corresponds to the set of core knowledge points within the candidate knowledge group. It can be calculated using the following formula. The core knowledge points in Coverage ratio : The molecule represents and The number of core knowledge points commonly included, represented by the denominator. The total number of core knowledge points included. The closer the value is to 1, the better. More and more core knowledge points are included in middle.

[0087] When this ratio is greater than or equal to a preset coverage threshold, in one implementation method, it can be taken as follows: And when the following conditions are met simultaneously, establish a larger problem anchor point. Pointing to problem anchors with limited coverage The disassembly relationship: (1) It can be judged from the problem statement The coverage area is greater than ; (2) Also includes Core knowledge points not covered; (3) The two textbooks have common textbook titles, common learning objectives, the same examples, the same cases, the same textbook introductions, or similar sources.

[0088] By comparing each pair of problem anchors, we can identify which smaller problem anchors have a decomposition relationship with the larger problem anchors. Each decomposition relationship indicates that a smaller problem anchor corresponds to a portion of the content of the larger problem anchor.

[0089] To provide a basis for subsequent candidate knowledge group decomposition, this stage also needs to determine whether multiple smaller question anchors combined can cover the main content of the larger question anchor. Therefore, it is necessary to summarize all smaller question anchors that have established decomposition relationships with the larger question anchor and calculate their overall coverage of the larger question anchor.

[0090] Suppose and The anchor points for establishing the decomposition relationship are: in, Indicates and The combination of smaller problem anchor points that establish a decomposition relationship. Indicates and Establish the r-th smaller problem anchor point for the decomposition relationship, where r represents the number of smaller problem anchor points.

[0091] To determine whether these smaller problem anchors can be combined to cover the issue The main content, calculating their impact on Overall coverage : In the formula, This represents all the core knowledge points contained within all the smaller problem anchors; This represents the i-th smaller problem anchor point. The numerator represents... Able to cover The number of core knowledge points; the denominator represents The total number of core knowledge points included. The closer the value of is to 1, the more complete the coverage of the larger problem anchor points is by the combination of these smaller problem anchor points.

[0092] When the overall coverage ratio is greater than or equal to a preset threshold, in one implementation method, it can be taken as follows: This group of smaller problem anchors is recorded as an executable splitting combination. This combination indicates that the larger problem anchor can be actually split into these smaller problem anchors.

[0093] For example, for problem anchors "How can repetitive tasks be expressed programmatically?", if they have already been established separately: How to express a task that is repeated a fixed number of times? How to express condition-driven repetitive tasks?

[0094] when and The core knowledge points combined can cover When most of the core knowledge points are covered, and both contain unique knowledge points related to fixed-number loops and conditional loops respectively, then... Record as Executable splitting and combining. This represents the B-th problem anchor point, which is The smaller problem anchor point.

[0095] S3.32: Establishment of merge relationship.

[0096] The merge relation is used to indicate that two problem anchors are suitable to be placed in the same teaching knowledge unit. The merge relation is used to merge subsequent candidate knowledge groups when there are too many, the content is too small, or the relationship is too strong.

[0097] For the same pair of question anchors, if one question's coverage clearly encompasses the other, a split relationship is established. A merge relationship is used for two question anchors with similar coverage and complementary content. Therefore, the same pair of question anchors cannot have both a split and a merge relationship established simultaneously.

[0098] Merging relationships can be established when there is no obvious scope inclusion between two problem anchors, but their coverage is similar and they focus on different parts of the same teaching content.

[0099] Two problem anchors are merged when both satisfy the following conditions: (1) The two problem anchors discuss the same or closely related objects, the proposed tasks belong to different aspects or different processing steps of the same teaching task, and there is no inclusion relationship in the scope of coverage between the two; (2) The core knowledge points corresponding to the two problem anchors focus on different parts of the same teaching content; (3) There is a compositional relationship, co-occurrence relationship or similarity relationship between the core knowledge points corresponding to the two problem anchor points; (4) The two problem anchors have textbook evidence such as common textbook title, common learning objectives, the same example, the same case, the same textbook introduction, or similar source location.

[0100] For example, the questions "How should the loop variable be set?" and "How should the loop body be designed?" both discuss loops with a fixed number of iterations, and a merging relationship can be established between the two.

[0101] S3.33: Establishment of sequential relationships.

[0102] The sequence relationship is used to indicate that the teaching content corresponding to one problem anchor should be taught before the teaching content corresponding to another problem anchor. For example, "How to express a task that is repeated a fixed number of times?" should usually be taught before "How to express multiple levels of repetition through nested loops?".

[0103] When establishing the sequence relationship, priority is given to calculating the prerequisite relationships of knowledge points between candidate knowledge groups corresponding to two question anchors. For the i-th candidate knowledge group... and the j-th candidate knowledge group ,set up for China does not belong to The number of knowledge points, for China does not belong to The number of knowledge points. Among the knowledge points, The number of knowledge points for which at least one knowledge point is a direct prerequisite is denoted as . The same knowledge point is counted only once. for Provides the coverage ratio of direct prerequisite knowledge. for: .

[0104] Similarly, calculate for Provides the coverage ratio of direct prerequisite knowledge. When the denominator is zero, the corresponding proportion is zero. When the following conditions are met: and Then establish The corresponding problem anchor point precedes The order of the anchor points corresponds to the order of the questions, and the reverse is also true.

[0105] In one implementation, the priority relationship threshold A value of 0.6 is acceptable. When both directions reach or fail to reach the threshold for sequential relationship, the judgment should be made by referring to the basic level of the local order, heading level, and question anchor point in the original textbook.

[0106] After deduplication and relational analysis, this stage yields a set of relationships between problem anchor points. This set of relationships, along with candidate knowledge groups, serves as the basis for determining the number of teaching knowledge units, merging candidate knowledge groups, splitting candidate knowledge groups, and ranking teaching knowledge units in the next stage.

[0107] In one implementation, the set of relationships between problem anchors is calculated using a large language model, taking the problem anchors, their corresponding candidate knowledge groups, initial cross-layer mapping relationships, basic knowledge point graphs, and textbook units as inputs.

[0108] S4: Determining the Number of Teaching Knowledge Units and Generating a Division Scheme. Determine the target number of teaching knowledge units; for each target number of teaching knowledge units, divide candidate knowledge groups based on the relationships between problem anchors, generating candidate division schemes.

[0109] The stage of determining the number of teaching knowledge units and generating a division scheme is used to generate a set of candidate target numbers. Under each target number of teaching knowledge units, the candidate knowledge groups obtained in the previous stage are merged or split to form a candidate division scheme.

[0110] S4.1: Generate a set of candidate target numbers.

[0111] This method first determines whether the user has specified a target number of teaching knowledge units. If the user has specified a target number of teaching knowledge units, the specified number is added to the candidate target number set and used as the priority target number. If the teacher has not specified a target number of teaching knowledge units, a candidate target number set is generated based on the total load of the core knowledge points in the entire book and the allowable load range of a single teaching knowledge unit.

[0112] set up Represents the set of candidate target numbers. This represents the number of target teaching knowledge units in the candidate target quantity set. If the teacher specifies the number of target teaching knowledge units as... Then, the specified quantity will be added to the candidate target quantity set: .

[0113] If the user does not specify the target number of teaching knowledge units, or if the specified number is not feasible, the candidate number range will be determined according to the following formula: in, This indicates a lower bound on the number of candidate targets. This indicates an upper bound on the number of candidate targets. This indicates the total load of the core knowledge points in the entire book. This indicates the maximum allowable load for a single unit of instructional knowledge. This represents the minimum allowable workload for a single unit of instructional knowledge. This indicates that the total number of anchor points for the main problem is equal to the total number of candidate knowledge groups after processing by S3. Indicates rounding up. This indicates taking the minimum value. This indicates rounding down to the nearest integer.

[0114] Based on the above range, the set of candidate target numbers It can be represented as: .

[0115] The term "user-specified number not feasible" means that the candidate partitioning scheme generated based on the number of target teaching knowledge units specified by the user is not feasible (determined by S5).

[0116] S4.2: Based on the relationship between the problem anchor points, determine the candidate knowledge group to be used when generating candidate partitioning schemes, and obtain the current candidate knowledge group set.

[0117] For a combination of problem anchors with a decomposition relationship, the candidate knowledge group corresponding to the smaller problem anchor that constitutes the executable decomposition combination is temporarily removed from the candidate knowledge group obtained after processing in S3, and the current candidate knowledge group set is obtained.

[0118] Candidate knowledge groups corresponding to smaller problem anchors that constitute executable splitting combinations are not permanently deleted, but are only used during S4.3 splitting.

[0119] For example, and There is a disassembly relationship. for The executable splitting and combining, then Included in the current candidate knowledge group set. It is not included in the current candidate knowledge group set.

[0120] For multi-level executable splitting combinations, when determining the current candidate knowledge group set, only the candidate knowledge group corresponding to the largest problem anchor point at the top level is used. The candidate knowledge groups corresponding to the problem anchor points at each lower level are retained for use in subsequent splitting. When splitting is required in S4.3, the candidate knowledge group corresponding to the top-level problem anchor point is split first. After completing one split, if the obtained problem anchor points also have executable splitting combinations, the conditions in S4.3 are used to determine whether to continue splitting.

[0121] S4.3: Given the target number of teaching knowledge units, divide the candidate knowledge groups based on the current set of candidate knowledge groups and perform merging or splitting operations.

[0122] For the set of candidate target numbers For each target teaching knowledge unit in the dataset, this method generates corresponding candidate partitioning schemes.

[0123] In this stage, candidate knowledge groups are merged or split within the target number of teaching knowledge units, and the set of question anchors and the main question anchor contained in the candidate knowledge group are updated according to the merging or splitting results. When multiple candidate knowledge groups are merged, the new candidate knowledge group inherits and merges the set of question anchors of the original candidate knowledge group; when a candidate knowledge group (corresponding to a larger question anchor) is split, each sub-candidate knowledge group (corresponding to a smaller question anchor) uses the corresponding sub-question anchor in the selected executable splitting combination as the main question anchor.

[0124] set up This represents the current set of candidate knowledge groups. This indicates the number of candidate knowledge groups in the current candidate knowledge group set. ,but: .

[0125] like Then, a candidate partitioning scheme can be formed directly based on the current set of candidate knowledge groups.

[0126] like This indicates that the number of candidate knowledge groups exceeds the number of target teaching knowledge units, and the candidate knowledge groups should be merged.

[0127] Multiple candidate knowledge groups can be merged when they simultaneously meet the following merging conditions: (1) Merging relationships have been established between the main question anchors of multiple candidate knowledge groups; (2) The combined teaching load shall not exceed the maximum allowable load of a single teaching knowledge unit.

[0128] In another implementation, combinations of candidate knowledge groups that meet the above merging criteria can be further selected in the following order: (1) Prioritize combinations in which at least one candidate knowledge group has a load lower than the minimum allowable load for a single instructional knowledge unit; (2) Prioritize combinations with more substantial evidence of shared teaching materials recorded in the previous stage; (3) Prioritize the combination of two candidate knowledge groups with more knowledge layer relationships and stronger connections.

[0129] set up and This represents two candidate knowledge groups to be merged. If we represent the new candidate knowledge group formed after the merger, then: in, Represents candidate knowledge group i. Represents candidate knowledge group j.

[0130] The merged candidate knowledge group inherits and merges the problem anchors, knowledge points, core knowledge points, and source paths from the original candidate knowledge group. Simultaneously, the main problem anchor of the new candidate knowledge group is selected as the problem anchor with a high load of core knowledge points, covered through knowledge support relationships.

[0131] like This indicates that the number of candidate knowledge groups is less than the number of target teaching knowledge units, and candidate knowledge group splitting should be performed.

[0132] A candidate knowledge group can be split if it meets all of the following splitting conditions: (1) The main question anchor point of the candidate knowledge group has executable splitting and combination; (2) After the candidate knowledge group is replaced by multiple sub-candidate knowledge groups, the number of candidate knowledge groups in the current candidate knowledge group set shall not exceed the number of target teaching knowledge units; (3) The core knowledge points in the candidate knowledge group can be assigned to at least one sub-candidate knowledge group and retain the corresponding source path.

[0133] In another implementation, candidate knowledge groups that meet the above-mentioned splitting conditions can be further selected in the following order: (1) Prioritize combinations where the number of candidate knowledge groups after splitting is closer to the number of target teaching knowledge units; (2) Prioritize candidate knowledge groups with a large current teaching load; (3) Prioritize the selection of executable splitting and combination that has a higher coverage rate of the core knowledge points of the original candidate knowledge group; (4) Prioritize combinations of sub-candidate knowledge groups with less overlap in core knowledge points; (5) Prioritize the combination of sub-candidate knowledge groups with relatively balanced loads after splitting.

[0134] The above merging or splitting process can be repeated until... Until no suitable candidate knowledge groups are found to continue merging or splitting. If a quantity equal to [a certain number] cannot be obtained... If the current set of candidate knowledge groups is not found, then the partitioning scheme under this target number is marked as failed to generate. No candidate partitioning schemes will be generated under this target number, and other candidate target numbers will continue to be processed.

[0135] If, after merging or splitting, a set of candidate teaching knowledge units is obtained that is equal to the number of target teaching knowledge units, then a candidate partitioning scheme is formed under that number of target teaching knowledge units.

[0136] Each candidate teaching knowledge unit in the candidate partitioning scheme includes at least one main question anchor and one core knowledge point. A candidate teaching knowledge unit may contain one or more question anchors, or it may contain knowledge points from different chapters or sections of the original textbook. Each candidate teaching knowledge unit retains the knowledge point source path and the mapping path from the question anchor to the knowledge point.

[0137] After this stage of processing, candidate partitioning schemes corresponding to the number of target teaching knowledge units can be obtained. It should be noted that when merging and splitting candidate knowledge groups, multiple feasible schemes may exist; each feasible scheme will generate a corresponding candidate partitioning scheme. That is, in this stage, the number of target teaching knowledge units is one or more, and the number of candidate partitioning schemes corresponding to the number of target teaching knowledge units is one or more.

[0138] S5: Evaluation of the partitioning scheme and determination of teaching knowledge units. Based on the coverage rate of core knowledge points, the overlap rate of core knowledge points, the order violation rate, and the load variance, the optimal candidate partitioning scheme is selected, and the final teaching knowledge units and their teaching order suggestions are determined.

[0139] The evaluation and determination of teaching knowledge units stage is used to evaluate the candidate partitioning schemes generated in the previous stage, select the optimal candidate partitioning scheme, and determine the final number of teaching knowledge units, the final teaching knowledge units, and their suggested teaching order. This stage uses four calculable indicators to evaluate the candidate partitioning schemes: core knowledge point coverage rate, core knowledge point overlap rate, order violation rate, and load variance.

[0140] S5.1: Calculate the evaluation index of the partitioning scheme: Let Indicates the candidate partitioning scheme, Representing candidate partitioning schemes The global score is then: in, This indicates the coverage rate of core knowledge points. This indicates the overlap rate of core knowledge points. Indicates the order violation rate. Indicates the load variance; The preset weights represent the coverage of core knowledge points. The preset weights representing the overlap rate of core knowledge points. The preset weights representing the order violation rate The preset weights represent the load variance.

[0141] The global score is used to measure the overall quality of candidate partitioning schemes. Among these, the coverage rate of core knowledge points is crucial. This is a positive indicator used to encourage solutions to cover the core knowledge points in the textbook as much as possible; the overlap rate of core knowledge points. Order violation rate and load variance This is a penalty item used to reduce the score of candidate partitioning schemes that have too much repetition of knowledge points, unreasonable teaching order, or unbalanced unit load.

[0142] In one implementation, the following can be taken: The weights can be adjusted according to the course type, textbook size, or teaching organization requirements.

[0143] Core knowledge point coverage This is used to represent the degree to which candidate partitioning schemes cover the core knowledge points of the entire book, and its calculation method is as follows: in, Representing candidate partitioning schemes One of the candidate teaching knowledge units. Indicates candidate teaching knowledge units The core knowledge points in the text This represents the set of core knowledge points in the entire book. The purpose of this formula is to determine whether the candidate partitioning schemes adequately cover the main knowledge content of the textbook.

[0144] Core knowledge point overlap rate The degree of overlap in core knowledge points among different candidate teaching knowledge units is used to represent the extent of this overlap. The calculation method is as follows: in, This represents the sum of the number of core knowledge points in each candidate teaching knowledge unit. Core knowledge points that appear repeatedly will be counted repeatedly. This represents the number of core knowledge points obtained after merging and deduplicating the core knowledge points from all candidate teaching knowledge units. The difference between the two represents the number of core knowledge points that were over-counted due to duplicate allocation.

[0145] Order violation rate This is used to indicate whether a candidate partitioning scheme violates the priority relationship of knowledge points or the order of question anchor points. Since the order violation rate depends on the order of candidate teaching knowledge units, this method calculates... First, a temporary order is generated for the candidate partitioning schemes based on the sequential relationships between problem anchors, the prerequisite relationships between knowledge points, and the local order of the original textbook. The order violation rate is calculated based on this temporary order, and the final formal ranking is determined in subsequent steps of this stage.

[0146] set up Indicates the number of order constraints that were violated. Let the total number of sequence constraints be represented, then: Sequence constraints include prerequisite relationships between knowledge points and sequential relationships between problem anchors. If two objects with sequence constraints are located in the same candidate teaching knowledge unit, it is not considered a violation of the order between candidate teaching knowledge units, and the teaching order can be arranged within that candidate teaching knowledge unit. If there are no sequence constraints, then... .

[0147] Load variance This is used to indicate whether the teaching load is balanced among the candidate teaching knowledge units in the candidate partitioning scheme. Let... Indicates candidate teaching knowledge units The teaching load, To represent the average load, then: in, This represents the average workload of each candidate teaching knowledge unit in the candidate partitioning scheme. This formula determines whether some units have too much content and others too little content by calculating the deviation of each teaching knowledge unit's workload from the average workload. Since candidate teaching knowledge units must contain at least one core knowledge point, and the core knowledge point has a teaching workload, this is considered in feasible schemes. It should be greater than zero.

[0148] S5.2: Screening feasible candidate partitioning schemes: After calculating the evaluation index of the candidate partitioning schemes, this method uses the evaluation threshold to determine whether the candidate partitioning scheme is feasible.

[0149] set up This represents the threshold for coverage of core knowledge points. This represents the threshold for the overlap rate of core knowledge points. Indicates the order violation rate threshold. This represents the load variance threshold. Candidate partitioning schemes. The following conditions must be met simultaneously: in, This is to ensure that the core knowledge points are fully covered. Used to limit the repetition of knowledge points between teaching knowledge units. Used to restrict violations of the teaching sequence. Used to limit the load differences between units of instructional knowledge.

[0150] In one implementation, the following can be taken: The evaluation thresholds can be adjusted based on the size of the teaching materials, the type of course, or the teaching requirements.

[0151] In addition to the evaluation thresholds mentioned above, candidate partitioning schemes should also meet the following basic constraints: (1) Each candidate instructional knowledge unit has at least one main question anchor point; (2) Each candidate teaching knowledge unit contains at least one core knowledge point; (3) The core knowledge points included in the teaching knowledge units in the candidate division schemes should have a source path; (4) The number of candidate teaching knowledge units should be equal to the number of corresponding target teaching knowledge units; (5) There should be a cross-level mapping relationship or traceable evidence between the main problem anchor point and the core knowledge point in the teaching knowledge unit.

[0152] If a candidate partitioning scheme does not meet the above evaluation threshold or basic constraints, it will not be considered a feasible scheme.

[0153] S5.3: Determine the final partitioning scheme: After calculating the global score of each candidate partitioning scheme and selecting feasible candidate partitioning schemes according to the evaluation threshold or basic constraints, select the feasible candidate partitioning scheme with the highest global score as the final partitioning scheme, and the number of its candidate teaching knowledge units is taken as the final number of teaching knowledge units.

[0154] If no feasible candidate partitioning scheme exists, the scheme with the highest global score among the generated candidate partitioning schemes is selected as the approximate recommendation result to obtain the final partitioning scheme. The unmet feasibility conditions are marked in the output. This approximate recommendation result does not represent complete satisfaction of constraints, but rather the usable output result under the current textbook materials, candidate knowledge groups, and threshold settings.

[0155] If there are multiple candidate partitioning schemes with the same global score, the candidate partitioning scheme with the lower overlap rate of core knowledge points or the smaller load variance is selected as the final partitioning scheme.

[0156] S5.4: Final ordering of teaching knowledge units.

[0157] Each candidate teaching knowledge unit in the final partitioning scheme will be determined as the final teaching knowledge unit. Each teaching knowledge unit includes at least a main question anchor, a set of knowledge points, a set of core knowledge points, and a knowledge point source path. If a teaching knowledge unit contains multiple question anchors, the main question anchor will be used as the organizational center of the teaching knowledge unit, used for subsequent real-world question generation and naming of the teaching knowledge unit; other question anchors will be retained as related question anchors within the teaching knowledge unit.

[0158] Subsequently, this method sorts the final teaching knowledge units to obtain a suggested teaching order for the final teaching knowledge units. When sorting, priority is given to the sequential relationships between problem anchors and the prerequisite relationships between knowledge points; if no clear sequential or prerequisite relationships exist, the local order in the original textbook is referenced; if the order still cannot be determined under the above conditions, an auxiliary sorting can be performed according to the principle of prioritizing basic problems over comprehensive problems.

[0159] This stage of processing yields the final number of teaching knowledge units, the final set of teaching knowledge units, and a suggested order for teaching the teaching knowledge units.

[0160] S6: Real-world introduction question generation and result output stage. Real-world introduction questions are generated for each final teaching knowledge unit, and combined with the final teaching knowledge unit and its suggested teaching sequence, the teaching knowledge unit results are obtained by organizing around the question anchor points.

[0161] The stage of generating and outputting real-world introductory questions is used to generate corresponding real-world introductory questions for each teaching knowledge unit and output the final teaching knowledge unit division results.

[0162] This method generates realistic introductory questions for each final teaching knowledge unit. These questions are used to elicit the corresponding question anchors and knowledge point sets, serving as an introduction to the teaching process. The generation of realistic introductory questions is based on the main question anchors, knowledge point sets, core knowledge point sets, examples, case studies, and source texts for that teaching knowledge unit.

[0163] Real-world introductory questions should be rooted in real-life situations, academic disciplines, society, engineering projects, texts, or other authentic contexts. They should naturally lead to the corresponding main question anchor and connect to the core knowledge points of the final teaching unit. For example, for the main question anchor "How to express a repetitive task with a fixed number of repetitions," a real-world introductory question could be generated: "The class monitor needs to collect 8 homework assignments. How can they check each one without missing a single one?"

[0164] After a real-world import issue is generated, this method performs a lightweight validation. The validation includes: (1) Whether the real-world introductory question corresponds to the main question anchor point of the teaching knowledge unit; (2) Whether the real-world introductory question can be linked to the core knowledge points in this teaching unit; (3) Does the real-world introduction problem clearly introduce a large number of knowledge points from other teaching knowledge units? (4) Whether the real-world introductory question has a clear context and is suitable for introducing this teaching knowledge unit.

[0165] If the generated real-world import question cannot correspond to the main question anchor point, or cannot be associated with the core knowledge points in this teaching knowledge unit, this method can regenerate the real-world import question, or select a more suitable question from multiple candidate real-world import questions.

[0166] After generating and validating real-world import questions, this method outputs the final result: a set of teaching knowledge units organized around question anchors. This set includes: the final number of teaching knowledge units, the unit number, the main question anchor for each unit, the question anchors contained in each unit, the set of knowledge points for each unit, the set of core knowledge points for each unit, the source path for each knowledge point, the mapping path from question anchors to knowledge points, a suggested teaching order for each unit, and the real-world import question for each unit.

[0167] The teaching knowledge units organized around problem anchors are used to retain the source of knowledge points, the correspondence between problem anchors and knowledge points, and the recommended teaching order.

[0168] If the previous stage output an approximate recommendation, this stage can further mark the final result as an approximate recommendation and list the unmet feasibility conditions, such as insufficient coverage of core knowledge points, unbalanced teaching knowledge unit load, or high rate of sequence violation. This marking serves as a reminder to pay close attention to these issues when rewriting teaching materials, generating lesson plans, or using the materials in teaching.

[0169] Through this stage of processing, this method ultimately yields teaching knowledge units organized around problem anchors. These results retain the original sources of knowledge points from the reference materials and provide the main problem anchors for problem-driven learning, the organization of knowledge points, and real-world introductory questions. This can serve as a structured foundation for subsequent textbook rewriting, lesson plan generation, or classroom teaching design.

[0170] Compared to existing technologies such as knowledge graph construction, course problem graphs, lesson plan generation, and learning path recommendation, this invention focuses on the division of teaching knowledge units in problem-driven teaching. Its core approach lies in setting up a problem anchor layer on top of the basic knowledge point graph. Through cross-layer mapping relationships and problem anchor relationships, the discontinuous knowledge points in the original textbook are reorganized into teaching knowledge units centered around the core teaching problem. This is mainly manifested in: (1) Set up an independent problem anchor layer.

[0171] Existing course knowledge graphs or course problem graphs are typically used to express the relationships between course outlines, abilities, problems, knowledge points, and teaching resources, with the focus on forming a course knowledge system or resource organization structure.

[0172] This invention sets up an independent question anchor layer in addition to the basic knowledge point map. The basic knowledge point map is used to express knowledge points, textbook units, source evidence, and the prerequisite, composition, co-occurrence, and similarity relationships between knowledge points; the question anchor layer is used to express the central teaching question that can encompass several knowledge points.

[0173] The problem anchor points in this invention are different from ordinary knowledge point nodes in traditional knowledge point maps, as well as question entries such as practice questions and test questions. Their function is to serve as intermediate layer organization objects, supporting the generation of candidate knowledge groups, the combination and splitting of candidate knowledge, and the determination and sorting of teaching knowledge units.

[0174] (2) Establish cross-layer mapping relationship between problem anchor layer and knowledge point map layer.

[0175] Existing course problem graphs, competency graphs, or knowledge graphs can establish correspondences between problems, competencies, knowledge points, and teaching resources to express course structure, knowledge systems, or resource organization relationships. These relationships typically focus on illustrating the connections between different learners.

[0176] This invention establishes a cross-layer mapping relationship between an independent problem anchor layer and a lower-level knowledge point graph layer to connect the generation basis of the problem anchor, textbook evidence, and supporting knowledge points. Specifically, the initial trigger relationship records which knowledge points lead to the problem anchor, the textbook evidence relationship records which textbook content provides the basis for the problem anchor, and the knowledge support relationship records the knowledge points that can support the problem anchor after the candidate knowledge group is generated.

[0177] Based on this mapping relationship, the present invention can filter knowledge points that support the same problem anchor from the underlying knowledge point graph, and retain the corresponding path between the problem anchor, knowledge point and the original teaching material in the final teaching knowledge unit.

[0178] (3) Generate non-continuous candidate knowledge groups around the problem anchor point.

[0179] Existing textbook catalog knowledge graphs, course maps, or learning path recommendation methods are usually organized based on textbook catalogs, course teaching logic, prerequisite relationships of knowledge points, or learners' personalized learning paths.

[0180] This invention establishes a cross-layer mapping between the basic knowledge point map and the question anchor point layer. Then, it selects and aggregates knowledge points from the basic knowledge point map around the question anchor point to form candidate knowledge groups. These candidate knowledge groups are not required to originate from consecutive chapters or sections in the original textbook; instead, they are determined based on a comprehensive assessment of the evidence relevance, map relationship relevance, semantic relevance, and source location relevance between the knowledge points and the question anchor point. In other words, knowledge points scattered across different chapters or sections in the original textbook can be reorganized into candidate knowledge groups according to their support relationships for the same question anchor point. This approach differs from methods that simply use the textbook's table of contents for chapter segmentation or generate personalized learning paths solely based on knowledge point relationships.

[0181] (4) Apply the problem anchor relationship to the splitting, merging and sorting of candidate knowledge groups.

[0182] Existing problem chain or problem graph technologies typically focus on expressing the logical relationships between problems, or generating teaching content based on problem chains; while existing knowledge graph or learning path technologies typically focus on the prerequisite, sequential, or path relationships between knowledge points.

[0183] This invention establishes decomposition, merging, and sequential relationships between problem anchor points and directly applies them to the process of dividing teaching knowledge units. Specifically, the decomposition relationship is used to split candidate knowledge groups when there are many target teaching units or a heavy load on candidate knowledge groups; the merging relationship is used to merge candidate knowledge groups when there are few target teaching units or a light load on candidate knowledge groups; and the sequential relationship is used for sorting teaching knowledge units and calculating the order violation rate.

[0184] The relationship between the problem anchors in this invention is not only used to demonstrate problem logic or generate problem chain text, but also serves as the processing basis in the candidate knowledge group reorganization process, directly participating in the formation of teaching knowledge units.

[0185] (5) Generate and evaluate the division scheme under the constraint of the number of teaching units.

[0186] Existing knowledge graph-based lesson plan generation methods typically focus on how to combine textbook knowledge graphs, course information, student information, or class schedules to generate lesson plan content; learning path recommendation methods are usually geared towards individual learners, focusing on the learning order of knowledge points or personalized learning paths.

[0187] This invention addresses the problem of dividing teaching knowledge units before lesson plans or textbook texts are generated. Specifically, this invention first forms a set of candidate target quantities based on a number specified by the teacher or the result of algorithm calculation. Then, under each target teaching unit quantity, candidate knowledge groups are merged or split to form corresponding candidate division schemes. The candidate division schemes are evaluated through core knowledge point coverage, core knowledge point overlap, order violation rate, and load variance.

[0188] This invention can determine the appropriate number of teaching units, as well as the problem anchors and knowledge points contained in each teaching knowledge unit, thereby forming a structured division result for the generation of subsequent lesson plans or textbook texts.

[0189] (6) Output traceable teaching knowledge unit structure.

[0190] Existing teaching knowledge graph construction schemes typically organize textbook knowledge points, knowledge point attributes, and resources such as images, audio, and video into multimodal teaching knowledge graphs to enrich the presentation of knowledge and enhance classroom engagement. Existing lesson plan generation schemes usually focus on generating course lesson plan texts, teaching activity arrangements, or organizing teaching process texts according to question chains.

[0191] After the teaching knowledge units are determined, this invention outputs a structured result of the division of teaching knowledge units. This result includes the final number of teaching units, the problem anchor points of each teaching knowledge unit, the set of knowledge points, the source path of knowledge points, the cross-level mapping path between problem anchor points and knowledge points, the teaching order suggestion, and the real-world introduction question corresponding to each teaching knowledge unit.

[0192] The output of this invention preserves the corresponding path from the problem anchor point to the knowledge point and then to the source of the teaching materials, which can serve as a structured basis for subsequent textbook rewriting, lesson plan generation or problem-driven instructional design.

[0193] 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 reorganizing textbook knowledge points based on problem anchors and knowledge graphs, characterized in that, include: S1: Obtain structured teaching materials, analyze the teaching materials to obtain teaching material units, and construct a basic knowledge point map based on the teaching material units; S2: Construct problem anchors based on the textbook; And establish an initial cross-level mapping relationship between problem anchors and knowledge points in textbook units and basic knowledge point maps; S3: For each question anchor point, select knowledge points related to that question anchor point from the basic knowledge point graph to generate candidate knowledge groups; determine the relationship between question anchor points based on the candidate knowledge groups and the initial cross-layer mapping relationship; among which, the relationship between question anchor points includes decomposition relationship, merging relationship and sequential relationship; S4: Determine the number of teaching knowledge units; for each teaching knowledge unit, divide the candidate knowledge groups according to the relationship between the problem anchors and generate candidate division schemes; In S4, candidate knowledge groups are divided. When splitting candidate knowledge groups, larger problem anchors with executable splitting combinations are split into multiple smaller problem anchors. Among them, for larger and smaller problem anchors with splitting relationships, if a larger problem anchor corresponds to multiple smaller problem anchors, and the overall coverage ratio of the multiple smaller problem anchors to the larger problem anchor is greater than or equal to a preset threshold, then the multiple smaller problem anchors constitute an executable splitting combination. S5: Select the optimal candidate partitioning scheme based on the coverage rate of core knowledge points, the overlap rate of core knowledge points, the order violation rate, and the load variance, and determine the final teaching knowledge units and their teaching order suggestions; S6: Generate realistic introductory questions for each final teaching knowledge unit, and combine the final teaching knowledge unit with its teaching sequence suggestions to obtain the teaching knowledge unit results organized around the question anchors.

2. The method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs as described in claim 1, characterized in that, In S2, candidate question anchors are generated from the textbook title, learning objectives, examples, cases, textbook introduction, and textual descriptions that reflect the use or applicable context of knowledge points. Question anchors are then selected from the candidate question anchors based on the textbook unit and basic knowledge point map. The selection criteria are as follows: if a candidate question anchor is only the name of the knowledge point itself, or lacks the source textbook unit, the initial triggering knowledge point, or the evidence text, it will not be used as a question anchor.

3. The method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs as described in claim 1, characterized in that, In S3, the relevance of the question anchor and the knowledge point is determined based on the relevance of evidence, the relevance of the graph relationship, the semantic relevance, and the relevance of the source location. Among them, the relevance of evidence is used to determine whether the knowledge point and the question anchor point have direct textbook evidence; The graph relational degree is used to represent the strength of the relationship between the initial triggering knowledge points of knowledge points and problem anchors in the basic knowledge point graph; Semantic relevance is used to represent the semantic similarity between the name, description, and source evidence text of a knowledge point and the name and evidence text of the question anchor. Source location relevance is used to indicate how close the source location of knowledge points and question anchors is in the textbook.

4. The method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs as described in claim 1, characterized in that, In S3, Comparing two problem anchors, if their content is similar, one problem anchor has fewer restrictions and can cover more situations, so it is regarded as a larger problem anchor; the other problem anchor adds specific conditions, specific components or specific processing steps, so it is regarded as a smaller problem anchor. A decomposition relationship is established from the larger problem anchor to the smaller problem anchor when the proportion of core knowledge points in the smaller problem anchor that are covered by the larger problem anchor is greater than or equal to a preset coverage threshold, and the following conditions are met simultaneously: Condition 1: Judging from the problem description, the coverage of the larger problem anchor point is greater than that of the smaller problem anchor point; Condition 2: The larger problem anchor point contains core knowledge points that are not covered by the smaller problem anchor point; Condition 3: Evidence that the two materials share the same title, common learning objectives, the same example, the same case, the same introduction, or are from similar sources; For two problem anchors that do not have a decomposition relationship, determine whether the conditions for merging are met; If there is no obvious scope inclusion relationship between two question anchors, but their coverage is similar and they focus on different parts of the same teaching content, then the two question anchors have a merging relationship. The order of priority is determined based on the prerequisite relationships of knowledge points and the content of the textbook.

5. The method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs as described in claim 1, characterized in that, Overall coverage The calculation formula is: in, Indicates the problem anchor point The set of core knowledge points corresponding to the candidate knowledge group. This serves as an anchor point for larger issues. This represents all the core knowledge points contained within all the smaller problem anchors. Indicates and The combination of smaller problem anchor points that establish a decomposition relationship. This represents the i-th smaller problem anchor point.

6. The method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs as described in claim 1, characterized in that, In S5, the global score of the candidate partitioning scheme is calculated based on the core knowledge point coverage, core knowledge point overlap, order violation rate and load variance, and the candidate partitioning scheme with the highest global score is selected as the final partitioning scheme. Each candidate teaching knowledge unit in the final division scheme is the final teaching knowledge unit; Based on the sequential relationship between problem anchors and the prerequisite relationship between knowledge points, the final teaching knowledge units are ordered to obtain a suggested teaching order for the final teaching knowledge units.

7. The method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs as described in claim 6, characterized in that, The formula for calculating the global score is: in, Representing candidate partitioning schemes The overall score Indicates the coverage rate of core knowledge points. This indicates the overlap rate of core knowledge points. Indicates the order violation rate. Indicates the load variance; The preset weights represent the coverage of core knowledge points. The preset weights representing the overlap rate of core knowledge points. The preset weights representing the order violation rate Preset weights representing the load variance; Core knowledge point coverage rate is used to indicate the degree to which candidate partitioning schemes cover the core knowledge points of the entire book. The core knowledge point overlap rate is used to represent the degree of overlap in the core knowledge points among different candidate teaching knowledge units. The order violation rate is used to indicate whether a candidate partitioning scheme violates the prerequisite relationship of knowledge points or the order relationship of problem anchor points. Load variance is used to indicate whether the teaching load is balanced among the candidate teaching knowledge units in the candidate partitioning scheme.

8. The method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs as described in claim 7, characterized in that, The teaching load of a candidate knowledge unit is calculated based on the teaching load of its corresponding candidate knowledge group; The calculation process for the teaching load of candidate knowledge groups is as follows: determine the core knowledge points of the candidate knowledge group, calculate the teaching load of all core knowledge points, and sum them to obtain the teaching load of the candidate knowledge group. Among them, if a knowledge point is required by the learning objective, or if its structural salience or material support is greater than or equal to the corresponding threshold, then the knowledge point is a core knowledge point; the teaching load is the weighted sum of the structural salience, knowledge type complexity and material support of the knowledge point.

9. The method for reorganizing textbook knowledge points based on problem anchors and knowledge graphs as described in claim 1, characterized in that, In S1, knowledge points, knowledge layer relationships, and source evidence are extracted from textbook units to construct a basic knowledge point map; The basic knowledge point map includes knowledge point nodes, textbook unit nodes, knowledge layer relationships between knowledge points, and the source relationships between knowledge points and textbook units; Knowledge layer relationships include prerequisite relationships, composition relationships, co-occurrence relationships, and similarity relationships.

Images