A knowledge graph-based teaching resource database construction method and system

Abstract

The application relates to a knowledge graph teaching resource database construction method and system, and belongs to the technical field of data resource processing, which comprises the following steps: step S1: multi-source heterogeneous teaching resource acquisition and preprocessing: natural language processing is performed on unstructured text to identify the association of entities and key concepts, and user behavior logs are analyzed to extract implicit knowledge point association; step S2: multi-dimensional teaching knowledge graph mode layer definition: a multi-dimensional graph mode layer is constructed; step S3: collaborative extraction and fusion of graph entities and relationships: core teaching entities and specific relationship types are collaboratively extracted from the preprocessed teaching resources to construct a knowledge graph; and step S4: adaptive graph updating and teaching path generation: the constructed knowledge graph is used as a bottom-layer database to respond to user knowledge state queries; the application has the beneficial effect that massive, heterogeneous teaching resources and a knowledge graph are effectively fused.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data resource processing, and particularly relates to a knowledge graph-based teaching resource database construction method and system. BACKGROUND

[0002] With the rapid development of educational informatization, teaching resources are showing an explosive growth trend, covering various types such as courseware, test questions, videos and literature. However, there are many problems in the construction and application of current teaching resource databases. On the one hand, teaching resources are scattered in different platforms or systems, and there is a lack of effective association between resources, forming an "information island", making it difficult for teachers and students to quickly and accurately obtain comprehensive resources related to specific knowledge points; on the other hand, the classification of existing databases on teaching resources is mostly based on simple attribute tags such as subject and grade, which cannot deeply mine the knowledge logic relationship behind the resources, and cannot meet the needs of personalized teaching and intelligent learning.

[0003] As a technology that can effectively organize, manage and display knowledge, knowledge graph technology has strong knowledge association and semantic expression ability. Applying knowledge graph technology to teaching resource database construction is expected to solve the above problems. However, at present, how to effectively integrate massive and heterogeneous teaching resources with knowledge graphs to realize accurate association and intelligent retrieval of teaching resources has not yet formed a mature and efficient construction method and system. Therefore, there is an urgent need for a knowledge graph-based teaching resource database construction method to improve the utilization efficiency and service quality of teaching resources. SUMMARY

[0004] The application provides a knowledge graph-based teaching resource database construction method and system, which is used to solve the technical problem of how to effectively integrate massive and heterogeneous teaching resources with knowledge graphs, and to construct a knowledge graph by cooperatively extracting core teaching entities and specific relationship types from preprocessed teaching resources, thereby realizing the effective integration of massive and heterogeneous teaching resources with knowledge graphs.

[0005] In order to achieve the above purpose, the application realizes the following technical scheme:

[0006] A knowledge graph-based teaching resource database construction method, comprising the following steps:

[0007] Step S1: Multi-source heterogeneous teaching resource acquisition and preprocessing: obtaining original teaching resources from at least two different types of sources, the original teaching resources including structured course outlines, unstructured electronic teaching materials and user behavior logs of online teaching platforms; preprocessing the original teaching resources, including natural language processing of unstructured text to identify the association of entities and key concepts, and analyzing user behavior logs to extract implicit knowledge point associations;

[0008] Step S2: Multi-dimensional teaching knowledge graph schema layer definition: a multi-dimensional schema layer is constructed, which not only defines the core teaching entity types and their hierarchical relationships, but also defines specific relationship types for describing teaching logic, including at least "prerequisite relationship", "teaching depth increasing relationship" and "common co-occurrence misunderstanding relationship";

[0009] Step S3: collaborative extraction and fusion of graph entities and relationships: based on the schema layer, core teaching entities and specific relationship types are collaboratively extracted from the pre-processed teaching resources to construct a knowledge graph; among them, the "prerequisite relationship" and "teaching depth increasing relationship" between entities are extracted and verified by combining the analysis of the structured data of the course outline and the semantic analysis results of the teaching materials content; and the "common co-occurrence misunderstanding relationship" is automatically discovered and generated by analyzing the answer sequences and error patterns in the massive user behavior logs through a machine learning model;

[0010] Step S4: adaptive graph updating and teaching path generation: the constructed knowledge graph is used as a bottom database to respond to user knowledge state queries; according to the user's knowledge mastery, a personalized optimal learning path spanning multiple knowledge entities is dynamically generated by real-time traversal of the knowledge graph using graph algorithms, and each node in the learning path is associated with the most suitable teaching resource recommendation to the user, and the user's behavior feedback on the new learning path is used as incremental data.

[0011] Optionally, in step S1, the natural language processing of unstructured text to identify the association of entities and key concepts is the logical relationship of the association of entities / concepts in the teaching text, which needs to be extracted by a relationship classification model for calculating the relationship of the text, and the specific formula is as follows:

[0012] The association of the text belongs to the relationship feature :

[0013] ;

[0014] Wherein, represents the probability of the given text / entity belonging to the relationship feature when the relationship feature belongs to the relationship category set ;

[0015] BERT embedding vector of text , is the BERT embedding vector of text ​​​The vector representation of the text is encoded by the BERT pre-trained model and is the model's vectorized capture of the text semantics. Relational features The corresponding weight vector is used to measure relation features. With text The degree of matching; Relational features The bias term; It is an exponential function used to convert the matching degree into a non-negative value and amplify the numerical difference, making high matching degrees stand out more.

[0016] For each relation feature With the given text / entity The correlation is calculated, specifically, The vector dot product measures the characteristics of the relationship. Weight vector and text The similarity of BERT embedding vectors, plus the bias This is to allow the model to adjust relational features more flexibly. Exponential function It is monotonically increasing; the larger the exponent, the stronger the relationship characteristics. With text The higher the matching degree;

[0017] The normalization term ensures that the sum of the probabilities of all relational features equals 1, satisfying the fundamental property of probability; specifically, This refers to traversing relation features, which belong to the set of relation categories. Each relationship in; Relational features The corresponding weight vector is used to measure the traversal relationship features. With text The degree of matching; A vector representation of the text; Features of traversal relations The bias term is used to adjust the probability offset and is also a parameter learned by the model. It is a set of relation categories All traversal relation features Summing the results;

[0018] First, calculate the matching index of a single relation to ensure the result is non-negative, amplifying the difference between high-matching and low-matching relations. Then, sum the index results of all relations to form the relation category set. All relational features The exponential results are all added together, and the numerator is a certain relationship. The denominator is all relationship features The sum of the exponential results of the numerator and denominator can convert the text belonging to each relationship into relevance .

[0019] Optionally, in step S1, the implicit knowledge point association is extracted by calculating the cosine similarity between the implicit knowledge point vector and the knowledge point concept vector to determine the association degree:

[0020] ;

[0021] Among them, The similarity value ranges from -1 to 1. The closer the value is to 1, the more relevant the association degree is. The closer the value is to -1, the less relevant the association degree is.

[0022] is the knowledge point concept; is the implicit knowledge point; is the text embedding vector of the knowledge point concept; is the feature vector of the implicit knowledge point; and are the vectors converted from the features of the objects;

[0023] is the dot product of the two vectors, which is calculated by multiplying the corresponding dimension components and summing them up, reflecting the degree of coincidence of the vectors in the direction;

[0024] and are the lengths of the text embedding vector of the knowledge point concept and the feature vector of the implicit knowledge point, and are used to normalize the vectors to eliminate the influence of length difference;

[0025] is used to measure the association degree between the two vectors and . The cosine similarity measures the association degree by the cosine value of the vector angle.

[0026] Optionally, in step S2, the multi-dimensional graph pattern layer is a teaching activity sequence from shallow to deep, which is automatically filtered to find the optimal combination sequence based on the "cognitive ladder" and "activity effect feedback":

[0027] ;

[0028] Among them, represents the total return or total value of the optimal sequence of depth, and the sequence that can maximize the total return is found , For all teaching activity sequences, i.e. Therefore, use to represent, to represent the maximum independent variable, is the maximum sequence of the sum of cognitive ladder and effect feedback weighted sum in all sequences;

[0029] is the first teaching activity corresponding cognitive ladder level; is the effect feedback value from to ; is the effect feedback weight coefficient; is the value of a single action, the action in the sequence; It itself measures the value of a single action itself, and the sum is the sum of the self-value of all actions in the sequence, obtaining the total value of single-step;

[0030] is the value of the transition between actions, is the value of the transition between actions to ;

[0031] is the value of the transition between actions, there are adjacent action pairs in the sequence, and the sum of the values of all adjacent transitions;

[0032] The automatic screening of the optimal combination sequence is to find an action sequence , so that the sum of the self-value of all single-step actions and the sum of the value of all adjacent action transitions multiplied by the weight , the total sum is the largest, the screened sequence not only ensures that each action itself is good enough, but also ensures that the connection between actions is reasonable enough, so as to become the optimal sequence.

[0033] Optionally, in step S3, for the preprocessed teaching resources, the core teaching entities and specific relationship types are extracted in coordination to construct a knowledge graph, which is a process of constructing a knowledge graph from unordered resources to ordered structures, so as to quantify the dynamic cooperative interaction strength of the coordinated extraction , the dynamic cooperative effect of different extraction stages is quantified through interactive response, which reflects the process value of cooperation:

[0034] ;

[0035] Among them, is Dynamic synergy interaction intensity at time t, representing the overall intensity of multiple subjects interacting with each other at time t;

[0036] For the subject With Synergy correlation factor or interaction trigger coefficient; if And There is a synergistic relationship at Time, ; otherwise 0; For the subject At The state change at time t, the fluctuation of activity and contribution over time;

[0037] To avoid the denominator being 0, a small amount Is introduced to ensure calculability;

[0038] Time decay term, Is the decay rate; Is the absolute value of the time difference between the subject And Interaction, reflecting the rule that the longer the interaction, the weaker the contribution to the current synergy strength;

[0039] Double summation, representing the traversal calculation of all subjects on , And Without repeating, and then accumulating the results, There are Actions in the sequence.

[0040] Optionally, the ordered structure evolves into a knowledge graph that is dynamically evolving. Traditional order degree only evaluates static results, ignoring the long-term evolution of collaborative extraction on knowledge graph. By quantifying the long-term value of collaborative extraction on the ordered structure of the knowledge graph, the dynamic order is reflected:

[0041] ;

[0042] Among them, Indicates the change in free energy during the evolution of the ordered structure; Summation from the initial time To the end time , reflecting the cumulative effect of the whole process;

[0043] The order degree at time Under the synergistic effect, synergy is the behavior of internal elements cooperating and promoting order;

[0044] is a time weight factor or discount factor, is a rate constant or decay coefficient, which controls the weight decay speed of the long-term effect.

[0045] is a time weight factor or discount factor, is a rate constant or decay coefficient, which controls the weight decay speed of the long-term effect.

[0046] Optionally, the extraction and verification of the "prerequisite relationship" and "teaching depth increasing relationship" between entities are achieved by the method of "structured data anchoring + semantic analysis completion":

[0047] Step a: preliminary anchoring of the relationship from the course outline structured data;

[0048] Step b: verify the validity of the relationship by combining the semantic analysis of the teaching material content;

[0049] Step c: cross-verification and relationship correction.

[0050] Optionally, in step S4, in step S4, follow the path of "perceiving user state → generating personalized scheme → providing resource support → absorbing feedback iteration".

[0051] Optionally, the specific steps are as follows:

[0052] Step I: Role positioning of the underlying database of the knowledge graph;

[0053] Step II: Generation of personalized optimal learning path based on user knowledge state;

[0054] Step III: Association recommendation of learning path nodes and adaptive teaching resources.

[0055] A knowledge graph-based teaching resource database construction system, comprising:

[0056] a plurality of source heterogeneous teaching resource processing modules connected in sequence, a multi-dimensional teaching knowledge graph pattern layer definition module, a graph entity and relationship collaborative processing module, and a self-adaptive graph application module;

[0057] The multi-source heterogeneous teaching resource processing module is configured to obtain original teaching resources from at least two different types of sources, the original teaching resources including structured course outlines, unstructured electronic teaching materials, and user behavior logs of online teaching platforms, and to preprocess the original teaching resources, the preprocessing operation including performing natural language processing on unstructured text to identify the association of entities and key concepts, and analyzing the user behavior logs to extract implicit knowledge point associations;

[0058] The multi-dimensional teaching knowledge graph schema layer definition module is configured to build a multi-dimensional graph schema layer, which defines not only core teaching entity types and hierarchical relationships thereof, but also specific relationship types for describing teaching logic, the specific relationship types at least including a prerequisite relationship, a teaching depth increasing relationship and a common co-occurrence misunderstanding relationship;

[0059] The graph entity and relationship collaborative processing module is configured to collaboratively extract core teaching entities and specific relationship types from the pre-processed teaching resources based on the graph schema layer to build a knowledge graph, wherein the prerequisite relationship and the teaching depth increasing relationship between entities are extracted and verified by combining analysis of structured data of a course outline and semantic analysis results of teaching material content, and the common co-occurrence misunderstanding relationship is automatically discovered and generated by a machine learning model analyzing answer sequences and error patterns in massive user behavior logs;

[0060] The adaptive graph application module is configured to take the built knowledge graph as a bottom layer database, respond to user knowledge state queries, and according to the user's knowledge mastery, utilize graph algorithms to traverse the knowledge graph in real time, dynamically generate an individualized optimal learning path spanning multiple knowledge entities, associate each node in the learning path with the most suitable teaching resource recommendation to the user, and take the user's behavior feedback on the new learning path as incremental data.

[0061] The beneficial effects of the present application are as follows:

[0062] The present application is based on a graph schema layer, collaboratively extracts core teaching entities and specific relationship types from pre-processed teaching resources to build a knowledge graph; specifically, the prerequisite relationship and the teaching depth increasing relationship between entities are extracted and verified by combining analysis of structured data of a course outline and semantic analysis results of teaching material content; and the common co-occurrence misunderstanding relationship is automatically discovered and generated by a machine learning model analyzing answer sequences and error patterns in massive user behavior logs, realizing effective fusion of massive, heterogeneous teaching resources and knowledge graphs. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0064] Figure 1 The figure is a schematic diagram of the system structure of the present application;

[0065] Figure 2 This is a flowchart of the process of the present invention. Detailed Implementation

[0066] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0067] Example 1;

[0068] like Figure 1 As shown, this embodiment provides a knowledge graph-based teaching resource database construction system, including:

[0069] The module consists of a multi-source heterogeneous teaching resource processing module, a multi-dimensional teaching knowledge graph pattern layer definition module, a graph entity and relation collaborative processing module, and an adaptive graph application module, which are connected in sequence.

[0070] The multi-source heterogeneous teaching resource processing module is configured to obtain original teaching resources from at least two different types of sources. The original teaching resources include structured course outlines, unstructured electronic textbooks, and user behavior logs from online teaching platforms. The module also preprocesses the original teaching resources. The preprocessing operations include performing natural language processing on unstructured text to identify the association between entities and key concepts, and parsing user behavior logs to extract implicit knowledge point associations.

[0071] The multidimensional teaching knowledge graph pattern layer definition module is configured to construct a multidimensional graph pattern layer. The graph pattern layer defines both the core teaching entity types and their hierarchical relationships, as well as the specific relationship types used to describe the teaching logic. The specific relationship types include at least "prerequisite relationship", "increasing teaching depth relationship" and "common co-occurrence misunderstanding relationship".

[0072] The knowledge graph entity and relation collaborative processing module is configured to extract core teaching entities and specific relation types from preprocessed teaching resources based on the knowledge graph pattern layer to construct a knowledge graph. Among them, the "prerequisite relation" and "increasing teaching depth relation" between entities are extracted and verified by combining the analysis of the structured data of the course syllabus with the semantic analysis results of the textbook content. The "common co-occurrence misunderstanding relation" is automatically discovered and generated by analyzing the answer sequences and error patterns in massive user behavior logs through machine learning models.

[0073] The adaptive knowledge graph application module is configured to use the constructed knowledge graph as the underlying database, respond to users' knowledge status queries, and dynamically generate a personalized optimal learning path that spans multiple knowledge entities by using graph algorithms to traverse the knowledge graph in real time based on the user's knowledge mastery. At the same time, it recommends the most suitable teaching resources to the user by associating each node in the learning path with the most suitable teaching resources, and uses the user's behavior feedback on the new learning path as incremental data.

[0074] Example 2;

[0075] As Figure 2 shown, the embodiment provides a knowledge graph-based teaching resource database construction method, comprising the following steps:

[0076] Step S1: Multi-source heterogeneous teaching resource acquisition and preprocessing: obtaining original teaching resources from at least two different types of sources, the original teaching resources including structured course syllabus, unstructured electronic teaching materials and user behavior logs of online teaching platforms; preprocessing the original teaching resources, including natural language processing of unstructured text to identify the association of entities and key concepts, and parsing user behavior logs to extract implicit knowledge point associations;

[0077] Step S2: Definition of multi-dimensional teaching knowledge graph pattern layer: constructing a multi-dimensional graph pattern layer, which not only defines the core teaching entity types and their hierarchical relationships, but also defines specific relationship types for describing teaching logic, including at least "prerequisite relationship", "teaching depth increasing relationship" and "common co-occurrence misunderstanding relationship";

[0078] Step S3: Collaborative extraction and fusion of graph entities and relationships: based on the graph pattern layer, the core teaching entities and specific relationship types are collaboratively extracted from the preprocessed teaching resources to construct the knowledge graph; among them, the "prerequisite relationship" and "teaching depth increasing relationship" between entities are extracted and verified by combining the analysis of the structured data of the course syllabus and the semantic analysis results of the teaching materials content; while the "common co-occurrence misunderstanding relationship" is automatically discovered and generated by analyzing the answer sequences and error patterns in the massive user behavior logs through a machine learning model;

[0079] Step S4: Adaptive graph updating and teaching path generation: taking the constructed knowledge graph as the underlying database, responding to the user's knowledge state query; according to the user's knowledge mastery, using graph algorithms to traverse the knowledge graph in real time, dynamically generating an individualized optimal learning path spanning multiple knowledge entities, and recommending the most suitable teaching resources for each node in the learning path to the user, and using the user's behavior feedback on the new learning path as incremental data.

[0080] Embodiment 3;

[0081] Based on embodiment 2, in step S1, the natural language processing of unstructured text to identify the association of entities and key concepts is the logical relationship of entity / concept association in teaching text, which needs to be extracted through a relationship classification model to calculate the relationship of the text, and the specific formula is as follows:

[0082] Let the text segment to be analyzed be (for example: calculating the side length of a right triangle), and the relationship category set be , the text is calculated by a pre-trained model (i.e. BERT model) belongs to the relationship feature of relevance :

[0083] ;

[0084] wherein represents the probability that the given text / entity belongs to the relationship feature when (for example: the probability that the text in the text classification belongs to a certain category and the two entities in the knowledge graph have a relationship feature ). is the relationship feature belongs to the relationship category set (the relationship category set is all possible categories / relationship types, such as belonging to the same category);

[0085] is the BERT embedding vector of the text , and is the vector representation of the text , which is encoded by the BERT pre-training model, which is the vectorization capture of the text semantic by the model; is the weight vector corresponding to the relationship feature (the parameters learned by the model for each relationship), which is used to measure the matching degree of the relationship feature and the text ; is the bias term of the relationship feature (adjusting the offset of the probability, which is also a parameter learned by the model); is an exponential function, which is used to convert the "matching degree" into a non-negative value, and to amplify the value difference, so that the high matching degree is more prominent;

[0086] is the relevance of each relationship feature to the given text / entity , which is calculated as follows, is the vector dot product, which measures the similarity of the weight vector of the relationship feature and the BERT embedding vector of the text (the dot product is greater, the semantic matching degree is higher), plus the bias , in order to make the model more flexible to adjust the relationship feature ; the exponential function is monotonically increasing, and the larger the exponential result is, the higher the matching degree of the relationship feature and the text ;

[0087] ​is a normalization term, which makes the probability of all relationship features sum to 1, satisfying the basic property of probability; specifically, is the traversal relationship feature, which belongs to each relationship in the relationship category set ; is the relationship feature corresponding weight vector, which is used to measure the matching degree of the traversal relationship feature and the text ; is the vector representation of the text (or entity) (encoded by the BERT pre-training model, which is the vectorization result of the text semantics); is the bias term of the traversal relationship feature , which is used to adjust the offset of the probability, and is also a parameter learned by the model, is the sum of the results of all traversal relationship features in the relationship category set ;

[0088] First, calculate the matching index of a single relationship to ensure that the result is non-negative (because the output of the exponential function is always greater than 0), and amplify the difference between high and low matching degree relationships (the exponential function is monotonically increasing, the larger the exponential result, the higher the matching degree), then sum the exponential results of all relationships, that is, add up all the exponential results of the relationship features in the relationship category set ; The numerator is the exponential result of a certain relationship , and the denominator is the sum of the exponential results of all relationship features , and by "numerator ÷ denominator", "text belongs to each relationship" can be transformed into relevance (satisfying "non-negative" and "sum to 1").

[0089] From the original teaching text to the structured entity / concept, the whole process is as follows:

[0090] Original text Tokenization result Text vector Entity recognition Key concept relationship Semantic relationship, covering the whole link of "preprocessing- feature- semantic extraction" of teaching text NLP, which realizes the structuring of teaching resources (such as knowledge point library construction, intelligent retrieval).

[0091] Extracting implicit knowledge point association is to calculate the cosine similarity between the implicit knowledge point vector and the knowledge point concept vector to determine the degree of association:

[0092] ; ​

[0093] where, is the similarity value, ranging from [−1, 1], the closer the value to 1, the more relevant the correlation, the closer the value to -1, the less relevant the correlation;

[0094] is the knowledge point concept; is the implicit knowledge point; is the text embedding vector of the knowledge point concept (encoded by the BERT model on the concept name + definition); is the feature vector of the implicit knowledge point; and is the vector form of the object's features, such as: text word frequency vector, item attribute vector;

[0095] is the dot product (inner product) of the two vectors, calculated by multiplying the corresponding dimension components and summing them up, reflecting the degree of coincidence of the vectors in the direction;

[0096] and are the length of the text embedding vector of the knowledge point concept, the length of the feature vector of the implicit knowledge point, and are used to normalize the vector and eliminate the influence of length difference;

[0097] is used to measure the correlation (similarity) between two vectors and Cosine similarity measures the correlation by the cosine value of the vector angle:

[0098] If the directions of the two vectors are exactly the same (the angle is 0°), at this time , indicates and high similarity; if the directions of the two vectors are exactly opposite (the angle is 180°), at this time , indicates and high dissimilarity; if the two vectors are perpendicular (the angle is 90°), at this time , indicates and no similarity.

[0099] Used in the field of text similarity analysis (such as: the similarity of the themes of two articles, calculated by word frequency vector), recommendation system (such as: similarity matching of item attribute vectors), image feature matching, the core is to measure the correlation by "direction similarity" rather than "numerical size".

[0100] Embodiment 4;

[0101] Based on Embodiment 2, in step S2, the knowledge graph is divided into Schema Layer and Data Layer:

[0102] The Data Layer is specific “teaching knowledge instances” (such as: “one-variable linear equation”, “Newton’s first law”, “subject judgment of ‘ba’ sentence”), as well as specific associations between instances (such as: “one-variable linear equation is prerequisite knowledge of two-variable linear equation”);

[0103] The Schema Layer is a higher-dimensional “abstract framework”, which does not focus on specific knowledge points, but defines in advance “which types of things can become core entities in teaching”, “how these entities are layered”, and “which relationships can be used to describe teaching logic”, which is equivalent to establishing “uniform standards” for subsequent filling of specific teaching content.

[0104] And multi-dimension emphasizes that the design of the Schema Layer needs to cover multiple key dimensions of the teaching process (such as: knowledge progression dimension, cognitive difficulty dimension, and error diagnosis dimension), and not focus on a single “knowledge point classification”.

[0105] The core task of the Schema Layer of the graph is to “define core teaching entity types and hierarchical relationships” and “define specific relationship types that describe teaching logic”, both of which jointly support the knowledge organization of the teaching scenario.

[0106] Core teaching entity types refer to the most critical and basic knowledge carrier categories in the teaching scenario, rather than specific knowledge points; hierarchical relationships are the containing / subordinate / abstract-concrete relationships between these types, which ensure the structuring of the knowledge system;

[0107] The specific relationship types of teaching logic are that entity types are the framework, and specific relationship types give the framework life, and these relationships must fit the core logic of teaching (such as: knowledge progression, difficulty escalation, and error association), rather than the general “association”, “containment”, and at least three types of relationships, as follows:

[0108] Prerequisite-Relation is used to solve the logic of "what to learn first, what to learn next", describes the dependency between two knowledge points (or knowledge modules) "must master A first, then learn B", is the core basis of teaching sequence design (corresponding to "the order of knowledge"), the relationship direction is one-way (A→B, A is the prerequisite of B, B depends on A); avoid "teaching ahead" (such as: don't teach "polynomial addition and subtraction" directly teach "one-variable linear equation", students can't understand the transfer), auxiliary to develop individualized learning path (such as: students can't learn "two-variable linear equation", can trace back whether the prerequisite knowledge "one-variable linear equation" is mastered).

[0109] Depth-Increment-Relation is used to solve the logic of "how deep to learn, in how many steps", describes the "from shallow to deep" progressive relationship of the same knowledge point (or the same topic knowledge) in "cognitive difficulty, application range, examination requirement", corresponding to the gradual improvement of students' cognitive ability (different from the "different knowledge order" of "prerequisite relationship", this relationship is "depth layering of the same knowledge"). The relationship direction is one-way (shallow A→deep B), avoiding "one-size-fits-all" teaching (such as: teaching "shallow application" to students with weak foundation first, teaching "deep analysis" to high achievers); auxiliary design layered homework, stage test (such as: unit test focuses on "middle layer understanding", final test focuses on "deep layer application").

[0110] Co-Occurrence-Misconception-Relation is used to solve the logic of "which errors often occur together", describes the association between two or more "common misconceptions" that "highly co-occur", that is, if students have misconception A, they are likely to have misconception B as well. The essence is the "associativity of error cognition" (derived from the internal relationship between knowledge points or students' common cognitive bias), the relationship direction is bidirectional or multidirectional (A↔B, A and B co-occur).

[0111] The multi-dimensional graph mode layer is a sequence of teaching activities from shallow to deep, which is based on "cognitive ladder" and "activity effect feedback", and automatically selects the optimal combination sequence:

[0112] ;

[0113] Among them, represents the total return or total value of the depth-optimal sequence, and the sequence that maximizes the total return needs to be found , is all teaching activity sequences, that is , so represents represents the independent variable that maximizes Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences;

[0114] Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences;

[0115] Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences;

[0116] Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences;

[0117] Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences;

[0118] Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences;

[0119] Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences;

[0120] Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences; Select the sequence with the maximum sum of cognitive ladder sum + effect feedback weighted sum in all sequences;

[0121] The overall goal of automatically selecting the optimal combination sequence is to find a sequence of actions. This is such that the sum of the individual values ​​of all single-step actions and the sum of the values ​​of all adjacent action transitions, multiplied by a weight. The sequence that yields the largest sum is the optimal sequence, ensuring that each action is good enough and that the connections between actions are reasonable enough.

[0122] Example 5;

[0123] Based on Example 2, in step S3, the core teaching entities and specific relationship types are collaboratively extracted from the preprocessed teaching resources to construct a knowledge graph. This process of constructing the knowledge graph is a transformation from disordered resources to an ordered structure, thereby quantifying the dynamic collaborative interaction intensity of the collaborative extraction. By quantifying the dynamic synergistic effects of different extraction stages (such as entity recognition stage and relation classification stage) through interactive responses, the value of the synergistic process is demonstrated.

[0124] ;

[0125] in, for The dynamic collaborative interaction intensity at time t represents the overall intensity of the collaborative effect of multiple entities at time t.

[0126] as the main body and Synergistic correlation factors or interaction triggering coefficients; if and exist There is always a collaborative relationship (such as cooperation and information transmission). Otherwise, it is 0;

[0127] as the main body exist The change in state at any given moment represents the fluctuation of activity and contribution over time.

[0128] To avoid the denominator being 0, a minimum quantity is introduced. This ensures computability;

[0129] For time decay term, It is the attenuation rate ( >0, the larger the value, the faster the effect of the time difference decays. It is the main body and The absolute value of the time difference between interactions reflects the principle that the longer the interaction takes place, the weaker its contribution to the current level of collaboration.

[0130] For double summation, represents the traversal calculation of all pairs of subjects ( And No repetition) is added, For sequence has Actions;

[0131] Calculate the dynamic synergy contribution of each pair of subjects, and then get the overall synergy strength by summation, for each pair of subjects , the relative activity of the subject is measured (because of the addition , the denominator is meaningless when time is divided) ; whether the pair of subjects has a current synergy relationship is determined by , and the influence of time difference is attenuated by , which reflects that recent interaction has a greater impact on current synergy than remote interaction. Adding the synergy contribution of all pairs of subjects, we get the total strength of dynamic synergy interaction at ;

[0132] The information propagation of the node is used to measure the dynamic strength of information synergy interaction between nodes in the network, optimize information flow, and quantify the dynamic synergy degree of multiple subjects from three dimensions of "individual state change", "interaction association" and "time decay"; The overall strength of real-time dynamic synergy of multiple subjects not only considers which subjects are synergized, but also considers how much each subject contributes, and whether the interaction is new enough in time. Finally, through double summation, the synergy degree of the system is obtained.

[0133] Further, the ordered structure evolves into a knowledge graph that is dynamically evolving (e.g., new data will supplement triples, and errors will be corrected), and the traditional order degree only evaluates static results, ignoring the long-term evolution of the knowledge graph (e.g., high-consistency triples extracted by synergy extraction can reduce conflicts and accelerate fusion when new data is added later). By quantifying the long-term value of the continuous optimization of the ordered structure of the knowledge graph by synergy extraction, the dynamic order is reflected:

[0134] ;

[0135] Among them, The change in free energy during the evolution of ordered structure, i.e., free energy is the core quantity in thermodynamics to measure available energy, and structure evolution is usually accompanied by energy change;

[0136] is the time from the initial time to the end time ​​Summation is performed to reflect the cumulative effect of the entire process; For a moment The degree of orderliness under synergy; synergy refers to the behavior of internal elements cooperating with each other to promote order.

[0137] For a moment The degree of order under independent action, that is, the degree of order when elements have no cooperation and each behaves randomly, is a reference benchmark for disorder or weak order.

[0138] It is a time-weighted factor or a discount factor. It is a rate constant (or decay coefficient) that controls the rate of decay of the weighted effect in the long term; The larger the scale, the more emphasis is placed on the near term, and the faster the long-term contribution diminishes. Indicates from time until the end time The remaining time, the longer the remaining time is The smaller the value, the closer it is to the initial moment. The larger, The smaller the value, the less the contribution of early actions to the final free energy is discounted; conversely, The closer , The smaller, The closer the value is to 1, the more attention is paid to the contribution of late-stage behavior.

[0139] First calculate each time step The difference between cooperative order and independent order. This reflects the difference between collaboration and independence. The degree of orderliness of contributions made at any given moment; then multiplied by the time weight. This reflects the differences in the weight of contributions at different times in the final free energy; that is, the early contributions are weakened by time decay, while the later contributions are more prominent, across the entire time range. Summing these values ​​yields the total change in free energy resulting from the synergistic effect throughout the entire evolutionary process.

[0140] From a dynamic evolutionary perspective, it breaks through the limitations of static evaluation, linking collaborative optimization with the long-term viability of the knowledge graph. For example, high-quality triples extracted collaboratively can reduce conflicts in subsequent data fusion and lower maintenance costs; the time discount term aligns with practical application scenarios (e.g., knowledge graphs focus more on the orderliness of recent data, as it directly affects current decisions; the value range is...). , The larger the value, the stronger the benefit of collaborative extraction to the long-term orderly evolution of the knowledge graph, and the more persistent the optimization effect. (Ordered structure evolution) Evaluation of the long-term value optimization of collaborative extraction.

[0141] The extraction and verification of "prerequisite relationship" and "teaching depth increasing relationship" between entities, which means that a certain teaching entity (such as knowledge point "one-time equation") needs to be based on another entity (such as knowledge point "operation of rational numbers") to effectively master, is achieved through "structured data anchoring + semantic analysis completion":

[0142] Step a: Preliminary anchoring of relationship from course outline structured data;

[0143] As the core document of teaching planning, the course outline contains structured information such as clear teaching module sequence and prerequisite requirements. By analyzing the "prerequisite knowledge" and "preparation module" label fields in the outline, the preliminary prerequisite relationship pairs are directly extracted.

[0144] Step b: Verify the validity of the relationship by combining the semantic analysis of the teaching material content;

[0145] As the core carrier of teaching implementation, the content arrangement and logical connection of the teaching material directly reflect the prerequisite dependency between entities. Through natural language processing technology, the teaching material text is analyzed semantically to verify the prerequisite relationship from three aspects: concept definition, example analysis, and problem association.

[0146] Step c: Cross-verification and relationship correction;

[0147] The prerequisite relationships extracted from the course outline are cross-compared with the relationships verified by the semantic analysis of the teaching material. The consistent relationships are directly confirmed, and the relationships with differences are verified again.

[0148] "Common co-occurrence misunderstanding relationship" easily confuses two or more teaching entities (such as "sufficient condition" and "necessary condition"), leading to incorrect simultaneous association relationships. Through the discovery and generation of a machine learning model based on the deep analysis of massive user behavior logs, the specific process is as follows:

[0149] User behavior logs contain massive unstructured and semi-structured data such as learners' answer records (question ID, answer time, answer correctness, and answer duration), knowledge point association records (question corresponding to the examined knowledge point), and error feedback records (wrong question annotation, error cause description). Data cleaning and feature extraction are required:

[0150] The answer sequence and error pattern in the user behavior log are automatically discovered as follows. First, invalid logs (such as misoperation records with an answer time less than 3 seconds, and repeated submission records) are removed, and valid answer sequences (user answer ID and result in chronological order) are retained. Second, a “question-knowledge point” mapping matrix is constructed, and the core knowledge point corresponding to each question (such as the question “Judge the condition type of ‘if x > 2, then x > 1’” corresponding to the knowledge point “sufficient condition, necessary condition”) is determined through teaching expert annotation and text matching technology. Finally, key features are extracted: “co-occurrence frequency of wrong questions” (the number of times that two knowledge points corresponding to questions are answered incorrectly by the same user within the same answer period), “error interval time” (the time difference between the answers of two knowledge points corresponding to questions by the same user, if the interval is less than 10 minutes, it is considered as short-term co-occurrence), and “error pattern similarity” (through natural language processing to analyze the error description in the user's wrong question feedback, such as “confusion between the definitions of sufficient condition and necessary condition” and “reversing the direction of sufficient condition judgment”, the semantic similarity of the error description is calculated), as well as user attribute features (such as grade, learning progress), providing multi-dimensional feature data for subsequent model training.

[0151] Embodiment 6;

[0152] Based on Embodiment 2, in step S4, the path of “perceiving user state → generating personalized scheme → providing resource support → absorbing feedback iteration” is followed, and the specific steps are as follows:

[0153] Step I: Role positioning of the underlying database of the knowledge graph;

[0154] The constructed knowledge graph (including core teaching entities, “prerequisite relationships”, and “misunderstanding relationships”) is the data basis, and these information is the core basis for subsequent personalized services.

[0155] At the same time, the graph supports real-time query capability: when the system receives a user's knowledge state query request (such as a student completing a post-class test or a teacher querying a student's weak link), it can quickly match the user's current mastered / unmastered knowledge entities from the graph to provide user portraits for subsequent analysis.

[0156] Step II: Generation of personalized optimal learning path based on user knowledge state;

[0157] The goal is to design the most suitable learning order for users with different knowledge levels, avoiding one-size-fits-all teaching schemes. The key points are as follows:

[0158] Input: User knowledge mastery The system needs to obtain the user's knowledge state data, common sources include:

[0159] Explicit data: test scores (e.g., 100% accuracy for knowledge point 1, 30% accuracy for knowledge point 2), homework completion, course progress, etc.

[0160] Implicit data: learning duration (e.g., the time spent on knowledge point 2 is three times that of knowledge point 1), number of hesitations in answering questions (e.g., repeatedly modifying answers to questions on knowledge point 2).

[0161] Core technology: Real-time traversal system using graph algorithms (e.g., shortest path algorithm, weight-based path planning algorithm) in combination with relationship rules in the knowledge graph to dynamically generate learning paths. For example:

[0162] Assuming the user does not currently master knowledge point 2, the graph clearly states that "the prerequisite for knowledge point 2 is knowledge point 1" and "the common misunderstanding of knowledge point 2 is confusion with knowledge point 3";

[0163] The graph algorithm will first determine whether the user has mastered knowledge point 1; if not, the starting point of the path is set to knowledge point 1; if so, it directly points to knowledge point 2, while adding a step to distinguish knowledge point 2 from knowledge point 3 (to avoid misunderstanding);

[0164] The final personalized path may be: learn knowledge point 1 → distinguish knowledge point 2 from knowledge point 3 → learn knowledge point 2 → practice typical examples of knowledge point 2 (rather than directly learning knowledge point 2 to avoid learning difficulties due to insufficient prerequisites or misunderstanding). The "optimality" of the path is not "shortest," but "most suitable for user efficiency and effectiveness."

[0165] Step III: Association and recommendation of learning path nodes and adaptive teaching resources;

[0166] After generating the learning path, the system needs to match "the most suitable teaching resources" for each "node" (i.e., each knowledge point to be learned) in the path to ensure that the user can efficiently master the node knowledge. The core logic is "precise binding of resources and knowledge entities":

[0167] Personalized adaptation: not only matching "knowledge point corresponding resources," but also adjusting based on user characteristics. For example:

[0168] For visual learners, recommend animated explanation videos for knowledge point 2;

[0169] For users with weak foundation, recommend introductory-level examples (rather than difficult problems) for knowledge point 2;

[0170] For users who easily confuse knowledge point 2 and knowledge point 3, additionally recommend comparison and analysis documents for knowledge point 2 and knowledge point 3, i.e., use the "co-occurrence misunderstanding relationship" in the graph to compare and analyze documents.

[0171] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1.A method for constructing a knowledge graph-based teaching resource database, characterized in that, Comprise the following steps: Step S1: Multi-source heterogeneous teaching resource acquisition and preprocessing: obtaining original teaching resources from at least two different types of sources, the original teaching resources including structured course outline, unstructured electronic teaching materials and user behavior logs of online teaching platform; preprocessing the original teaching resources, including natural language processing on unstructured text to identify the association of entities and key concepts, and parsing the user behavior logs to extract implicit knowledge point association; Step S2: Definition of multi-dimensional teaching knowledge graph pattern layer: a multi-dimensional graph pattern layer is constructed, which not only defines the core teaching entity types and their hierarchical relationships, but also defines specific relationship types for describing teaching logic, including at least "prerequisite relationship", "teaching depth increasing relationship" and "common co-occurrence misunderstanding relationship"; In the step S2, the multi-dimensional graph pattern layer is a shallow-to-deep teaching activity sequence, which is automatically filtered based on "cognitive ladder" and "activity effect feedback" to select the optimal combination sequence: ； where, Total return or total value of the depth optimal sequence, find the sequence that maximizes the total return , For all teaching activity sequences, i.e. So use to represent, Indicates the independent variable that maximizes, In all sequences, select the maximum sequence of the total sum of cognitive ladders + effect feedback weighted sum; is the teaching activity corresponding cognitive step level; is the effect feedback value from ; is the effect feedback weight coefficient; is the value of a single action, the action in the sequence; is the value of the action itself, summed is the sum of the values of all actions in the sequence, resulting in the total value of the single step; value for inter-action transition, is a measure of action to transition value; For a sequence of n actions, there are n-1 adjacent action pairs, and the value is summed over all adjacent transitions. For a sequence of n actions, there are n-1 adjacent action pairs, and the value is summed over all adjacent transitions. For a sequence of n actions, there are n-1 adjacent action pairs, and the value is summed over all adjacent transitions The automatic screening of the optimal combination sequence is to find a motion sequence such that the sum of the values of all single-step motions and the sum of the values of all adjacent motion transitions multiplied by the weight are the largest, the screened sequence guarantees that each motion is good enough and the connection between motions is reasonable enough, thereby becoming the optimal sequence; Step S3: Collaborative extraction and fusion of graph entities and relationships: based on the graph pattern layer, core teaching entities and specific relationship types are collaboratively extracted from the preprocessed teaching resources to construct a knowledge graph; among them, the "prerequisite relationship" and "teaching depth increasing relationship" between entities are extracted and verified by combining the structured data of the course outline with the semantic analysis results of the teaching material content; and the "common co-occurrence misunderstanding relationship" is automatically discovered and generated by analyzing the answer sequence and error mode in the massive user behavior logs through a machine learning model; Step S4: Adaptive graph updating and teaching path generation: the constructed knowledge graph is used as a bottom database to respond to user's knowledge state query; according to the user's knowledge mastery, a personalized optimal learning path spanning multiple knowledge entities is dynamically generated by real-time traversal of the knowledge graph using graph algorithms, and each node in the learning path is associated with the most suitable teaching resource recommendation to the user, and the user's behavior feedback on the new learning path is used as incremental data. 2.The knowledge graph-based teaching resource database construction method according to claim 1, characterized in that, In the step S1, the natural language processing on the unstructured text to identify the association of entities and key concepts is the logical relationship of entity / concept in teaching text, which needs to be extracted by a relationship classification model for calculating the relationship of the text, and the specific formula is as follows: Computing text by pre-trained model Belongingness to relationship feature Relevance : ； wherein P (R | T, E) denotes the probability that a given text / entity belongs to a relation characteristic , is the probability that a relation characteristic belongs to a set of relation categories ;​ BERT embedding vector of the text is a vector representation of the text , which is obtained by the BERT pre-training model and is a vectorization capture of the semantic of the text by the model; a relationship feature corresponding weight vector, used to measure the matching degree of the relationship feature and the text ; a bias term of the relationship feature ; an exponential function, used to convert the matching degree into a non-negative numerical value and amplify the numerical difference, so that the high matching degree is more prominent; is calculated for each relation feature with the given text / entity , specifically, is the vector dot product, measuring the similarity between the weight vector of the relation feature and the BERT embedding vector of the text , plus the bias , to make the model more flexible to adjust the relation feature ; the exponential function is monotonically increasing, the larger the exponential result, the higher the matching degree of the relation feature and the text ; The normalization term ensures that the sum of the probabilities of all relational features equals 1, satisfying the fundamental property of probability; specifically, This refers to traversing relation features, which belong to the set of relation categories. Each relationship in; Relational features The corresponding weight vector is used to measure the traversal relationship features. With text The degree of matching; A vector representation of the text; Features of traversal relations The bias term is used to adjust the probability offset and is also a parameter learned by the model. It is a set of relation categories All traversal relation features Summing the results; First, the matching index of a single relationship is calculated to ensure the result is non-negative, to enlarge the gap between high matching degree relationships and low matching degree relationships, and then the sum of the index results of all relationships is calculated, the numerator is the index result of a certain relationship, the denominator is the sum of the index results of all relationship features in the relationship category set , and the correlation between the text and each relationship is converted into relevance by "numerator ÷ denominator" . ​​​ 3.The knowledge graph-based teaching resource database construction method of claim 1, wherein, In the step S1, the extraction of implicit knowledge point association is to calculate the cosine similarity between the implicit knowledge point vector and the knowledge point concept vector to judge the correlation degree: ； wherein, is a similarity value, ranging from [-1, 1], the value closer to 1 means the correlation degree is more relevant, and the value closer to -1 means the correlation degree is less relevant; is a knowledge point concept; is an implicit knowledge point; is a text embedding vector for a knowledge point concept; is a feature vector for an implicit knowledge point; and is a vector form that converts the features of an object into; The dot product of two vectors is a scalar quantity that is calculated by multiplying the corresponding components of the vectors and summing the results. It reflects the degree of alignment between the vectors. and are the lengths of the text embedding vectors of the knowledge point concepts, the lengths of the feature vectors of the implicit knowledge points, and are used to normalize the vectors and eliminate the influence of length difference; For measuring the degree of association between two vectors and Cosine similarity measures the degree of association between two vectors by the cosine of the angle between them. 4.The knowledge graph-based teaching resource database construction method of claim 1, wherein, In the step S3, for the core teaching entity and the specific relationship type in the pre-processed teaching resource, the knowledge graph is constructed as a process from unordered resource to ordered structure to quantify the dynamic collaborative interaction strength of collaborative extraction , the dynamic collaborative effect of different extraction stages is quantified by interactive response, which reflects the process value of collaboration: ； wherein, is the dynamic synergistic interaction intensity at the moment of time t represents the overall intensity of the synergistic interaction of the plurality of subjects at time t; is the subject with the synergy correlation factor or interaction trigger coefficient; if and there is a synergistic relationship at , then ; otherwise, 0; is the subject the state change at , the fluctuation of activity and contribution over time; To avoid a denominator of zero, a small amount is introduced to ensure calculability; is a time decay term, is a decay rate; is a subject with the absolute value of the time difference between the interaction, embodies the law that the more remote in time the interaction, the weaker the contribution to the current synergy strength. For double summation, represents the summation over all pairs of subjects , and without repeating the traversal calculation, the results are accumulated, for the sequence has actions. 5.The knowledge graph-based teaching resource database construction method of claim 4, characterized in that, The ordered structure evolves into a dynamic evolving knowledge graph, and the traditional order degree only evaluates the static result, ignoring the long-term evolution of collaborative extraction on the knowledge graph. By quantifying the long-term value of collaborative extraction on the continuous optimization of the ordered structure of the knowledge graph, the dynamic order is reflected: ； wherein represents the change in free energy during the evolution of the ordered structure; is summed over time from an initial time to a final time to embody the cumulative effect of the entire process; for time order degree under synergy, synergy is the internal elements to coordinate and promote the orderly behavior; for time The order degree under independent action, i.e. the order degree of elements without cooperation and random behavior, is the reference benchmark of disorder or weak order. is a time weighting factor or discount factor, is a rate constant or decay coefficient that controls the rate of weight decay of the long-term effect. 6.The knowledge graph-based teaching resource database construction method of claim 4, characterized in that, The extraction and verification of the "prerequisite relationship" and "teaching depth increasing relationship" between entities is realized by the method of "structured data anchoring + semantic analysis completion": Step a: Preliminary anchoring of relationship from structured data of course outline; Step b: Verify the validity of the relationship in combination with the semantic analysis of the teaching material content; Step c: cross-checking and relationship correction. 7.The knowledge graph-based teaching resource database construction method of claim 1, wherein, In the step S4, the path of "perceiving user state → generating personalized scheme → providing resource support → absorbing feedback iteration" is followed. 8.The knowledge graph-based teaching resource database construction method of claim 7, wherein, The specific steps are as follows: Step I: role positioning of the underlying database of the knowledge graph; Step II: generation of personalized optimal learning path based on user knowledge state; Step III: associated recommendation of learning path nodes and adaptive teaching resources. 9.A knowledge graph-based teaching resource database construction system, configured to perform the knowledge graph-based teaching resource database construction method according to any one of claims 1-8. Comprise: sequentially connected multi-source heterogeneous teaching resource processing module, multi-dimensional teaching knowledge graph mode layer definition module, graph entity and relationship collaborative processing module, and adaptive graph application module; The multi-source heterogeneous teaching resource processing module is configured to obtain original teaching resources from at least two different types of sources, wherein the original teaching resources include structured course outline, unstructured electronic teaching material and user behavior log of online teaching platform, and the original teaching resources are preprocessed, and the preprocessing operation includes performing natural language processing on unstructured text to identify the association of entities and key concepts, and analyzing user behavior log to extract implicit knowledge point association; The multi-dimensional teaching knowledge graph mode layer definition module is configured to build a multi-dimensional graph mode layer, which defines core teaching entity types and hierarchical relationships, and also defines specific relationship types for describing teaching logic, including "prerequisite relationship", "teaching depth increasing relationship" and "common co-occurrence misunderstanding relationship"; The graph entity and relationship collaborative processing module is configured to extract core teaching entities and specific relationship types from the preprocessed teaching resources based on the graph mode layer to build a knowledge graph, wherein the "prerequisite relationship" and "teaching depth increasing relationship" between entities are extracted and verified by combining the structured data of the course outline with the semantic analysis results of the teaching material content, and the "common co-occurrence misunderstanding relationship" is automatically discovered and generated by machine learning model analyzing the answer sequence and error mode in massive user behavior log; The adaptive graph application module is configured to use the built knowledge graph as the underlying database to respond to user's knowledge state query, and according to the user's knowledge mastery, use graph algorithm to traverse the knowledge graph in real time, dynamically generate a personalized optimal learning path spanning multiple knowledge entities, and recommend the most suitable teaching resources for each node in the learning path to the user, and use the user's behavior feedback on the new learning path as incremental data.

Images