Teaching quality evaluation method and system based on large model

Abstract

The invention provides a teaching quality evaluation method and system based on a large model, and the method comprises the steps: obtaining a teacher explanation text, a teaching behavior log and student feedback data, and constructing a triple structured representation; constructing a double-class node directed graph according to the triple structured representation; performing knowledge graph state evolution according to the knowledge point subsets, generating knowledge graph substructures at continuous moments, calculating structural change values of the knowledge graph substructures at adjacent moments, performing regular constraint processing on all the structural change values in combination with a high-risk path set, and generating a teaching behavior causal contribution matrix; wherein each row of the teaching behavior causal contribution matrix represents the structural influence of one teaching behavior on a plurality of knowledge points; and combining the teaching behavior causal contribution matrix with the teaching behavior set and the knowledge point set, inputting a large language model, generating natural language suggestion content, and generating a teaching suggestion according to the natural language suggestion content.

Description

Technical Field

[0001] The present invention belongs to the field of large models, and in particular relates to a teaching quality evaluation method and system based on large models. Background Art

[0002] Currently, teaching quality evaluation, a key feedback and improvement component of the education system, has long relied on manual assessments, standardized tests, and classroom observations. While these methods have provided some degree of oversight and feedback, they are generally subject to high subjectivity, coarse granularity, and poor dynamic responsiveness. Especially with the increasing popularity of digital education, online classrooms, and personalized learning, traditional evaluation methods are no longer able to meet the demands of accurate, real-time, and personalized analysis of teaching processes and outcomes in these new educational contexts. In recent years, artificial intelligence technologies have begun to be applied to educational assessment, such as behavioral analysis systems based on data mining and performance prediction using simple machine learning models. While these methods have achieved some progress in some scenarios, they generally rely on static features and lack the ability to model the underlying mechanisms between teaching behavior and student responses. Consequently, they are unable to accurately reveal the underlying logic linking teaching intent, teaching strategies, and student learning outcomes.

[0003] Furthermore, most existing systems assess teaching quality through an "outcome-oriented" approach, focusing on endpoint data such as student test scores and homework completion, while neglecting to delve deeper into "process information"—that is, the dynamic impact of factors such as interaction methods, the order of knowledge presentation, and classroom feedback on learning outcomes. This prevents teachers from flexibly adjusting their teaching strategies based on real-time student feedback, and makes it difficult to obtain refined, structured optimization recommendations. More importantly, these evaluation mechanisms rely almost entirely on empirical judgment or data correlation analysis, lacking causal reasoning capabilities and unable to explain why a certain teaching method is ineffective for certain types of students or how to precisely intervene in teaching behaviors to improve student performance. With the development of large language model (LLM) technology, although research has explored its applications in areas such as automatic grading and intelligent question-answering, its potential for systematic modeling of teaching quality, causal judgment, and personalized recommendation generation has yet to be fully realized.

[0004] Therefore, how to build a teaching quality evaluation system based on a large model that can understand teaching intentions, model knowledge evolution paths, and reveal the mechanistic connection between teaching behaviors and learning outcomes has become an important technical challenge in the current development of intelligent education. Summary of the Invention

[0005] The purpose of this invention is to propose a teaching quality evaluation method and system based on a large model. The invention further integrates the expression of knowledge structure and the understanding of teaching context, so that the evaluation system has the ability to flexibly adapt to different subjects, different teaching styles and different student groups, greatly improving the intelligence level and promotion and application value of the system.

[0006] In order to achieve the above object, a first aspect of the present invention provides a teaching quality evaluation method based on a large model, the method comprising the following steps:

[0007] Obtaining teacher explanation texts, teaching behavior logs, and student feedback data to construct a triplet structured representation, wherein the triplet structured representation includes teaching behavior, knowledge point set, and student feedback;

[0008] A two-class node directed graph is constructed based on the triple structured representation; the nodes of the two-class node directed graph include a teaching behavior node set and a student feedback status node set, as well as directed edges, representing the student status that may be caused by the teaching behavior. The two-class node directed graph is obtained, and the causal effect is calculated using the knowledge point constraint as the context filtering condition. If the conditional causal effect of the causal path is positive and significant, and the corresponding student status belongs to the negative feedback label set, it is marked as a high-risk path, and the high-risk path set is output;

[0009] Determine, based on the teaching behavior, a subset of knowledge points affected by the teaching behavior, perform knowledge graph state evolution based on the knowledge point subset, generate a knowledge graph substructure at consecutive moments, and calculate the structural change value of the knowledge graph substructure at adjacent moments to quantify the impact of the teaching behavior on the structural state of the knowledge points. Combine all structural change values with a set of high-risk paths for regularization constraint processing to generate a teaching behavior causal contribution matrix, wherein each row of the teaching behavior causal contribution matrix represents the structural impact of a teaching behavior on multiple knowledge points;

[0010] The teaching behavior causal contribution matrix is combined with the teaching behavior set and the knowledge point set, and input into a large language model to generate natural language suggestion content, and then the teaching suggestion is generated using the natural language suggestion content.

[0011] Furthermore, the acquisition of teacher explanation texts, teaching behavior logs, and student feedback data to construct a triple structured representation specifically includes:

[0012] Using the large model, the teacher's explanation text, teaching behavior log and student feedback data are extracted and embedded to generate teacher semantic embedding, teaching behavior embedding and student feedback embedding;

[0013] A structural regularization term is introduced to process the teacher semantic embedding, teaching behavior embedding, and student feedback embedding; the structural regularization term is used to ensure that the teacher's explanation, teaching behavior, and feedback maintain content semantic consistency, and to encourage the semantic direction between teaching behavior and student feedback to maintain differences, thereby avoiding homogenization;

[0014] The teaching behavior label set mapped from the teaching behavior embedding is used as the teaching behavior, the knowledge point set identified by comparing the teacher semantic embedding with the standard knowledge base is used as the knowledge point set, and the student mastery status label inferred from the student feedback embedding is used as the student feedback;

[0015] The teaching behavior, knowledge point set and student feedback are structured as triples.

[0016] Furthermore, the calculation of causal effects using knowledge point constraints as context filtering conditions specifically includes:

[0017] Obtain the set of knowledge points involved in the current teaching to mark the contextual consistency of behavior and feedback;

[0018] Determining, based on the knowledge point set involved in the current teaching, the first conditional expectation of the student state when implementing the behavioral teaching behavior;

[0019] Based on the set of knowledge points involved in the current teaching, the second conditional expectation of the student state when no behavioral teaching behavior is implemented;

[0020] The difference between the first conditional expectation and the second conditional expectation is taken as the intervention effect under the contextual condition and the causal effect.

[0021] Furthermore, the knowledge point subset corresponding to the teaching behavior affected by the teaching behavior is determined based on the teaching behavior, and is obtained through semantic matching of behavior and knowledge points:

[0022] The knowledge point subset is obtained by calculating the similarity of the embedding vector of the large model, that is, matching the behavior semantics with the knowledge point description and selecting the knowledge points with close semantics.

[0023] Furthermore, the step of performing knowledge graph state evolution according to the knowledge point subset to generate a continuous-time knowledge graph substructure specifically includes:

[0024] The knowledge point set involved in the current teaching content is the knowledge point node set; each knowledge point node represents a specific knowledge point;

[0025] Obtain the scores of questions corresponding to the knowledge point node in the formal exam, the scores of questions involving the knowledge point in the homework, and the scores of questions related to the knowledge point in the quiz, and determine the mastery score corresponding to the knowledge point node through weighted summation;

[0026] Add directed edges between knowledge point nodes based on the knowledge point dependencies in the standard syllabus;

[0027] A continuous periodic knowledge graph substructure is constructed based on the knowledge point node set and the directed edge set.

[0028] Furthermore, directed edges are added between knowledge point nodes based on the knowledge point dependency relationship in the standard teaching syllabus. If the first knowledge point node is a prerequisite for the second knowledge point node, a directed edge is added between the first knowledge point node and the second knowledge point node in the graph, indicating that mastering the first knowledge point node is a prerequisite for mastering the second knowledge point node.

[0029] Furthermore, the structural change value of the knowledge graph substructure at adjacent moments is calculated as:

[0030]

[0031] Where, ΔS ij Indicates teaching behavior For knowledge point node k j The change in mastery, i.e., the structural change value; Represents the attention weight, which indicates the teaching behavior on the knowledge point node k j The semantic relevance of v is in the range [0,1]; t (k j ) represents the student's understanding of knowledge point node k at time t j Mastery of v t-1 (k j ) represents the student’s previous knowledge point node k j degree of mastery;

[0032] Among them, if all structural change values are combined with the high-risk path set for regular constraint processing, the causal contribution matrix of teaching behavior is generated, which specifically includes:

[0033] If the current path is a high-risk path, the structural change value is adjusted by the regularization coefficient to adjust the teaching behavior causal contribution matrix;

[0034] If it is not a high-risk path, the structural change value is the causal contribution matrix of the teaching behavior.

[0035] Furthermore, the teaching behavior causal contribution matrix is combined with the teaching behavior set and the knowledge point set, inputted into a large language model, and natural language suggestion content is generated, and teaching suggestions are generated using the natural language suggestion content, specifically including:

[0036] Identify teaching behaviors with significant impact and optimization value based on the teaching behavior causal contribution matrix, and construct a feedback candidate set;

[0037] For each teaching action in the feedback candidate set, a structural semantic joint input vector is constructed: the teaching action, the causal influence sub-vector of the behavior on the knowledge point, the associated important knowledge point set, and the template prompt word;

[0038] Based on the structural semantics and the combined input vector, a large model is used to generate behavioral optimization suggestions;

[0039] Generate standard suggestion entries based on the behavior optimization suggestions, including: teaching behavior, suggestion content, knowledge point set, priority score of feedback candidate set, and execution status field;

[0040] Summarize all the suggested items to form the feedback set for the current round.

[0041] Furthermore, the teaching behaviors with significant impact and optimization value are identified based on the teaching behavior causal contribution matrix, and a feedback candidate set is constructed, specifically:

[0042]

[0043] in, is the priority score, δ is the negative contribution threshold set by the system, Ω(k j ) is the knowledge point k j The importance rating of is the indicator function;

[0044] like γ is the minimum feedback trigger value set, then the teaching action Considered as a high-priority feedback object; ultimately forming a set of recommended behaviors

[0045] In a second aspect of the present invention, a teaching quality evaluation system based on a large model is provided, the system comprising:

[0046] A structuring unit is used to obtain teacher explanation texts, teaching behavior logs, and student feedback data to construct a triple structured representation, which includes teaching behavior, knowledge point set, and student feedback;

[0047] A teaching risk construction set is used to construct a two-class node directed graph based on the triple structured representation; the nodes of the two-class node directed graph include a teaching behavior node set and a student feedback status node set, as well as directed edges, representing the student status that may be caused by the teaching behavior, to obtain a two-class node directed graph, and calculate the causal effect using the knowledge point constraint as the context filtering condition. If the conditional causal effect of the causal path is positive and significant, and the corresponding student status belongs to the negative feedback label set, it is marked as a high-risk path, and the high-risk path set is output;

[0048] A risk evolution set is used to determine, based on the teaching behavior, a subset of knowledge points affected by the teaching behavior, perform knowledge graph state evolution based on the knowledge point subset, generate knowledge graph substructures at consecutive moments, and calculate the structural change values of the knowledge graph substructures at adjacent moments to quantify the impact of the teaching behavior on the structural state of the knowledge points. All structural change values are combined with a set of high-risk paths for regularization constraint processing to generate a teaching behavior causal contribution matrix, wherein each row of the teaching behavior causal contribution matrix represents the structural impact of a teaching behavior on multiple knowledge points;

[0049] A teaching suggestion generation set is used to input the teaching behavior causal contribution matrix, the teaching behavior set, and the knowledge point set into a large language model to generate natural language suggestion content, and generate teaching suggestions using the natural language suggestion content.

[0050] The beneficial technical effects of the present invention are at least as follows:

[0051] In response to the many problems existing in the above-mentioned prior art, the present invention provides a teaching quality evaluation method and system based on a large model, focusing on solving the problems existing in the current teaching evaluation, such as "missing process, unclear cause and effect, non-real-time feedback, and non-optimizable strategy". The core of the present invention is to introduce a large-scale pre-trained model with semantic understanding and generation capabilities as the cognitive core, and combine teaching behavior data with student feedback information to construct a structured and dynamically evolving teaching state representation framework. The system can automatically extract key behaviors and cognitive features from the actual teaching process, characterize the dynamic changes in students' knowledge mastery, and establish a mechanistic connection between teaching behavior and learning outcomes. Furthermore, the present invention realizes an in-depth evaluation of the effectiveness of teaching behavior by modeling the structured causal relationship in the teaching process, and combines the model generation capability to provide teachers with personalized and actionable teaching optimization suggestions, thereby realizing a complete teaching quality evaluation closed loop from "observing results" to "understanding process" to "generating suggestions".

[0052] Compared to existing results-oriented, relevance-driven evaluation systems, this invention emphasizes three key capabilities: process perception, structural modeling, and mechanism reasoning. This enables teaching evaluation to move beyond the aggregation of static indicators and instead continuously perceive teaching dynamics, predict student responses, explain teaching effectiveness, and proactively provide improvement strategies. Building on the powerful language modeling capabilities of large models, this invention further integrates knowledge structure expression with understanding of teaching context, enabling the evaluation system to flexibly adapt to different disciplines, teaching styles, and student groups, significantly enhancing the system's intelligence and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative effort.

[0054] Figure 1 This is a flow chart of the teaching quality evaluation method based on a large model of the present invention. DETAILED DESCRIPTION

[0055] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0056] like Figure 1 As shown, an embodiment of the present invention provides a teaching quality evaluation method based on a large model, the method comprising:

[0057] S1. Obtain teacher explanation text, teaching behavior log, and student feedback data, and construct a triple structured representation, where the triple structured representation includes teaching behavior, knowledge point set, and student feedback.

[0058] Specifically, this step proposes a "Triple-Mapping Semantic Structure Model (TMSS)" to map the original teaching process data into a structured representation S t ={A t ,K t ,B t}, serving as the semantic starting point input for the entire patent system.

[0059] Input Data Description

[0060] Teacher explains text T t : Teaching content in natural language form;

[0061] Teaching Behavior Log t : Structured behavioral data, such as speech rate, interaction frequency, multimedia use, etc.;

[0062] Student feedback data t : Students’ answers, classroom reactions, wrong question texts, etc.

[0063] Furthermore, using the large model f θ T t ,L t ,R t Extract the embedding representation to generate:

[0064] Teacher semantic embedding T =f θ (T t );

[0065] Teaching behavior embedded L =f θ (L t );

[0066] Student feedback embedded in e R =f θ (R t ).

[0067] Fusion representation and structural regularization term introduction (Formula (1))

[0068]

[0069] Among them, E t : The final fused structural semantic representation; W1, W2, W3: The linear weight of each type of embedding, reflecting its contribution; λ: Regularization strength control parameter; The structural regularization term constrains the consistency and decoupling of the three embeddings in the semantic space.

[0070] Among them, the structural regularization term is defined as:

[0071]

[0072] Where, α: content alignment coefficient (e.g., 1.0); β: role decoupling coefficient (e.g., 0.7);

[0073] Part 1: Maintaining semantic consistency between teachers’ explanations, teaching behaviors, and feedback;

[0074] Part 2: Encourage differences in the semantic direction between teaching behavior and student feedback to avoid homogenization.

[0075] Furthermore, the structured output variables are constructed as follows:

[0076] Teaching Behavior A t :From e L The set of mapped teaching behavior labels (e.g., “low interaction,” “fast pace,” etc.);

[0077] Knowledge point set K t : Through e T The set of knowledge points identified by comparison with the standard knowledge base;

[0078] Student Feedback B t :From e R Inferred labels of students’ mastery status (e.g., “high error rate,” “attention fluctuation”);

[0079] Final output structure

[0080] S t ={A t ,K t ,B t} (3).

[0081] S2. Construct a two-class node directed graph based on the triple structured representation; the nodes of the two-class node directed graph include a teaching behavior node set and a student feedback status node set, as well as directed edges, which represent the student status that may be caused by the teaching behavior, to obtain a two-class node directed graph, and calculate the causal effect using the knowledge point constraint as the context filtering condition. If the conditional causal effect of the causal path is positive and significant, and the corresponding student status belongs to the negative feedback label set, it is marked as a high-risk path, and the high-risk path set is output.

[0082] Specifically, this step aims to construct a causal behavior graph G C , used to model structured teaching behavior A t Feedback status B with students t The causal relationship between them and the identification of causal paths that may have negative impacts P i This is the key step in connecting “structured input” and “mechanism understanding” in this patented solution, providing a causal basis for subsequent teaching strategy feedback.

[0083] Input Description: A t : the set of teaching behaviors taken by the teacher at time t (e.g., “fast speaking speed”, “few blackboard writing”, “low interaction”); B t : The set of feedback states that students show after the teaching behavior (such as "high error rate", "attention drift", "knowledge point k i not mastered”); K t : A collection of knowledge points involved in the current teaching, used to mark the contextual consistency of behavior and feedback.

[0084] Furthermore, a two-class node directed graph G is constructedC =(V A ∪V B ,E), where: V A is the set of teaching behavior nodes; V B is a set of student feedback status nodes; E is a directed edge, indicating that “a certain teaching behavior may cause a certain student feedback state”; edge e ij ∈E represents teaching behavior May result in student status The weight indicates the strength of the causal effect.

[0085] Specifically, this step proposes a context-aligned intervention estimation mechanism (CAIE) to extract effective causal relationships from teaching sequences. The present invention introduces knowledge point constraints as context filtering conditions, and only in K t Causal effects are calculated when the data are consistent or highly overlapping to avoid misleading correlations across knowledge content. The intervention effect under contextual conditions is calculated as follows:

[0086]

[0087] Among them, τ ij : Teaching behavior Student Status The conditional causal effect of In the knowledge point context K t Under the implementation behavior hour, expectations; The expected value when the behavior is not implemented in the same context.

[0088] Furthermore, we define the high-risk causal path P i The judgment mechanism:

[0089] If an edge e ij The causal effect τ ij is positive and significant, and Belongs to the system-defined "negative feedback label set" This path is considered to have potential risks;

[0090] The system combines all such edges into a high-risk path set P = {P i}, each path P i Shaped like The high-risk causal path screening criteria are as follows:

[0091]

[0092] Among them, P i:By teaching behavior Negative state of students caused Limited to the knowledge point context K t Medium; δ: the lower limit of the risk causal effect (e.g., δ = 0.2); A collection of negative feedback, including labels such as "decreased mastery", "increased error rate", and "decreased engagement".

[0093] Final output: causal behavior graph structure G C (Visualized teaching impact structure); High-risk path set P = {P i}(for use in subsequent large model generation suggestions).

[0094] S3. Determine a subset of knowledge points affected by the teaching behavior based on the teaching behavior, evolve the knowledge graph state based on the knowledge point subset, generate a knowledge graph substructure at consecutive moments, and calculate the structural change value of the knowledge graph substructure at adjacent moments to quantify the impact of the teaching behavior on the structural state of the knowledge points. Combine all structural change values with a set of high-risk paths for regular constraint processing to generate a teaching behavior causal contribution matrix, wherein each row of the teaching behavior causal contribution matrix represents the structural impact of a teaching behavior on multiple knowledge points.

[0095] Specifically, this step "Causal modeling of teaching behavior on knowledge evolution" is the central structural step in the technical route of this patent, which is located after "Modeling the impact of teaching behavior (Step 2)" and before "Generating feedback suggestions (Step 4)". Its function is to establish the teaching behavior A t and knowledge structure graph evolution G t The mechanistic mapping relationship between the learning process and the learning process is established to clarify how specific behaviors lead to positive or negative evolution of students' knowledge structure. The results of this mechanism modeling directly determine the "content orientation" and "behavior adjustment basis" of subsequent feedback suggestions. In real teaching scenarios, whether students have truly "learned" a knowledge point is not determined by a right or wrong judgment, but by whether the state of the point in their knowledge network, the connection with other points, and the stability of their mastery are improved. Therefore, this step proposes an intervention explanation mechanism based on changes in the knowledge graph structure to quantify "teaching behavior The state evolution of the knowledge graph G t mechanism of action".

[0096] The final output of the present invention is not a simple score, but a graph structure causal weight matrix W K , which provides a causal contribution score of teaching behavior to the evolution of knowledge structure, and becomes the “data-driven explanatory basis” for the subsequent generation of large-scale model strategies.

[0097] Furthermore, for each teaching behavior The present invention first determines the subset of knowledge points that are mainly affected by the behavior through semantic matching of the behavior and knowledge points.

[0098] This subset can be obtained through the large model f θ The embedding vector similarity calculation is implemented by matching the behavior semantics with the knowledge point description and selecting the knowledge points with similar semantics.

[0099] This step does not perform embedded modeling, but only reuses the large model f defined in step 1. θ and knowledge base.

[0100] Further, construct the knowledge graph substructure and

[0101] In the graph G t and G t-1 In the , extract the corresponding The subgraph structure of

[0102] This subgraph contains knowledge point node k j and its edges (indicating dependency or precedence), and node values indicating mastery (e.g., v t (k j ) represents the current moment for knowledge point k j degree of mastery).

[0103] Among them, G t and G t-1 It is dynamically generated by the system based on the students' standardized learning performance data at the end of each teaching cycle. The specific construction method is as follows:

[0104] Node construction: Based on the knowledge point set K involved in the current teaching content t is a set of nodes; each node k j ∈K t Indicates a specific knowledge point.

[0105] Node attribute assignment (mastery calculation): For each knowledge point k j The system calculates the mastery score v of this knowledge point based on the following indicators t (k j ):ExamScore(k j ):The score of the question corresponding to this knowledge point in the formal examinations such as midterm and final exams; HomeworkScore(k j ): score of questions involving this knowledge point in daily homework; QuizScore(k j ): The score of the questions related to this knowledge point in the quiz.

[0106] The mastery level is calculated using the following weighted average formula:

[0107] v t (k j )=0.5×ExamScore(k j )+0.3×HomeworkScore(k j )+0.2×QuizScore(k j )(6)

[0108] Among them: all scores are normalized to the interval [0,1]; weights (0.5, 0.3, 0.2) can be flexibly adjusted according to actual teaching arrangements.

[0109] Edge relationship construction: According to the knowledge point dependency relationship in the standard teaching syllabus, add directed edges between knowledge points; if knowledge point k a It is knowledge point k b , then add directed edges (k a →k b ), indicating mastery of k a Is to master k b premise.

[0110] Historical graph snapshot: After each teaching cycle, the system will build a complete knowledge state graph G t Save as a snapshot and use it as input G for the next cycle t-1 ; Ensure that subsequent analysis of the differences in the impact of teaching behavior on knowledge evolution can be carried out.

[0111] Furthermore, the structural change value ΔS is calculated ij :

[0112] Next, the present invention compares and The structural differences of the knowledge points are quantified to quantify the impact of teaching behavior on the structural state of the knowledge points;

[0113] The differences include changes in mastery, changes in the connectivity of the graph structure, etc. The following formula is defined for calculation:

[0114]

[0115] Where, ΔS ij :Indicates teaching behavior For knowledge point node k j The change in mastery; Attention weight, which represents the effect of the teaching behavior on the knowledge point node k j The semantic relevance of v is in the range [0,1], which is given by semantic matching; t (k j ): student’s knowledge point node k at time tj Mastery of v t-1 (k j ): The student’s previous knowledge point node k j degree of mastery.

[0116] It can be understood that this formula is essentially a weighted modeling of “knowledge changes caused by behavior”, focusing only on the sub-knowledge points that are actually affected by the behavior.

[0117] Furthermore, after obtaining all ΔS ij Afterwards, the present invention further combines the high-risk path P provided in the previous step;

[0118] If the behavior Located in a high-risk pathway (i.e., leading to ), and knowledge point node It is considered that the impact needs to be subject to a structural penalty factor to prevent the subsequent system from misjudging the behavior as a "positive contribution"; the regularization process is as follows:

[0119]

[0120] Among them, W K (i,j): Final teaching behavior For knowledge point node k j λ: regularization coefficient (recommended to be 0.2 to 0.3 times the unit change of mastery); The indicator function takes the value 1 only when the triple path is a risk path;

[0121] This mechanism reflects the creative design added by the present invention in this step: combining behavior-structure modeling with causal risk path regularization to make behavior modeling more causally consistent and robust.

[0122] Finally, the present invention outputs a behavior-knowledge point causal contribution matrix Each row represents the structural impact of a teaching behavior on multiple knowledge points.

[0123] S4. The teaching behavior causal contribution matrix is combined with the teaching behavior set and the knowledge point set, and input into a large language model to generate natural language suggestion content, and then generate teaching suggestions using the natural language suggestion content.

[0124] Specifically, this step aims to calculate the causal contribution matrix W of teaching behavior based on the output of step 3. K , combined with the behavior label A in this round of teaching t And the knowledge point set K t , generating structured teaching suggestions It can be adopted by teachers or systems, thus completing the "assessment-reasoning-feedback" closed loop of the teaching quality evaluation system.

[0125] Furthermore, this step receives three types of structured input data from the previous stage: teaching behavior set A t , represents the teaching behavior performed by the teacher in the current time period; the knowledge point set K t , indicating all the knowledge points involved in this round of teaching;

[0126] Causal influence matrix of teaching behavior on knowledge points Represents each teaching behavior For knowledge point k j The positive and negative impact strength;

[0127] Optional input: student's current knowledge graph state G t , used to provide background explanation support in suggestion generation.

[0128] Furthermore, the specific steps are as follows:

[0129] Identify the target behavior that needs feedback: The system K Identify teaching behaviors with significant impact and optimization value, and build a feedback candidate set

[0130] For every teaching behavior Calculate the cumulative value of its negative causal contribution:

[0131]

[0132] Among them, δ is the negative contribution threshold set by the system, Ω(k j ) is the knowledge point k j Importance rating (e.g. position in the course structure or level of difficulty), is an indicator function; if (γ is the set minimum feedback trigger value), then Considered as a high-priority feedback object; ultimately forming a set of recommended behaviors

[0133] Prepare structure-semantic fusion input: For each The system builds a structural semantic joint input vector: including behavior labels The causal influence sub-vector W of behavior on knowledge points K (i,:); the set of important knowledge points associated with it System template prompts, such as "Please propose specific strategies to improve this teaching behavior to avoid negative impacts on key knowledge points."

[0134] Generating suggested text: Using large model f θ , generate behavioral optimization suggestions:

[0135]

[0136] Among them, f θ The large language model defined in this patent is used to generate natural language suggestion content; suggestion results It is a clear-structured and targeted teaching improvement suggestion, for example, "When explaining 'monotonicity of functions', it is recommended to increase blackboard writing and add real-life examples."

[0137] Establishing the proposed structure means: Structured into standard suggested items for easy display, execution, and tracking, including the following fields: Original teaching behavior: Suggested content: Involved knowledge point collection: Priority Rating: Normalization processing results; Execution status field: initially empty, the teacher or system will fill in whether it is adopted later.

[0138] Summarize all the suggested items to form the feedback set for this round:

[0139]

[0140] Finally, there is a feedback and adoption mechanism: teachers receive feedback through the system interface. You can choose whether to adopt the suggestions in the A. If you adopt them, the generated teaching behavior adjustment will be used as A. t+1 The components of the suggestion are input into the next round of teaching behavior flow; if not adopted, the system will record the feedback status and dynamically adjust the weight and priority threshold in the next suggestion generation;

[0141] This adoption mechanism forms a self-closed-loop teaching quality optimization process of "evaluation → reasoning → feedback → execution → re-evaluation".

[0142] The embodiment of the present invention further provides a teaching quality evaluation system based on a large model, the system comprising:

[0143] A user behavior credibility modeling module is used to obtain user-side data, environmental context, and charging station status information, combine them into a feature vector for the current charging station, input the feature vector into a neural network, output the probability of the user actually appearing and starting charging and the expected real power load value within the current scheduling period of the corresponding charging station, and determine the corresponding actual expected power load, which is the weighted product of the probability of the user actually appearing and starting charging and the expected real power load value within the current scheduling period of the corresponding charging station;

[0144] A spatial thermal potential map construction module is used to embed the actual expected power load into a graph structure to perform thermal potential diffusion analysis and obtain thermal potential values of corresponding nodes;

[0145] The resource scheduling optimization module is used to design a resource scheduling optimization function based on thermal potential diffusion analysis. The resource scheduling optimization function determines the actual power allocation value available to each pile position based on the thermal potential value and constraints. The constraints include that the total power scheduling of the entire station must not exceed the system available value, the allocated power of each pile position cannot exceed its access physical limit, and the system will no longer allocate power to pile positions with extremely low confidence levels, thereby avoiding resource waste due to behavioral prediction errors.

[0146] The control instruction execution and feedback module is used to send the actual power allocation value available for each charging pile position to the actual charging pile control device, drive it to execute power output behavior according to the system strategy within the current scheduling cycle, and monitor the actual control execution status of each pile position in real time, and make dynamic adjustments when necessary.

[0147] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0148] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a division of logical functions. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0149] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0150] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A teaching quality evaluation method based on a large model, characterized by: The method comprises the following steps: Obtaining teacher explanation texts, teaching behavior logs, and student feedback data to construct a triplet structured representation, wherein the triplet structured representation includes teaching behavior, knowledge point set, and student feedback; A two-class node directed graph is constructed based on the triple structured representation; the nodes of the two-class node directed graph include a teaching behavior node set and a student feedback status node set, as well as directed edges, representing student states that may be caused by the teaching behavior. The two-class node directed graph is obtained, and the causal effect is calculated using the knowledge point constraint as the context filtering condition. If the conditional causal effect of the causal path is positive and significant, and the corresponding student state belongs to the negative feedback label set, it is marked as a high-risk path, and the high-risk path set is output; Determine, based on the teaching behavior, a subset of knowledge points affected by the teaching behavior, perform knowledge graph state evolution based on the knowledge point subset, generate a knowledge graph substructure at consecutive moments, and calculate the structural change value of the knowledge graph substructure at adjacent moments to quantify the impact of the teaching behavior on the structural state of the knowledge points. Combine all structural change values with a set of high-risk paths for regularization constraint processing to generate a teaching behavior causal contribution matrix, wherein each row of the teaching behavior causal contribution matrix represents the structural impact of a teaching behavior on multiple knowledge points; The teaching behavior causal contribution matrix is combined with the teaching behavior set and the knowledge point set, and input into a large language model to generate natural language suggestion content, and then the teaching suggestion is generated using the natural language suggestion content.

2. A teaching quality evaluation method based on a large model according to claim 1, characterized in that: The acquisition of teacher explanation text, teaching behavior log and student feedback data to construct a triple structured representation specifically includes: Using the large model, the teacher's explanation text, teaching behavior log and student feedback data are extracted and embedded to generate teacher semantic embedding, teaching behavior embedding and student feedback embedding; A structural regularization term is introduced to process the teacher semantic embedding, teaching behavior embedding, and student feedback embedding; the structural regularization term is used to ensure that the teacher's explanation, teaching behavior, and feedback maintain content semantic consistency, and to encourage the semantic direction between teaching behavior and student feedback to maintain differences, thereby avoiding homogenization; The teaching behavior label set mapped from the teaching behavior embedding is used as the teaching behavior, the knowledge point set identified by comparing the teacher semantic embedding with the standard knowledge base is used as the knowledge point set, and the student mastery status label inferred from the student feedback embedding is used as the student feedback; The teaching behavior, knowledge point set and student feedback are structured as triples.

3. A teaching quality evaluation method based on a large model according to claim 1, characterized in that: The calculation of causal effects using knowledge point constraints as context filtering conditions specifically includes: Obtain the set of knowledge points involved in the current teaching to mark the contextual consistency of behavior and feedback; Determining, based on the knowledge point set involved in the current teaching, the first conditional expectation of the student state when implementing the behavioral teaching behavior; Based on the set of knowledge points involved in the current teaching, the second conditional expectation of the student state when no behavioral teaching behavior is implemented; The difference between the first conditional expectation and the second conditional expectation is taken as the intervention effect under the contextual condition and the causal effect.

4. The teaching quality evaluation method based on a large model according to claim 1 is characterized in that: The knowledge point subset corresponding to the teaching behavior is determined based on the teaching behavior, and is obtained by semantic matching between behavior and knowledge points: The knowledge point subset is obtained by calculating the similarity of the embedding vector of the large model, that is, matching the behavior semantics with the knowledge point description and selecting the knowledge points with close semantics.

5. A teaching quality evaluation method based on a large model according to claim 4, characterized in that: The step of evolving the knowledge graph state according to the knowledge point subset to generate a continuous-time knowledge graph substructure specifically includes: The knowledge point set involved in the current teaching content is the knowledge point node set; each knowledge point node represents a specific knowledge point; Obtain the scores of questions corresponding to the knowledge point node in the formal exam, the scores of questions involving the knowledge point in the homework, and the scores of questions related to the knowledge point in the quiz, and determine the mastery score corresponding to the knowledge point node through weighted summation; Add directed edges between knowledge point nodes based on the knowledge point dependencies in the standard syllabus; A continuous periodic knowledge graph substructure is constructed based on the knowledge point node set and the directed edge set.

6. A teaching quality evaluation method based on a large model according to claim 5, characterized in that: According to the knowledge point dependency relationship in the standard teaching syllabus, directed edges are added between knowledge point nodes. If the first knowledge point node is a prerequisite for the second knowledge point node, a directed edge is added between the first knowledge point node and the second knowledge point node in the graph, indicating that mastering the first knowledge point node is a prerequisite for mastering the second knowledge point node.

7. The teaching quality evaluation method based on a large model according to claim 1 is characterized in that: The structural change value of the knowledge graph substructure at adjacent moments is calculated as: Where, ΔS ij Indicates teaching behavior For knowledge point node k j The change in mastery, i.e., the structural change value; Represents the attention weight, which indicates the teaching behavior on the knowledge point node k j The semantic relevance of v is in the range [0,1]; t (k j ) represents the student's understanding of knowledge point node k at time t j Mastery of v t-1 (k j ) represents the student’s previous knowledge point node k j degree of mastery; Among them, if all structural change values are combined with the high-risk path set for regular constraint processing, the causal contribution matrix of teaching behavior is generated, which specifically includes: If the current path is a high-risk path, the structural change value is adjusted by the regularization coefficient to adjust the teaching behavior causal contribution matrix; If it is not a high-risk path, the structural change value is the causal contribution matrix of the teaching behavior.

8. The teaching quality evaluation method based on a large model according to claim 1 is characterized in that: The step of inputting the teaching behavior causal contribution matrix, the teaching behavior set, and the knowledge point set into a large language model to generate natural language suggestion content, and generating teaching suggestions using the natural language suggestion content specifically includes: Identify teaching behaviors with significant impact and optimization value based on the teaching behavior causal contribution matrix, and construct a feedback candidate set; For each teaching action in the feedback candidate set, a structural semantic joint input vector is constructed: the teaching action, the causal influence sub-vector of the behavior on the knowledge point, the associated important knowledge point set, and the template prompt word; Based on the structural semantics and the combined input vector, a large model is used to generate behavioral optimization suggestions; Generate standard suggestion entries based on the behavior optimization suggestions, including: teaching behavior, suggestion content, knowledge point set, priority score of feedback candidate set, and execution status field; Summarize all the suggested items to form the feedback set for the current round.

9. A teaching quality evaluation method based on a large model according to claim 8, characterized in that: The method of identifying teaching behaviors with significant influence and optimization value based on the teaching behavior causal contribution matrix and constructing a feedback candidate set is as follows: in, is the priority score, δ is the negative contribution threshold set by the system, Ω(k j ) is the knowledge point k j The importance rating of is the indicator function; like γ is the minimum feedback trigger value set, then the teaching action Considered as a high-priority feedback object; ultimately forming a set of recommended behaviors 10. A teaching quality evaluation system based on a large model, characterized by: The system comprises: A structuring unit is used to obtain teacher explanation texts, teaching behavior logs, and student feedback data to construct a triple structured representation, which includes teaching behavior, knowledge point set, and student feedback; A teaching risk construction set is used to construct a two-class node directed graph based on the triple structured representation; the nodes of the two-class node directed graph include a teaching behavior node set and a student feedback status node set, as well as directed edges, representing the student status that may be caused by the teaching behavior, to obtain a two-class node directed graph, and calculate the causal effect using the knowledge point constraint as the context filtering condition. If the conditional causal effect of the causal path is positive and significant, and the corresponding student status belongs to the negative feedback label set, it is marked as a high-risk path, and the high-risk path set is output; A risk evolution set is used to determine, based on the teaching behavior, a subset of knowledge points affected by the teaching behavior, perform knowledge graph state evolution based on the knowledge point subset, generate knowledge graph substructures at consecutive moments, and calculate the structural change values of the knowledge graph substructures at adjacent moments to quantify the impact of the teaching behavior on the structural state of the knowledge points. All structural change values are combined with a set of high-risk paths for regularization constraint processing to generate a teaching behavior causal contribution matrix, wherein each row of the teaching behavior causal contribution matrix represents the structural impact of a teaching behavior on multiple knowledge points; A teaching suggestion generation set is used to input the teaching behavior causal contribution matrix, the teaching behavior set, and the knowledge point set into a large language model to generate natural language suggestion content, and generate teaching suggestions using the natural language suggestion content.

Images