Teaching simulation training method and device, electronic equipment and storage medium

Abstract

The invention provides a teaching simulation training method and device, electronic equipment and a storage medium, and belongs to the technical field of education, and the method comprises the steps: inputting a classroom teaching plan into a pre-trained large language model, and enabling the large language model to analyze the classroom teaching plan and recognize a plurality of interaction nodes, generating a plurality of question templates for each interaction node; when it is detected that the classroom progress reaches any interaction node, one or more virtual student bodies with different communication characteristics are selected from a virtual student body set, and a student question is generated based on a question template corresponding to the interaction node; and recording interaction data to form an interaction log, and generating teaching feedback based on the interaction log, the high-simulation virtual student group system which has diversified communication characteristics, can generate structured interaction behaviors and supports quantitative feedback is constructed, and low-cost, extensible and high-fidelity normal school student teaching skill training is realized.

Description

Technical Field

[0001] This application relates to the field of educational technology, and in particular to a teaching simulation training method, device, electronic device, and storage medium. Background Technology

[0002] With the deepening of teacher education reform, the systematic training of teaching skills for teacher trainees is receiving increasing attention. Teaching skills encompass not only the ability to impart knowledge but also complex dimensions such as classroom organization, teacher-student interaction, and differentiated instruction. Effective teaching skills training requires repeated practice in near-real classroom settings, achieving skill improvement through a closed-loop mechanism of trial teaching—feedback—further practice. Therefore, constructing a high-fidelity, repeatable, and assessable teaching simulation training environment has become a key path to improving the professional competence of teacher trainees and is of great significance to ensuring the quality of future basic education.

[0003] Currently, teacher training for student teachers primarily relies on offline microteaching or video-based simulations. These methods typically involve a small group of real students or classmates playing student roles in trial lessons, with supervising teachers observing and providing experiential feedback. However, this model has significant limitations: First, real student groups are highly homogeneous in cognitive level, learning style, and personality traits, making it difficult to replicate the diversity of students in a real classroom; second, teaching feedback largely depends on teachers' subjective experience, lacking standardized, structured, and time-series evaluation indicators, making it difficult to track the improvement trajectory of teaching behavior; third, simulated scenarios are highly simplified, failing to effectively reproduce dynamic classroom interactions, such as unexpected questions, disciplinary disruptions, and differentiated responses; fourth, each training session requires significant human resources, resulting in high costs and difficulty in scaling up. Although some recent research has attempted to introduce educational technology, such as rule-based virtual students or simple chatbots, their rigid interaction logic and lack of personality traits still cannot support multi-round, complex, and personalized classroom-level teaching simulations. In addition, while general-purpose large language models can generate dialogues, they lack progress triggering mechanisms for teaching scenarios, student differentiation modeling with controllable communication characteristics, and structured evaluation output based on interactive evidence chains, making them difficult to use for reproducible and quantifiable teaching skills training.

[0004] In summary, how to achieve low-cost, scalable, and high-fidelity teaching skills training for teacher trainees has become an urgent technical problem to be solved. Summary of the Invention

[0005] This application provides a teaching simulation training method, device, electronic device, and storage medium to address the shortcomings of existing teaching simulation training methods, which struggle to reproduce the diversity of real classrooms and are costly.

[0006] This application provides a teaching simulation training method, including the following steps: The lesson plan is input into a pre-trained large language model, which parses the lesson plan, identifies multiple interactive nodes, and generates several question templates for each interactive node. When the classroom progress reaches any of the aforementioned interactive nodes, one or more virtual student entities with different communication characteristics are selected from the set of virtual student entities, and student questions are generated based on the question template corresponding to the interactive node. Interaction data is recorded to form an interaction log, and teaching feedback is generated based on the interaction log.

[0007] According to a teaching simulation training method provided in this application, the large language model parses the classroom lesson plan and identifies multiple interactive nodes, and generates several question templates for each interactive node, including: The large language model parses the classroom lesson plan and divides it into multiple teaching units, each of which contains one or more knowledge points. Based on the interactivity of each knowledge point, the knowledge points that need to be interacted with are selected as interactive nodes. One or more question templates are generated for each of the interactive nodes. The question templates contain the question intent and fillable slot fields.

[0008] According to the teaching simulation training method provided in this application, the interaction adaptability is calculated based on the following steps: The interaction fit is obtained by weighting and aggregating cognitive load, error susceptibility, talkability, and contextualization potential. Among them, the cognitive load reflects the degree of psychological effort required for students to understand knowledge points, the error susceptibility reflects the possibility of students misunderstanding knowledge points when learning them, the talkability reflects whether knowledge points are suitable for open dialogue, and the contextualization potential reflects whether knowledge points can be embedded in real-life scenarios.

[0009] According to the teaching simulation training method provided in this application, before selecting one or more virtual student entities with different communication characteristics from the set of virtual student entities and generating student questions based on the question template corresponding to the interactive node, the method further includes: Each virtual student is assigned a habit vector, which reflects the communication characteristics of the virtual student. Based on predefined rules, the habit vectors of each virtual student are converted into specific behavioral parameters, which are used to constrain the text generation strategy of the large language model.

[0010] According to the teaching simulation training method provided in this application, when the classroom progress reaches any of the interactive nodes, one or more virtual student entities with different communication characteristics are selected from the set of virtual student entities, and student questions are generated based on the question template corresponding to the interactive node, including: When the classroom progress reaches any of the aforementioned interactive nodes, one or more virtual student entities are randomly selected, or one or more virtual student entities are selected based on the probability distribution constructed from the aforementioned behavioral parameters. For any selected virtual student, a student question is generated based on the virtual student's behavioral parameters and the question template corresponding to the interaction node.

[0011] According to the teaching simulation training method provided in this application, after generating student questions based on the question template corresponding to the interactive node, the method further includes: Semantic analysis of teacher feedback yields an explanatory completeness index and / or an evidence sufficiency index; When the completeness of explanation index or the sufficiency of evidence index does not meet the preset conditions and the behavior parameters of the virtual student body meet the follow-up question triggering conditions, follow-up questions are generated based on the behavior parameters of the virtual student body according to the teacher's feedback.

[0012] This application also provides a teaching simulation training device, including the following modules: The template generation module is used to: input the classroom lesson plan into a pre-trained large language model, wherein the large language model parses the classroom lesson plan and identifies multiple interactive nodes, and generates several question templates for each interactive node; The question generation module is used to: when the classroom progress reaches any of the interactive nodes, select one or more virtual student entities with different communication characteristics from the set of virtual student entities, and generate student questions based on the question template corresponding to the interactive node; The feedback and evaluation module is used to: record interactive data to form an interaction log, and generate teaching feedback based on the interaction log.

[0013] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the teaching simulation training methods described above.

[0014] This application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the teaching simulation training method as described above.

[0015] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the teaching simulation training method as described above.

[0016] The teaching simulation training method, device, electronic equipment, and storage medium provided in this application input classroom lesson plans into a pre-trained large language model. The large language model parses the classroom lesson plans and identifies multiple interactive nodes, generating several question templates for each interactive node. When the classroom progress reaches any of the interactive nodes, one or more virtual students with different communication characteristics are selected from the set of virtual students. Based on the question templates corresponding to the interactive node, student questions are generated. In other words, different virtual students are given different communication characteristics, enabling them to exhibit differentiated questioning styles, response speeds, and emotional tendencies during interactions, thereby simulating the diversity of students in a real classroom and overcoming the uncontrollability of student groups. Interaction data is recorded to form an interaction log, and teaching feedback is generated based on the interaction log. That is, at each interactive node, the teaching responses of the teacher trainees and the feedback behavior of the virtual students are automatically recorded, forming a structured and time-series interaction log, providing a data foundation for the subsequent construction of a causal evaluation model of teaching behavior, and realizing the quantification and traceability of feedback. Furthermore, leveraging the contextual understanding and generation capabilities of a large language model, the virtual student body can dynamically adjust its questioning strategies in multi-round dialogues, simulating complex interactive scenarios in a real classroom, such as follow-up questions, challenges, and inattentiveness. This significantly enhances the realism and challenge of the training. Moreover, the entire training process does not require the presence of real students or teachers throughout; only a single lesson plan input is needed to automatically generate multiple rounds of personalized practice, greatly reducing manpower and time costs and supporting large-scale deployment. In summary, this application constructs a highly realistic virtual student group system with diverse communication characteristics, capable of generating structured interactive behaviors, and supporting quantitative feedback, achieving low-cost, scalable, and high-fidelity teaching skills training for teacher trainees. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the teaching simulation training method provided in this application; Figure 2 This is a schematic diagram of the teaching simulation training device provided in this application; Figure 3 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] It should be noted that in the description of the embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, they can be fixed connections, detachable connections, or integral connections; they can be mechanical connections or electrical connections; they can be direct connections or indirect connections through an intermediate medium; and they can be internal connections between two elements. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0021] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects have an "or" relationship.

[0022] The following is combined Figures 1-3 This application describes the teaching simulation training method, apparatus, electronic device, and storage medium provided in the embodiments of this application.

[0023] Figure 1 This is a flowchart illustrating the teaching simulation training method provided in this application, such as... Figure 1 As shown, the method includes the following: S110, The lesson plan is input into a pre-trained large language model, which parses the lesson plan and identifies multiple interactive nodes, and generates several question templates for each interactive node; S120, when the classroom progress reaches any of the interactive nodes, select one or more virtual student entities with different communication characteristics from the set of virtual student entities, and generate student questions based on the question template corresponding to the interactive node; S130, record interactive data to form an interaction log, and generate teaching feedback based on the interaction log.

[0024] In this embodiment, the executing entity of the teaching simulation training method can be a teaching simulation training device, which may include, but is not limited to, servers, computer devices such as mobile phones, tablets, laptops, handheld computers, in-vehicle electronic devices, wearable devices, ultra-mobile personal computers (UMPCs), netbooks, or personal digital assistants (PDAs). The executing entity of the teaching simulation training method can also be a teaching simulation training system, which is subordinate to the teaching simulation training device.

[0025] In this embodiment, a lesson plan refers to a teaching plan pre-designed by the teacher for implementing classroom teaching. It typically includes structured information such as teaching objectives, content, steps, time allocation, question design, and activity arrangements. Specific formats may include, but are not limited to, a teaching syllabus, the lesson plan text, a PowerPoint presentation, lecture notes, and blackboard design. In this solution, the lesson plan is provided as input text to the large language model to identify potential opportunities for teacher-student interaction.

[0026] Here, the pre-trained Large Language Model (LLM) is a deep neural network model pre-trained on a large-scale corpus, such as GPT, LLaMA, and Qwen, which possesses powerful natural language understanding and generation capabilities. In this scheme, the LLM is used as an instructional semantic analyzer and a question template generator.

[0027] Here, interactive nodes are key teaching moments in the lesson plan that are pre-set or identified by the model and are suitable for teacher-student interaction. Examples include after the introduction of a concept, after the explanation of an example problem, before group discussion, and during the summary and generalization stage. Each interactive node represents a teaching anchor point that can trigger student questions or feedback.

[0028] Here, the question template is a set of structured or semi-structured question frameworks generated by the LLM based on the context of the interaction nodes. It contains the core question intent and variable parameters, such as knowledge points, context, and difficulty level. For example, the template "I don't quite understand the difference between [concept A] and [concept B], could you explain it again?" is generated by the large language model under the condition of communication characteristic neutrality and can be instantiated by different virtual student entities into personalized questions that include the communication characteristics of the student entity.

[0029] Here, the virtual student is an AI agent with specific personality traits, cognitive level, and communication style, modeled parametrically based on psychological theories. Upon receiving a question template, each virtual student generates specific questions that align with its own personality and behavioral logic.

[0030] Here, teaching feedback refers to the structured evaluation report automatically generated by the system based on the teacher trainees' responses to questions from virtual students, combined with interactive data such as response time, answer accuracy, and guidance strategies. This report can be used to diagnose shortcomings in teaching ability and guide improvements.

[0031] In practice, lesson plans need to undergo standardized textual and structural processing before being input into a pre-trained large language model. Specifically, PPT files require text extraction and page structure recognition, all text content is uniformly encoded into UTF-8, and noise such as footers, page numbers, and irrelevant watermarks is removed.

[0032] In S110, the lesson plan is input into the pre-trained large language model. The large language model is prompted to identify key positions suitable for teacher-student interaction through prompting engineering or fine-tuning. The large language model outputs a list of interactive nodes with timestamps or teaching step numbers.

[0033] Furthermore, for each identified interactive node, the large language model generates several diverse question templates based on the context. Preferably, the question templates cover different cognitive levels and common student confusion types, such as different levels of understanding, application, and analysis, and include easily confused questions and extended questions.

[0034] In the S120, during the simulated teaching process, the lecturer actually lectures to the classroom / student body (a simulated student body on a laptop, desktop computer, or an embodied robot student body, etc.) / camera / microphone / blackboard. The system tracks the lecturer's progress in real time and can detect the current classroom progress through speech-to-text, PPT page turning signals, speech recognition technology, semantic matching, and other methods.

[0035] Furthermore, when the system detects that the current class progress matches a specific interactive node, it triggers the interaction flow for that node. The system randomly selects one or more virtual student entities from the question template library corresponding to that interactive node, or selects a question template for each active virtual student entity. It then personalizes the template based on the communication characteristics of each virtual student entity, generating student questions in text or voice format. For example, for extroverted students, it generates direct and concise questions; for open-ended students, it generates extended questions.

[0036] In S130, instructors respond via voice or text, and the system records their responses, response times, and behavioral data regarding whether they used teaching strategies such as asking questions, providing examples, or engaging in group discussions. All interaction data is stored in the database, which may include, but is not limited to, interaction node IDs, student questions, communication characteristics of virtual students, instructor responses, timestamps, and other data, forming a traceable sequence of teaching behaviors.

[0037] Furthermore, based on preset rules or trained evaluation models, multi-dimensional teaching feedback can be generated. This may include, but is not limited to, whether the answers are scientific and error-free, whether the language was adjusted according to the students' cognitive level, whether it stimulated further thinking or resolved confusion, whether differentiated response strategies were adopted for different types of students, and further improvement suggestions can be generated.

[0038] The teaching simulation training method provided in this application involves inputting a classroom lesson plan into a pre-trained large language model. The large language model parses the lesson plan and identifies multiple interactive nodes, generating several question templates for each interactive node. When the classroom progress reaches any of the interactive nodes, one or more virtual students with different communication characteristics are selected from the set of virtual students. Based on the question templates corresponding to the interactive node, student questions are generated. In other words, different virtual students are given different communication characteristics, enabling them to exhibit differentiated questioning styles, response speeds, and emotional tendencies during interactions. This simulates the diversity of students in a real classroom and overcomes the uncontrollability of student groups. Interaction data is recorded to form an interaction log, and teaching feedback is generated based on the interaction log. That is, at each interactive node, the teaching responses of the teacher trainees and the feedback behavior of the virtual students are automatically recorded, forming a structured and time-series interaction log. This provides a data foundation for the subsequent construction of a causal evaluation model of teaching behavior, achieving quantifiable and traceable feedback. Furthermore, leveraging the contextual understanding and generation capabilities of a large language model, the virtual student body can dynamically adjust its questioning strategies in multi-round dialogues, simulating complex interactive scenarios in a real classroom, such as follow-up questions, challenges, and inattentiveness. This significantly enhances the realism and challenge of the training. Moreover, the entire training process does not require the presence of real students or teachers throughout; only a single lesson plan input is needed to automatically generate multiple rounds of personalized practice, greatly reducing manpower and time costs and supporting large-scale deployment. In summary, this application constructs a highly realistic virtual student group system with diverse communication characteristics, capable of generating structured interactive behaviors, and supporting quantitative feedback, achieving low-cost, scalable, and high-fidelity teaching skills training for teacher trainees.

[0039] In an optional embodiment, the large language model parses the lesson plan and identifies multiple interactive nodes, and generates several question templates for each interactive node, including: The large language model parses the classroom lesson plan and divides it into multiple teaching units, each of which contains one or more knowledge points. Based on the interactivity of each knowledge point, the knowledge points that need to be interacted with are selected as interactive nodes. One or more question templates are generated for each of the interactive nodes. The question templates contain the question intent and fillable slot fields.

[0040] In this embodiment, the large language model performs semantic analysis on the lesson plan, identifying the teaching stages such as introduction, new instruction, practice, and summary, and further dividing each stage into several teaching units. A teaching unit is the smallest teaching segment organized around a core teaching objective, typically containing one or more closely related knowledge points. For example, in a middle school mathematics lesson, the solution to a linear equation in one variable can be considered a teaching unit, which includes knowledge points such as the transposition rule, properties of equality, and solution steps.

[0041] Here, "interaction suitability" refers to whether a particular knowledge point is suitable for deepening understanding through teacher-student Q&A, discussion, and questioning. Not all knowledge points are equally suitable for interaction. For example, purely memorized facts have low interaction value, such as π≈3.14, while conceptual clarification has high interaction potential, such as the difference between speed and rate, and the application of principles.

[0042] Furthermore, the large language model automatically estimates the interactivity of knowledge points based on their semantic features and teaching objectives. Specifically, based on the abstractness, confusion level, and application scenario complexity of knowledge points, as well as the different levels of teaching objectives such as understanding, analysis, and evaluation required for each knowledge point, the interactivity of each knowledge point is evaluated according to preset rules or calculation methods. A threshold is set, and knowledge points with high interactivity are marked as interactive nodes. That is, knowledge points with interactivity greater than the set threshold are marked as interactive nodes, and each interactive node corresponds to its teaching unit and context.

[0043] Furthermore, for each selected interactive node, the large language model generates multiple question templates, which can be generated by combining information including but not limited to: knowledge point content and its common misunderstandings, teaching objectives of the teaching unit to which it belongs, etc.

[0044] In this embodiment, it is required to generate question templates that cover different question types, such as clarifying questions, application and transfer questions, follow-up questions on exceptions, and value and context discussion questions.

[0045] The teaching simulation training method provided in this application focuses interactions on content that truly needs clarification, construction, or deepening of understanding through dialogue, through knowledge point-level interaction adaptation assessment. This ensures that instructors are no longer faced with random or formalized questions, but rather with genuine teaching significance, thereby enhancing the relevance and professional depth of training. The generation of question templates is anchored within a framework of specific knowledge points, teaching objectives, and cognitive levels, giving the question templates a natural cognitive hierarchy. This supports instructors in practicing different levels of teaching response strategies, promoting the development of their advanced teaching abilities. By first dividing teaching units and then identifying interaction nodes, the system ensures that each interaction is embedded in a reasonable teaching process, avoiding abrupt questions that are out of context, making the rhythm of the virtual classroom more consistent with real teaching principles. Furthermore, since each interaction node is associated with a clear knowledge point, teaching unit, and cognitive objective, the system can accurately assess the teaching performance of student teachers in specific knowledge dimensions after recording their responses, providing reliable data for teaching diagnosis and improvement.

[0046] In an optional embodiment, the interaction adaptability is calculated based on the following steps: The interaction fit is obtained by weighting and aggregating cognitive load, error susceptibility, talkability, and contextualization potential. Among them, the cognitive load reflects the degree of psychological effort required for students to understand knowledge points, the error susceptibility reflects the possibility of students misunderstanding knowledge points when learning them, the talkability reflects whether knowledge points are suitable for open dialogue, and the contextualization potential reflects whether knowledge points can be embedded in real-life scenarios.

[0047] In this embodiment, the system utilizes a general large model and text clustering technology to divide the content of a lesson into several teaching units U={u1,u2,…,u…}. M Within each teaching unit u, several knowledge points K are further extracted. j ={k j1 ,k j2 ,…,k jn}, j=1,…,M; Generate a semantic vector representation v for each knowledge point k. k =Enc(k), where Enc(⋅) is the hidden embedding function of the sentence vector encoder or large language model. To identify knowledge points suitable for teacher-student interaction, the system calculates an interaction fit score I(k) for each knowledge point k. Based on teaching experience and historical data, the following feature dimensions are selected: Cognitive load C load(k): Indicates the level of psychological resources required for students to understand this knowledge point. By analyzing the number of preconceptions and the density of technical terms upon which this knowledge point relies, it is estimated that highly abstract knowledge points, those involving multi-step reasoning, and those requiring cross-concept integration typically have a high cognitive load. For example, solving a system of two linear equations in two variables has a higher cognitive load than recognizing triangle types.

[0048] Error Predisposition E error (k): Indicates the likelihood that students will make typical misunderstandings or operational errors when learning or applying this knowledge point. The probability of students being prone to misunderstanding or confusion is inferred through historical classroom data or large-scale models. Specifically, the large-scale language model calls upon a database of common educational psychology myths, such as "the greater the speed, the greater the acceleration," or "plants release carbon dioxide at night, therefore they are harmful." Alternatively, it uses high-frequency error patterns from historical teaching data to determine whether the knowledge point belongs to the error-prone type. For example, negative number operations, the relationship between photosynthesis and respiration, and passive voice structures are all highly error-prone / confusing knowledge points.

[0049] Discretion D disc (k): Indicates whether the knowledge point supports open-ended, multi-perspective, and non-unique answer dialogues. It is determined based on the large model's scoring of the existence of multiple viewpoints and the availability of open discussion space.

[0050] Contextualization potential S scenario (k): Indicates whether the knowledge point can be naturally embedded in real life, social issues, or interdisciplinary contexts for teaching. For example, whether it is easy to connect with life scenarios and subject-specific cases.

[0051] In this embodiment, the system assigns weights to the above four dimensions. w 1, w 2, w 3, w 4, satisfying ∑ w =1, and the weights can be dynamically adjusted according to the characteristics of the subject or the training objectives. For example, science subjects emphasize cognitive load and error susceptibility, while humanities subjects emphasize discussability and contextualization potential. The formula for calculating the interaction fit I(k) is: ; Set threshold ,when When this knowledge point is selected, it will be included as an interactive node in the interactive node list. Among them, C... load (k), E error (k), D disc (k), S scenario (k) are all scores normalized to [0,1].

[0052] The teaching simulation training method provided in this application quantifies the interactive value of knowledge points through four-dimensional indicators. Interaction is triggered only when a knowledge point has high cognitive challenge, high error risk, high discussion space, or high relevance to real life, making each question have real teaching significance. In addition, the four-dimensional attributes are not only used to filter nodes, but also serve as contextual constraints for the generation of virtual student questions. For example, for knowledge points with high error susceptibility, the system can allow some virtual students to actively raise typical erroneous viewpoints, thereby guiding the lecturer to clarify misconceptions. Furthermore, since each question template implicitly contains the four-dimensional label of its source knowledge point, after recording the teacher trainees' responses, the system can reverse-analyze their teaching shortcomings in specific cognitive dimensions, providing a basis for personalized training.

[0053] In an optional embodiment, before selecting one or more virtual student entities with different communication characteristics from the set of virtual student entities and generating student questions based on the question template corresponding to the interactive node, the method further includes: Each virtual student is assigned a habit vector, which reflects the communication characteristics of the virtual student. Based on predefined rules, the habit vectors of each virtual student are converted into specific behavioral parameters, which are used to constrain the text generation strategy of the large language model.

[0054] In this embodiment, the habit vector is a low-dimensional real vector (e.g., 5-8 dimensions) used to structurally represent the stable communication tendencies of virtual students in classroom interactions. It encompasses both personality traits and learning styles and expression preferences. Habit vector dimensions may include, but are not limited to, initiative, clarity of expression, emotional sensitivity, cognitive style, social orientation, questioning tendency, and willingness to seek help. Initiative indicates whether a student tends to ask questions proactively or wait to be called on; clarity of expression refers to whether their language is clear and concise, and their vocabulary is accurate; emotional sensitivity refers to whether they are easily anxious, excited, or resistant to teacher feedback; cognitive style refers to whether they prefer concrete examples or abstract reasoning; social orientation refers to whether they frequently cite peer opinions or seek group validation; questioning tendency refers to whether they habitually challenge teacher statements or demand evidence; and willingness to seek help refers to whether they are willing to directly seek help when encountering difficulties. Different combinations constitute virtual students with different characteristics. The system can generate a set of students of class size S={s1,s2,…,s...} as needed. N}

[0055] Preferably, the five dimensions of the Big Five personality theory are used as habit vectors.

[0056] In the specific implementation process, virtual student bodies can be randomly initialized, randomly sampled according to an approximately normal distribution, or manually set by teachers in typical combinations, or selected based on a pre-set template library.

[0057] Furthermore, abstract habits are transformed into actionable generation control parameters to guide the specific expression of the large language model when instantiating question templates. The five dimensions (O, C, E, A, N) from the Big Five personality theory are used as habit vectors. For example, O corresponds to cognitive curiosity and the frequency of divergent questioning; virtual students with higher O values ​​are more likely to ask "why." C corresponds to learning focus and task adherence; virtual students with lower C values ​​are more likely to go off-topic or answer without following the steps. E corresponds to the rate of raising hands and speaking speed; virtual students with higher E values ​​are more likely to engage in informal discussions. A corresponds to cooperation and emotional support; virtual students with lower A values ​​are more likely to argue and ask challenging questions. N corresponds to emotional fluctuations and tension; virtual students with higher N values ​​are more likely to be anxious, sensitive to difficult questions, and embed emotional words in their question templates.

[0058] In practice, the aforementioned behavioral parameters are used as soft prompts, decoding constraints, or sampling strategy parameters to guide the large language model in generating specific question texts.

[0059] The teaching simulation training method provided in this application provides different teaching response strategies corresponding to different habit vectors. Because the system can stably reproduce these behavioral habit types, instructors can repeatedly practice response strategies in specific situations, accurately identify student types, and adopt appropriate strategies, thereby significantly improving their differentiated teaching and classroom interaction management capabilities. When an instructor fails to effectively respond to a certain type of virtual student, the system can perform attribution analysis based on the virtual student's habit vector, providing behavioral attribution basis for teaching feedback. For example, if multiple highly emotionally sensitive students withdraw after interaction, it indicates that the instructor may have used negative language.

[0060] In an optional embodiment, when the classroom progress reaches any of the interactive nodes, selecting one or more virtual student entities with different communication characteristics from the set of virtual student entities, and generating student questions based on the question template corresponding to the interactive node, includes: When the classroom progress reaches any of the aforementioned interactive nodes, one or more virtual student entities are randomly selected, or one or more virtual student entities are selected based on the probability distribution constructed from the aforementioned behavioral parameters. For any selected virtual student, a student question is generated based on the virtual student's behavioral parameters and the question template corresponding to the interaction node.

[0061] In this embodiment, the system tracks the teaching progress of student teachers in real time through multimodal signals, such as PPT page-turning events, speech-to-text keyword matching, and manual marking by teachers. When the current teaching content matches the context or timestamp of a preset interactive node, the system determines that the interactive node has been reached and activates the subsequent interactive process. The system maintains a class composed of multiple initialized virtual student entities, each of which has been assigned a habit vector and converted into a set of behavioral parameters Θ. i ={Proactiveness, Clarity of Expression, Emotional Sensitivity, ...}. One or more virtual students are randomly selected as speakers for this interaction. For each selected virtual student, a question template is randomly chosen, or selected based on the match between the behavioral parameter set and the question template type. The selected question template and behavioral parameters are input into a large language model, and the final student question is generated through controlled text. For example, if the clarity of expression is low, sentences containing redundant words, filler words, repetitive expressions, incomplete sentence structure, or unclear referents are generated.

[0062] Preferably, the system is configured to have different student body speaking modes, with different student body selection methods for different modes. In some embodiments, one or more students are randomly selected from multiple students as speakers for interactive nodes; in some embodiments, a sampling probability distribution is constructed based on the initiative parameters of each student, and virtual student bodies are selected based on this probability distribution; in other embodiments, at least one low-initiative student body is forcibly included to simulate scenarios where a silent person is called upon or speaks out accidentally.

[0063] The teaching simulation training method provided in this application ensures that even the same interactive node will generate different question subjects, questioning styles, and expression quality in different training rounds through random or probabilistic student selection and personalized generation driven by behavioral parameters. This prevents instructors from relying on memorized answers and requires them to truly understand the teaching content and have flexible adaptability, significantly improving the training effect. Since the behavioral parameters of the student subject from which each question is asked are recorded, the system can analyze the quality of the teacher's response to specific student types afterward, providing a basis for personalized training.

[0064] In an optional embodiment, after generating student questions based on the question template corresponding to the interactive node, the method further includes: Semantic analysis of teacher feedback yields an explanatory completeness index and / or an evidence sufficiency index; When the completeness of explanation index or the sufficiency of evidence index does not meet the preset conditions and the behavior parameters of the virtual student body meet the follow-up question triggering conditions, follow-up questions are generated based on the behavior parameters of the virtual student body according to the teacher's feedback.

[0065] In this embodiment, after the virtual student asks an initial question, the teacher responds via voice or text, i.e., teacher feedback. The system uses natural language understanding to perform structured analysis on the teacher feedback and extract key information, which may include, but is not limited to, content accuracy, explanation depth, response strategy, and emotional tendency. Content accuracy refers to whether the answer is scientific and free of factual errors. Explanation depth refers to whether only a conclusion is given, or whether a reasoning process, example, or analogy is provided. Response strategy refers to whether teaching behaviors such as asking questions, guiding, clarifying, or encouraging are used. Emotional tendency refers to whether the tone is supportive, patient, authoritative, or indifferent.

[0066] Furthermore, the system uses the behavioral parameters of the virtual student to determine whether to trigger follow-up questions. Understandably, not all student responses will involve follow-up questions; the system decides whether to generate follow-up questions and the type of follow-up question based on the virtual student's behavioral parameters. For example, for students with a high tendency to question, if the teacher's answer lacks evidence or contains logical leaps, a challenging follow-up question, such as "Is there any proof?", will be triggered.

[0067] The teaching simulation training method provided in this application, through a dynamic follow-up questioning mechanism based on teacher feedback, enables virtual students to intelligently follow up on the quality of their answers, forming a multi-round dialogue with a logical chain. This achieves a leap from single-round question-and-answer to multi-round teaching dialogue. On the one hand, it provides a chain of behavioral evidence for the assessment of advanced teaching abilities; on the other hand, it approximates real classroom interaction, forcing teachers to improve the depth of their concept explanations.

[0068] The teaching simulation training device provided in the embodiments of this application is described below. The teaching simulation training device described below can be referred to in correspondence with the teaching simulation training method described above.

[0069] Figure 2 This is a schematic diagram of the teaching simulation training device provided in this application, such as... Figure 2 As shown, the teaching simulation training device may include, but is not limited to: The template generation module 210 is used to: input the classroom lesson plan into a pre-trained large language model, wherein the large language model parses the classroom lesson plan and identifies multiple interactive nodes, and generates several question templates for each interactive node; The question generation module 220 is used to: when the classroom progress reaches any of the interactive nodes, select one or more virtual student entities with different communication characteristics from the set of virtual student entities, and generate student questions based on the question template corresponding to the interactive node; The feedback and evaluation module 230 is used to: record interactive data to form an interaction log, and generate teaching feedback based on the interaction log.

[0070] It should be noted that the teaching simulation training device provided in this application embodiment can execute the teaching simulation training method described in any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0071] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute a teaching simulation training method, which includes: The lesson plan is input into a pre-trained large language model, which parses the lesson plan, identifies multiple interactive nodes, and generates several question templates for each interactive node. When the classroom progress reaches any of the aforementioned interactive nodes, one or more virtual student entities with different communication characteristics are selected from the set of virtual student entities, and student questions are generated based on the question template corresponding to the interactive node. Interaction data is recorded to form an interaction log, and teaching feedback is generated based on the interaction log.

[0072] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0073] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the teaching simulation training method provided by the above methods, the method including: The lesson plan is input into a pre-trained large language model, which parses the lesson plan, identifies multiple interactive nodes, and generates several question templates for each interactive node. When the classroom progress reaches any of the aforementioned interactive nodes, one or more virtual student entities with different communication characteristics are selected from the set of virtual student entities, and student questions are generated based on the question template corresponding to the interactive node. Interaction data is recorded to form an interaction log, and teaching feedback is generated based on the interaction log.

[0074] Furthermore, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, is implemented to perform the teaching simulation training methods provided by the methods described above, the method comprising: The lesson plan is input into a pre-trained large language model, which parses the lesson plan, identifies multiple interactive nodes, and generates several question templates for each interactive node. When the classroom progress reaches any of the aforementioned interactive nodes, one or more virtual student entities with different communication characteristics are selected from the set of virtual student entities, and student questions are generated based on the question template corresponding to the interactive node. Interaction data is recorded to form an interaction log, and teaching feedback is generated based on the interaction log.

[0075] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0076] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A teaching simulation training method, characterized in that, include: The lesson plan is input into a pre-trained large language model, which parses the lesson plan, identifies multiple interactive nodes, and generates several question templates for each interactive node. When the classroom progress reaches any of the aforementioned interactive nodes, one or more virtual student entities with different communication characteristics are selected from the set of virtual student entities, and student questions are generated based on the question template corresponding to the interactive node. Interaction data is recorded to form an interaction log, and teaching feedback is generated based on the interaction log.

2. The teaching simulation training method according to claim 1, characterized in that, The large language model parses the lesson plan and identifies multiple interactive nodes, and generates several question templates for each interactive node, including: The large language model parses the classroom lesson plan and divides it into multiple teaching units, each of which contains one or more knowledge points. Based on the interactivity of each knowledge point, the knowledge points that need to be interacted with are selected as interactive nodes. One or more question templates are generated for each of the interactive nodes. The question templates contain the question intent and fillable slot fields.

3. The teaching simulation training method according to claim 2, characterized in that, The interaction adaptability is calculated based on the following steps: The interaction fit is obtained by weighting and aggregating cognitive load, error susceptibility, talkability, and contextualization potential. Among them, the cognitive load reflects the degree of psychological effort required for students to understand knowledge points, the error susceptibility reflects the possibility of students misunderstanding knowledge points when learning them, the talkability reflects whether knowledge points are suitable for open dialogue, and the contextualization potential reflects whether knowledge points can be embedded in real-life scenarios.

4. The teaching simulation training method according to claim 1, characterized in that, Before selecting one or more virtual student entities with different communication characteristics from the set of virtual student entities and generating student questions based on the question template corresponding to the interaction node, the method further includes: Each virtual student is assigned a habit vector, which reflects the communication characteristics of the virtual student. Based on predefined rules, the habit vectors of each virtual student are converted into specific behavioral parameters, which are used to constrain the text generation strategy of the large language model.

5. The teaching simulation training method according to claim 4, characterized in that, When the classroom progress reaches any of the interactive nodes, one or more virtual students with different communication characteristics are selected from the set of virtual students, and student questions are generated based on the question template corresponding to the interactive node, including: When the classroom progress reaches any of the aforementioned interactive nodes, one or more virtual student entities are randomly selected, or one or more virtual student entities are selected based on the probability distribution constructed from the aforementioned behavioral parameters. For any selected virtual student, a student question is generated based on the virtual student's behavioral parameters and the question template corresponding to the interaction node.

6. The teaching simulation training method according to claim 5, characterized in that, After generating student questions based on the question template corresponding to the interactive node, the process further includes: Semantic analysis of teacher feedback yields an explanatory completeness index and / or an evidence sufficiency index; When the completeness of explanation index or the sufficiency of evidence index does not meet the preset conditions and the behavior parameters of the virtual student body meet the follow-up question triggering conditions, follow-up questions are generated based on the behavior parameters of the virtual student body according to the teacher's feedback.

7. A teaching simulation training device, characterized in that, include: The template generation module is used to: input the classroom lesson plan into a pre-trained large language model, wherein the large language model parses the classroom lesson plan and identifies multiple interactive nodes, and generates several question templates for each interactive node; The question generation module is used to: when the classroom progress reaches any of the interactive nodes, select one or more virtual student entities with different communication characteristics from the set of virtual student entities, and generate student questions based on the question template corresponding to the interactive node; The feedback and evaluation module is used to: record interactive data to form an interaction log, and generate teaching feedback based on the interaction log.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the teaching simulation training method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the teaching simulation training method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the teaching simulation training method as described in any one of claims 1 to 6.

Images