Classroom data classification method and apparatus, and electronic device

Abstract

The application relates to the technical field of computers and discloses a classroom data classification method and device and electronic equipment. The method comprises the following steps: obtaining classroom data; preprocessing the classroom data to obtain preprocessed classroom data; according to a pre-set classroom structure model, the preprocessed classroom data is disassembled to obtain teaching links, teaching activities and teaching behaviors corresponding to the preprocessed classroom data; according to the teaching links and the teaching activities, theme content is obtained; a pre-set behavior combination is obtained according to the teaching behaviors, and the behavior combination is filtered according to the theme content to obtain a target behavior combination; the target behavior combination is classified to obtain a classification result corresponding to the classroom data. Through comprehensive processing and analysis, the application improves the accuracy and effectiveness of data classification and enhances the understanding of classroom situations.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus and electronic device for classifying classroom data. Background Technology

[0002] In classroom settings, students' questions and answers are typically analyzed and categorized using a bottom-up approach to obtain more specific and personalized student performance and feedback. This analytical and categorization method helps teachers better understand students' learning needs and performance, thereby providing more targeted guidance and support.

[0003] However, this method also has limitations. Because it focuses primarily on processing the students' questions or answers themselves, while ignoring the relevance of the questions or answers in the classroom context, the classification results are inaccurate. Summary of the Invention

[0004] One objective of this application is to provide a classroom data classification method, apparatus, and electronic device to solve the technical problem of low accuracy in classroom data classification results in related technologies.

[0005] To address the aforementioned technical problems, one technical solution adopted in this application is: providing a classroom data classification method, comprising: acquiring classroom data; preprocessing the classroom data to obtain preprocessed classroom data; decomposing the preprocessed classroom data according to a preset classroom structure model to obtain teaching segments, teaching activities, and teaching behaviors corresponding to the preprocessed classroom data; obtaining theme content based on the teaching segments and teaching activities; obtaining preset behavior combinations based on the teaching behaviors, and filtering the behavior combinations according to the theme content to obtain target behavior combinations; classifying the target behavior combinations to obtain classification results corresponding to the classroom data.

[0006] This classroom data classification method ensures the accuracy and completeness of classroom data through acquisition and preprocessing, providing a foundation for subsequent analysis. Based on a classroom structure model, the preprocessed data is broken down to identify teaching segments, activities, and behaviors, thus combining classroom data with the classroom context. This makes data analysis more closely aligned with actual teaching scenarios and facilitates a deeper understanding of the data's meaning and background. Thematic content is extracted from teaching segments and activities, which helps in understanding the classroom context and relevance, making subsequent analysis more targeted and effective. Pre-defined behavioral combinations are obtained based on teaching behaviors, and these combinations are filtered in conjunction with thematic content to select effective combinations that are relevant to the theme and consistent with the context. This helps narrow the scope of analysis and reduce the impact of noisy data. The filtered effective behavioral combinations are then classified to obtain the final classification results. These results reflect the relevance of questions or answers within the classroom context and their performance under specific themes and teaching behaviors. Therefore, this method, from data processing to obtaining the final classification results, considers the accuracy and completeness of the data. It combines classroom data with the classroom context, and further enhances understanding by extracting thematic content, filtering behavioral combinations, and finally classifying the data. This process helps to deepen the understanding of the data's background and meaning, aids in understanding the relevance of questions or answers within the classroom context, reduces the impact of noisy data, and improves the accuracy and relevance of the final classification results. This classroom data classification method, through comprehensive processing and analysis, improves the accuracy and effectiveness of data classification and enhances the understanding of the classroom context.

[0007] Optionally, the preprocessing of the classroom data to obtain preprocessed classroom data includes: obtaining text data based on the classroom data; processing the text data, including grammar checking, homophone handling, context understanding and optimization, and proper noun and terminology recognition, to obtain processed text data; and determining the speaker's identity in the processed text data to obtain preprocessed classroom data, wherein the preprocessed classroom data is optimized text content containing the speaker's identity. By comprehensively processing the classroom data, the quality and usability of the text data can be improved, and the accuracy and recall rate of student speech recognition can also be effectively improved.

[0008] Optionally, the step of decomposing the preprocessed classroom data according to a preset classroom structure model to obtain the teaching segments, teaching activities, and teaching behaviors corresponding to the preprocessed classroom data includes: inputting the optimized text content containing the speaker's identity into the preset classroom structure model; the classroom structure model analyzing the optimized text content containing the speaker's identity based on its learned structures and patterns to identify teaching segments; identifying the teaching activities contained in each teaching segment based on the identified teaching segments; and analyzing the content and characteristics of the identified teaching activities to identify the teaching behaviors in each teaching activity. Corresponding the preprocessed classroom data to teaching segments, activities, and behaviors allows for a more accurate understanding of teaching activities and behaviors at different stages, which helps in further analyzing and optimizing teaching practices.

[0009] Optionally, the step of inputting the optimized text content containing the speaker's identity into a preset classroom structure model, wherein the classroom structure model analyzes the optimized text content containing the speaker's identity based on its learned structures and patterns to identify teaching segments, includes: preparing the optimized text content containing the speaker's identity as data, the data preparation including text formatting, timestamp organization, and data cleaning to obtain text content that meets the input requirements of the preset classroom structure model; loading the preset classroom structure model, inputting the text content and timestamp information that meet the input requirements into the classroom structure model; extracting features from the text content that meets the input requirements to identify key behaviors, and determining the position of the key behaviors in the classroom data based on the timestamp information; classifying the key behaviors whose positions have been determined to be in the preset teaching segments, thereby obtaining teaching segments containing specific content. The model can fill specific classroom interactions and teaching activities into predefined teaching segment categories, thereby forming a complete and content-rich teaching segment structure.

[0010] Optionally, obtaining thematic content based on the teaching segments and activities includes: extracting text summaries from the interactive content of the teaching activities to obtain the activity theme of each teaching activity; obtaining the segment theme corresponding to each teaching segment based on the activity theme; and obtaining the classroom theme of the entire class content based on the segment theme. The activity theme, the segment theme, and the classroom theme constitute the thematic content. The obtained thematic content provides clear standards for subsequent identification of target behavior combinations, enabling the system to more accurately determine which behaviors are relevant to the teaching objectives and which should be filtered out.

[0011] Optionally, the step of obtaining a preset combination of behaviors based on the teaching behaviors and filtering the combination of behaviors based on the theme content to obtain a target combination of behaviors includes: obtaining a question-and-answer / evaluation behavior combination based on the teaching behaviors; wherein, the question-and-answer / evaluation behavior combination refers to a sequence of behaviors including questioning, answering, and evaluating, extracted from the theme content during teaching activities; inputting the question-and-answer / evaluation behavior combination and the theme content into a large language model, and outputting a valid question-and-answer / evaluation behavior combination through the large language model, wherein the valid question-and-answer / evaluation behavior combination refers to a behavior combination with a score greater than a threshold, i.e., the target combination of behaviors. By filtering and analyzing valid question-and-answer / evaluation behavior combinations, classroom data can be classified more accurately, ensuring the efficiency and accuracy of the classification results.

[0012] Optionally, classifying the target behavior combinations to obtain classification results corresponding to the classroom data includes: acquiring preset classification criteria; inputting the preset classification criteria into a large language model to provide a classification framework for the large language model; inputting the effective question-answer-evaluation behavior combinations and the corpus corresponding to the topic content into the large language model, so that the large language model outputs effective question classification, effective answer classification, and effective evaluation classification according to the classification framework. This allows for systematic classification analysis of classroom interactions, thereby obtaining detailed classification results for classroom data. This not only helps in understanding teaching effectiveness but also helps in optimizing teaching strategies and improving students' learning experience.

[0013] Optionally, the classroom data also includes video data from the classroom process. Determining the speaker's identity in the processed text data specifically involves: performing visual analysis on the video data to obtain student action data from the classroom process; and combining this student action data to determine the speaker's identity in the processed text data. By combining visual analysis of the video data from the classroom process, student action data can be obtained. Using this action data to identify the speaker improves the accuracy of the identification.

[0014] Optionally, the method further includes: analyzing the teaching process in conjunction with the topic content, and then presenting a summary and suggestions for the teaching process.

[0015] To address the aforementioned technical problems, one technical solution adopted in this application is as follows: a classroom data classification device is provided, comprising: a classroom data acquisition module for acquiring classroom data; a data preprocessing module for preprocessing the classroom data to obtain preprocessed classroom data; a classroom structure acquisition module for decomposing the preprocessed classroom data according to a preset classroom structure model to obtain the teaching segments, teaching activities, and teaching behaviors corresponding to the preprocessed classroom data; a theme acquisition module for obtaining theme content based on the teaching segments and teaching activities; a target behavior combination determination module for obtaining preset behavior combinations based on the teaching behaviors and filtering the behavior combinations according to the theme content to obtain target behavior combinations; and a classroom data classification module for classifying the target behavior combinations to obtain the classification results corresponding to the classroom data. This classroom data classification device has the beneficial effects corresponding to the aforementioned classroom data classification method.

[0016] To address the aforementioned technical problems, one technical solution adopted in this application is to provide an electronic device, including a memory and a processor. The memory is connected to the processor, and the processor is configured to execute one or more computer programs stored in the memory. When the processor executes the one or more computer programs, it causes the electronic device to implement a classroom data classification method applicable to electronic devices. This electronic device possesses the beneficial effects corresponding to the aforementioned classroom data classification method applicable to electronic devices.

[0017] To address the aforementioned technical problems, one technical solution adopted in this application is to provide a non-volatile computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are executed by an electronic device, the electronic device performs the classroom data classification method described above. This non-volatile computer-readable storage medium possesses the beneficial effects corresponding to the aforementioned classroom data classification method.

[0018] To address the aforementioned technical problems, one technical solution adopted in this application is to provide a computer program product, comprising a computer program stored on a non-volatile computer-readable storage medium. The computer program includes program instructions, which, when executed by an electronic device, cause the electronic device to perform the classroom data classification method described above. This computer program product possesses the beneficial effects corresponding to the aforementioned classroom data classification method. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the structure of a classroom data processing system provided in one embodiment of this application;

[0021] Figure 2 This is a flowchart of a classroom data classification method provided in an embodiment of this application;

[0022] Figure 3 This is a schematic diagram of the classroom structure model corresponding to the classroom data provided in the embodiments of this application;

[0023] Figure 4 This is a schematic diagram illustrating how the subject matter content is obtained, as provided in an embodiment of this application.

[0024] Figure 5 This is a schematic diagram illustrating the extraction of preset behavior combinations using a finite state machine, as provided in an embodiment of this application.

[0025] Figure 6 This is a schematic diagram illustrating the acquisition of question and answer content corresponding to each teaching activity, provided in an embodiment of this application.

[0026] Figure 7 This is a schematic diagram illustrating how an interactive topic, provided in an embodiment of this application, uses LLM to filter out valid questions, valid answers, and valid evaluations.

[0027] Figure 8 This is a schematic diagram illustrating the classification results obtained from classroom data according to an embodiment of this application;

[0028] Figure 9 This is a schematic diagram of the structure of a classroom data classification device provided in an embodiment of this application;

[0029] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0031] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. Moreover, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0032] Classroom data can encompass various forms of information and content, including text, audio, video, and image data. This data includes students' written answers, teachers' handouts, teaching materials, classroom recordings, recordings of student discussions or presentations, teacher lecture videos, student presentation videos, classroom interaction videos, student questions and answers, and records of interactive behaviors. By categorizing this classroom data, we can better understand and utilize the information within it, thereby supporting needs such as instructional improvement, personalized learning, and intelligent teaching assistance.

[0033] When classifying classroom data, related technologies typically employ a bottom-up approach, analyzing and categorizing student questions and answers. Specifically, in the context of classroom data classification, this bottom-up approach means starting with students' specific questioning and answering behaviors. By analyzing and categorizing these individual behaviors, the overall learning situation and needs are understood, and personalized learning support and teaching improvements are then implemented based on this understanding. This method emphasizes starting from underlying data and specific details of individual behavior, gradually deducing overall conclusions or action plans to better meet the needs of individual students and improve overall teaching effectiveness.

[0034] While bottom-up approaches have advantages in processing questions and answers in classroom data, they also have limitations. One major issue is that they neglect the relevance of questions or answers within the classroom context, potentially leading to inaccurate classification results. This limitation can negatively impact teaching and learning because the analysis and classification of classroom data should consider various factors and contextual information, not just individual student behavior. For example, ignoring the meaning and relevance of student questions or answers in a specific classroom context may lead to misunderstandings or incomplete understandings of student behavior; classification based solely on the content of student questions or answers may fail to adequately reflect important information such as students' thought processes, learning motivations, or cognitive states.

[0035] Therefore, to overcome these limitations, this application provides a classroom data classification method and apparatus, aiming to overcome the problem of neglecting contextual relevance in bottom-up methods and improve classification accuracy by comprehensively processing classroom data and contextual information. The classroom data classification method provided in this application emphasizes comprehensive processing and analysis of classroom data to better understand student behavior and teaching situations, thereby improving the accuracy and effectiveness of classification results. Specifically, it involves: acquiring and preprocessing classroom data to ensure data accuracy and completeness, providing a foundation for subsequent analysis; decomposing the preprocessed data according to a classroom structure model to identify teaching segments, teaching activities, and teaching behaviors, thus combining classroom data with the classroom context, making data analysis closer to actual teaching scenarios and helping to deeply understand the meaning and background of the data; extracting thematic content from teaching segments and activities, including interactive themes, segment themes, and classroom themes, which helps to understand classroom context and relevance, making subsequent analysis more targeted and effective; obtaining preset behavior combinations based on teaching behaviors, and filtering these behavior combinations in conjunction with thematic content to select effective behavior combinations that are relevant to the theme and consistent with the context, which helps to narrow the scope of analysis and reduce the impact of noisy data. The filtered, valid behavioral combinations are categorized to obtain final classification results. These results reflect the relevance of questions or answers within the classroom context and their performance under specific themes and teaching behaviors. Therefore, the entire scheme, from data processing to the acquisition of final classification results, considers the accuracy and completeness of the data, combines classroom data with the classroom context, and extracts thematic content, filters behavioral combinations, and finally classifies them. This process helps to deeply understand the background and meaning of the data, facilitates understanding the relevance of questions or answers within the classroom context, reduces the impact of noisy data, and improves the accuracy and relevance of the final classification results.

[0036] The classroom data classification method illustrated in this application embodiment is applied in a classroom data processing system.

[0037] Please see Figure 1 , Figure 1 This is a schematic diagram of a classroom data processing system provided in one embodiment of this application. The classroom data processing system includes a classroom data processing device 100, an audio acquisition device 200, and a video acquisition device 300. The audio acquisition device 200 and the video acquisition device 300 are respectively connected to the classroom data processing device 100.

[0038] Audio acquisition device 200 is used to record sound data in the classroom, such as student questions, teacher explanations, and student discussions. Video acquisition device 300 is used to record video data in the classroom, including visual information such as teacher lectures, student demonstrations, and classroom interactions. Classroom data processing device 100, as a central processing unit, receives and integrates data from audio acquisition device 200 and video acquisition device 300. This includes converting the data from audio acquisition device 200 and video acquisition device 300 into text data, preprocessing the text data (including grammar checking, homophone handling, context understanding and optimization, proper noun and terminology recognition, etc.), improving the accuracy of text recognition and thus the accuracy of question-and-answer classification; it also determines the speaker's identity based on the text data to improve the accuracy of speaker differentiation and increase the recall rate of student answers.

[0039] The classroom data processing device 100 also classifies the pre-processed classroom data, including breaking down the pre-processed classroom data into classroom activities and tasks according to a preset classroom structure model to obtain the teaching segments, teaching activities, and teaching behaviors corresponding to the pre-processed classroom data; obtaining thematic content based on the teaching segments and teaching activities, including interactive themes, segment themes, and classroom themes; obtaining preset behavior combinations based on teaching behaviors, and filtering the preset behavior combinations according to thematic content to obtain effective behavior combinations; and classifying the effective behavior combinations to obtain the classification results corresponding to the classroom data.

[0040] Classroom data processing equipment 100 can be an electronic device such as a computer server, dedicated data processing equipment, or integrated processor chip; the specific form of classroom data processing equipment 100 is not limited. Classroom data processing equipment 100 can be installed in various locations, depending on system design requirements and actual application scenarios. For example, the equipment can be installed in a corner or on a wall of the classroom for easy access and monitoring during teaching. For large data processing equipment or servers, it can be installed in a dedicated computer room in the school to ensure equipment security and stable network connections. If cloud computing services are used as the data processing platform, the equipment may actually be deployed in a remote data center or cloud server.

[0041] The audio acquisition device 200 can be an electronic device used to acquire sound sources in an application scenario. After acquiring the audio signal, the audio acquisition device 200 transmits it to the classroom data processing device 100. The audio acquisition device 200 can be an integrated microphone array installed in a classroom teaching scenario or several microphones evenly distributed and installed in a classroom teaching scenario.

[0042] The video capture device 300 can be an electronic device used to capture video and images from an application scenario. After capturing the video signal, the video capture device 300 transmits it to the classroom data processing device 100. The video capture device 300 can be a camera, video camera, screen recording device, etc., installed in a classroom teaching setting.

[0043] This application provides a classroom data classification method applied to a classroom data processing device 100. The classroom data processing device 100 can implement the classroom data classification method using pure software or a combination of software and hardware. The classroom data processing device 100 can run an application program for executing the classroom data classification method. This application program can be presented in a form adapted to the electronic device, such as an app application. In some examples, it can also be presented as a system plugin, web page plugin, etc.

[0044] The classroom data classification method proposed in this application is described below through specific embodiments.

[0045] Please see Figure 2 , Figure 2 This is a flowchart of a classroom data classification method provided in an embodiment of this application. The method includes the following steps:

[0046] S11. Obtain classroom data.

[0047] Classroom data can be obtained using the aforementioned audio and video capture devices. The content of the classroom data is detailed in the foregoing embodiments.

[0048] When implementing the classroom data classification method of this application embodiment, the classroom data can be real-time data collected during teaching sessions, or it can be historical classroom data. Alternatively, it can be a combination of real-time and historical classroom data.

[0049] S12. Preprocess the classroom data to obtain preprocessed classroom data.

[0050] The preprocessing of classroom data includes text optimization and speaker identification optimization. First, the acquired classroom data is converted into text data. This process includes converting audio and video into text, which may involve technologies such as automatic speech recognition and image recognition. After obtaining the corresponding text data, the text data undergoes optimization processing. This process includes grammar checking, homophone handling, context understanding and optimization, and proper noun and terminology recognition. Large Language Models (LLMs) can be used to comprehensively optimize the text data in terms of grammar rules, homophones, context understanding and optimization, punctuation checking, and proper noun and terminology recognition.

[0051] For example, the original sentence is: m equals b, so what we need to estimate is... What is that quantity, actually?

[0052] The optimized sentence is: m = ρV, and what we want to estimate is the mass of the air in the classroom.

[0053] The speaker identification process is further optimized. Classroom data also includes video data from the classroom process. To determine the speaker's identity in the processed text data, specifically: after visual analysis of the video data, student action data during the classroom process is obtained; combined with the student action data, the speaker's identity in the processed text data is determined. In this embodiment, to address the issue of low speaker identification accuracy, student standing behavior is included as part of the judgment rule to assist in speaker recognition. Through experimental pathways and analysis of a large amount of classroom data, this method demonstrates significant performance in teacher-student question-and-answer scenarios, greatly improving the accuracy and recall rate of student speech recognition.

[0054] When incorporating the condition of students standing up into the decision-making rules to determine the speaker's role, timestamps, semantic analysis, and image analysis can be combined. Specifically, video data is collected in the classroom using devices such as cameras, and timestamps and corresponding text data are recorded. Semantic analysis is performed on the text data to identify keywords and pronouns to determine the speaker's identity. Image analysis technology is used to monitor students' actions in the classroom, such as whether any students stand up. Based on the results of timestamps, semantic analysis, and image analysis, decision-making rules are formulated. For example, when a question appears in the text, combined with image analysis data, if a student is detected standing up within a certain period after the question is asked, the standing student can be identified as the speaker answering the question. Then, according to the decision-making rules, the standing student is identified as the speaker, and their identity information is associated with the corresponding text data. For example, after the teacher asks a question, student A stands up to answer; based on the timestamp, a certain period after the question is asked is determined; image analysis detects student A standing up; combined with semantic analysis, the answerer is confirmed to be student A; if a student stands up within 5 seconds of the question being asked, the standing student is marked as the speaker answering the question. In this way, by combining information such as semantic analysis, image analysis, and timestamps, the condition of students standing up can be added to the judgment rule, thereby determining the speaker's role in the classroom dialogue data.

[0055] The above-mentioned text optimization and speaker identity differentiation optimization processes can be carried out simultaneously, or text optimization can be prioritized, and speaker identity differentiation optimization can be performed based on the optimized text data, or speaker identity differentiation optimization can be performed first, and then text optimization can be performed based on the optimization result.

[0056] In some embodiments, text data is obtained based on classroom data; the text data is processed, including grammar checking, homophone handling, context understanding and optimization, and proper noun and terminology recognition, to obtain processed text data; the speaker's identity is determined in the processed text data to obtain preprocessed classroom data. Therefore, by comprehensively processing classroom data, the quality and usability of the text data can be improved, and the accuracy and recall of student speech recognition can be effectively improved, thus providing a more reliable and effective foundation for subsequent data analysis and applications.

[0057] S13. Based on the preset classroom structure model, the preprocessed classroom data is broken down to obtain the teaching links, teaching activities and teaching behaviors corresponding to the preprocessed classroom data.

[0058] In real-world classroom settings, effective classroom Q&A is closely linked to the teaching content. However, real classroom behavior is often complex and diverse, meaning that classroom Q&A is not always directly related to the teaching content. In this embodiment, classroom Q&A is divided into two categories: effective and ineffective. Effective Q&A refers to interactions closely related to the teaching content, while ineffective Q&A refers to interactions unrelated to the teaching content. To accurately capture effective Q&A, it is necessary to deeply analyze the design and composition of the classroom structure, break down each activity and task, analyze the theme of each stage, and gradually deduce the main teaching line, thereby identifying Q&A behaviors related to the teaching content and filtering out ineffective Q&A.

[0059] Therefore, in this embodiment, based on a preset classroom structure model, the preprocessed classroom data is broken down into classroom activities and tasks to obtain the corresponding teaching segments, teaching activities, and teaching behaviors. Teaching segments are the main stages or parts of the classroom teaching process, reflecting the overall structure and flow of teaching. Each teaching segment has specific goals and functions, helping teachers and students to progress systematically in the learning process. Teaching activities refer to the specific operations or interactive forms in which teachers and students jointly participate in a specific teaching segment. Teaching activities usually aim to achieve specific teaching objectives and promote student learning and participation. Teaching behaviors refer to the actual actions or performances taken by teachers and students in specific teaching activities. Teaching behaviors are the concrete manifestation of achieving teaching activities and can directly affect learning outcomes. By understanding teaching segments, teaching activities, and teaching behaviors, the structure and dynamics of classroom teaching can be analyzed more clearly, and the classroom flow can be clarified.

[0060] Please see Figure 3 , Figure 3 This is a schematic diagram of the classroom structure model corresponding to the classroom data provided in this application embodiment. The preset classroom structure model is obtained by training on a large amount of classroom data. When using this classroom structure model, the preprocessed classroom data is input into it, and the model generates content corresponding to the input data based on the learned structure and patterns. This generated content is generated according to the structure and characteristics of the classroom structure model, including specific teaching steps, teaching activities, and related data content (i.e., teaching behaviors).

[0061] like Figure 3As shown, the classroom structure model mainly includes teaching segments, teaching activities, and teaching behaviors. The teaching segments consist of classroom introduction, the core process of teaching and learning, knowledge transfer and integrated application, and classroom summary. The teaching activities corresponding to the classroom introduction include interactive teaching activities, and the corresponding teaching behaviors include teacher questioning. The teaching activities corresponding to the core process of teaching and learning include explanatory teaching activities, interactive teaching activities, and tasks such as inquiry-based tasks, assessment tasks, memorization tasks, and transfer and application tasks. The teaching behaviors corresponding to the core process of teaching and learning include teacher lecturing, teacher questioning, student answering, teacher feedback, student-student interaction, teacher observation, teacher assignment of tasks, student task completion, teacher guidance, student presentation of results, student sharing of ideas, teacher evaluation and feedback, student self-evaluation, and student peer evaluation. The teaching activities corresponding to knowledge transfer and integrated application include transfer and application tasks. The teaching activities corresponding to the classroom summary include knowledge summary activities and homework assignment activities. The teaching behaviors corresponding to the knowledge summary activities include teacher summarization and student summarization. The teaching behaviors corresponding to the homework assignment activities include teacher assignment of homework.

[0062] It can be obtained through the following methods Figure 3 The classroom structure model shown is as follows. First, a certain amount of classroom data is collected, including the text of the dialogue between teachers and students in the classroom. This can be achieved through Automatic Speech Recognition (ASR) technology, which can collect the text of the teacher-student dialogue in the classroom and convert the oral communication into text form.

[0063] Then, the obtained teacher-student dialogue texts in the classroom are classified using a pre-defined teaching behavior classifier. The classification principle can be based on a large model. For example, a training dataset is collected and prepared, including teacher-student dialogue texts and their corresponding teaching behavior classification labels; a large pre-trained model suitable for text classification tasks, such as GPT, is selected, and the model has strong semantic understanding capabilities by pre-training on large-scale text data; the selected large pre-trained model is fine-tuned on the collected teacher-student dialogue data, and the model parameters are adjusted on the training data to better adapt to the specific classification task; the fine-tuned model is used to train the data so that the model can learn to classify texts into different teaching behavior categories, such as teacher questions, student answers, and student-student interactions; the model's performance is then evaluated on an independent validation set, including metrics such as accuracy and recall, to ensure the model's generalization ability and accuracy; the trained model is applied to actual classroom dialogue text data to achieve automatic text classification, thereby identifying different teaching behaviors and obtaining teaching behavior classification results.

[0064] Next, based on the obtained teaching behavior classification results and the timestamp corresponding to each text in the teaching behavior classification results, the texts in the teaching behavior classification results are sorted. Then, key points such as teacher questioning, teacher task assignment, teacher lecturing, teacher summarizing, and teacher homework assignment are identified, and these key points are used as cutting points to extract classroom tasks from the teaching behavior classification results.

[0065] Next, the classroom tasks mentioned above are summarized and categorized into teaching activities using a task classifier, such as interactive teaching activities, explanatory teaching activities, knowledge summarization activities, and homework assignment activities. The task classifier can also be implemented based on a large model, with a similar implementation principle to the method described above.

[0066] Finally, based on the classification results of the teaching activities and the teaching behaviors included in each activity, they were linked to the teaching segments. For example, interactive teaching activities include teacher questioning, which is categorized as the lesson introduction segment. Similarly, the summarized teaching segments also include the core process of teaching and learning, knowledge transfer and comprehensive application, and lesson summary.

[0067] The names of teaching segments such as lesson introduction, core teaching and learning process progression, knowledge transfer and integrated application, and lesson summary can be determined based on consensus in teaching theories, instructional design frameworks, or educational research. These teaching segments typically represent different stages or important parts of the teaching process, each with a specific function and purpose. The selection of these names is usually based on widely accepted teaching terminology or teaching models, so that educators can better understand and share information about instructional design and implementation.

[0068] The above describes the process of reasoning to derive the classroom structure model, including data collection and text classification, timestamp processing and key point identification, task classification and teaching activity categorization, and association of teaching segments. Specifically, firstly, ASR technology is used to collect teacher-student dialogue texts in the classroom. Then, a pre-defined teaching behavior classifier is used to classify the text, identifying different categories of teaching behaviors. The text is then sorted according to timestamps, and key points such as teacher questions and lectures are identified as cutting points to segment the classified text into different classroom tasks. A task classifier is used to summarize and classify the segmented classroom tasks, resulting in different types of teaching activities. Based on the classification results of the teaching activities and the teaching behaviors they contain, each teaching activity is associated with a corresponding teaching segment, forming a correspondence between tasks and teaching segments.

[0069] The classroom structure model is derived through the above method. In the embodiments of this application, this classroom structure model can be directly applied, i.e., as follows: Figure 3As shown in the diagram, this classroom structure model allows for the breakdown of currently acquired classroom data, identifying teaching segments, activities, and behaviors. This integrates classroom data with the classroom context, making data analysis more closely aligned with actual teaching scenarios and facilitating a deeper understanding of the data's meaning and background.

[0070] according to Figure 3 The classroom structure model shown includes a breakdown of teaching segments, activities, and behaviors. The pre-processed classroom data can be considered part of this model. By inputting the pre-processed classroom data into this model, different teaching segments, activities, and behaviors can be mapped to them, allowing us to understand the position and role of the data within the overall teaching process and thus grasp the context of the lecture.

[0071] Specifically, based on a pre-defined classroom structure model, the pre-processed classroom data is broken down to obtain the corresponding teaching segments, activities, and behaviors, including:

[0072] The optimized text content containing the speaker's identity is input into a preset classroom structure model. The classroom structure model analyzes the optimized text content containing the speaker's identity based on the structure and pattern it has learned, so as to identify the teaching segment.

[0073] Based on the identified teaching segments, identify the teaching activities contained in each teaching segment;

[0074] Based on the identified teaching activities, the content and characteristics of the teaching activities are analyzed to identify the teaching behaviors in each teaching activity.

[0075] Specifically, the optimized text content containing the speaker's identity is input into a preset classroom structure model. The classroom structure model analyzes the optimized text content containing the speaker's identity based on its learned structures and patterns to identify teaching segments, including:

[0076] The optimized text content containing the speaker's identity is prepared as follows: data preparation includes text formatting, timestamp organization, and data cleaning to obtain text content that meets the input requirements of the preset classroom structure model; the preset classroom structure model is loaded by inputting the text content and timestamp information that meet the input requirements into the classroom structure model; feature extraction is performed on the text content that meets the input requirements to identify key behaviors, and the position of the key behaviors in the classroom data is determined based on the timestamp information; the key behaviors with determined positions are classified and categorized into preset teaching segments to obtain teaching segments containing specific content.

[0077] The above-mentioned data preparation involves optimizing the text content, including speaker identification, text formatting, timestamp organization, and data cleaning, to obtain text content that meets the input requirements of the classroom structure model. This text content, along with timestamp information, is then input into the model to identify key behaviors and determine their specific locations within the classroom data based on the timestamp information. The identified key behaviors are then categorized and assigned to pre-defined teaching segments, resulting in teaching segments containing specific content. The ultimate goal is to obtain teaching segments containing specific content, which can be categorized into four predefined categories: classroom introduction, core teaching and learning process progression, knowledge transfer and integrated application, and classroom summary. Through these steps, the model can fill specific classroom interactions and teaching activities into predefined teaching segment categories, thus forming a complete and content-rich teaching segment structure.

[0078] Furthermore, based on the preset classroom structure model, the preprocessed classroom data is divided into different teaching segments, including classroom introduction, core teaching and learning process advancement, knowledge transfer and comprehensive application, and classroom summary; the teaching activities corresponding to the classroom introduction, the core teaching and learning process advancement, the knowledge transfer and comprehensive application, and the classroom summary are identified; and teaching behaviors are identified based on the identified teaching activities.

[0079] Based on a pre-defined classroom structure model, the pre-processed classroom data is divided into different teaching segments, such as lesson introduction, core teaching process, knowledge transfer and application, and lesson summary. For each teaching segment, corresponding teaching activities are identified, such as lesson introduction activities, core teaching process activities, knowledge transfer and application activities, and lesson summary activities. Timestamp information is used to map the classification results of teaching behaviors to the time periods of the teaching segments, determining the time points when each teaching activity occurs. Based on the time points, the specific teaching activities included in each teaching segment are identified. Finally, the content and characteristics of the teaching activities are further analyzed to identify specific teaching behaviors, providing a deeper understanding of the details of teaching behaviors in each activity.

[0080] For example, the teaching activities corresponding to the introduction of a lesson include: guiding students with questions and setting the scene; the corresponding teaching behaviors include: teachers asking questions, students answering questions, and teachers explaining questions.

[0081] The core processes of teaching and learning include: explanation, practice, and discussion; the corresponding teaching behaviors include: teacher explanation, student practice, and student discussion.

[0082] The teaching activities corresponding to knowledge transfer and integrated application include: practical application and case analysis; the corresponding teaching behaviors include: teacher guidance, student practice, and student presentation.

[0083] Classroom summary corresponds to teaching activities including: knowledge review and homework assignment; and corresponding teaching behaviors include: teacher summary and teacher homework assignment.

[0084] By breaking down the data as described above, the preprocessed classroom data can be correlated with teaching segments, activities, and behaviors, allowing for a more accurate understanding of teaching activities and behaviors at different stages. This helps in further analyzing and optimizing teaching practices.

[0085] Among them, from Figure 3 In the classroom structure model shown, classroom Q&A will appear in "Teaching and Learning and Core Problem Advancement" and "Knowledge Transfer and Comprehensive Application". The Q&A format will be divided into two types: teacher-student interactive Q&A and teacher-student task-based Q&A. These two Q&A formats are different, and the analysis process will also be different:

[0086] Interactive Q&A between teachers and students: This type of scenario is quite common, which is analyzed using the "question-answer-evaluation" behavior combination. The teacher asks a question and signals the students to answer. After the students raise their hands, stand up, and answer, the teacher gives evaluation feedback.

[0087] Teacher-student task-based question and answer: The teacher will give a task theme, which is usually a clear question. Then, through organizing inquiry-assessment-memorization-transfer and application tasks, students will answer the question by completing the tasks. The teacher will then provide evaluation and feedback.

[0088] S14. Obtain the theme content based on the teaching steps and activities.

[0089] Based on the above steps, the classroom structure can be broken down into teaching segments and activities. A Large Language Model (LLM) can be used to summarize and refine the content of each segment and activity, extracting interaction themes, segment themes, and the overall classroom theme. These extracted themes are then used to further refine effective question-and-answer sessions in the following steps. Specifically, the LLM can be used to extract text summaries of the interactive content in teaching activities, identifying the interaction themes between students and teachers. This can be achieved through generative models or extractive summarization methods to summarize the interactive content, resulting in the activity themes corresponding to the teaching activities. Next, text summaries are extracted from the activity themes corresponding to each teaching segment, extracting the theme content of that segment, i.e., the segment theme. The LLM can identify key content and information from the activity themes corresponding to each segment, thus obtaining the segment theme. Finally, the thematic content of each teaching segment and activity is integrated to generate a summary of the entire lesson's theme, i.e., the classroom theme. The LLM can synthesize the themes of each segment to extract the overall classroom theme.

[0090] Another approach is to extract interactive themes, session themes, and classroom themes based on network recognition methods. This method leverages existing teaching datasets and uses network recognition techniques, such as deep learning models (like recurrent neural networks or Transformer models), to train the model to identify thematic content within teaching activities and sessions. The trained network model is then used to analyze the text of each teaching session and activity, identifying and extracting thematic content. This allows for a more accurate capture of the core theme of each session and activity.

[0091] Both large language models and network recognition methods can acquire the thematic content of each teaching segment and activity, including interactive themes, segment themes, and classroom themes. This acquired thematic content provides clear criteria for subsequent identification of target behavior combinations, enabling the system to more accurately determine which behaviors are relevant to the teaching objectives and which should be filtered out.

[0092] like Figure 4 As shown, the process includes: first, summarizing the theme of each teaching activity, i.e., the activity theme; then, summarizing the theme of each teaching segment based on these activity themes, i.e., the segment theme; and finally, summarizing the entire classroom content based on the segment themes to obtain the classroom theme. Here, the activity theme refers to the core content and objectives focused on in a specific teaching activity, such as problem-based learning, discussion, or practice. The segment theme refers to the core content and objectives supported by all activities in a specific teaching segment, summarizing the main knowledge points and teaching strategies of that segment. The classroom theme refers to the core idea and objectives of the entire lesson, reflecting the main content and learning outcomes of the course; it is a comprehensive distillation based on all segment themes.

[0093] S15. Obtain the pre-set behavior combinations based on teaching behaviors, and filter the behavior combinations according to the theme content to obtain the target behavior combinations.

[0094] Understandably, every teaching activity involves related teaching behaviors. As mentioned above, there are two types of question-and-answer interactions: interactive question-and-answer and task-based question-and-answer. In this step, we need to extract the "question-answer-evaluation" behavior combinations for these two types of questions and answers from the tasks. The pre-set behavior combinations include the "question-answer-evaluation" combination, and may also include other forms, such as asking a question, answering a question, providing feedback, guiding an exploration, summarizing, task-execution, evaluation, interaction, reflection, and adjustment.

[0095] The "question-answer-evaluate" behavioral combination refers to a sequence of behaviors in teaching activities that includes asking questions, answering questions, and evaluating content. For example, if the teaching topic is about animal ecosystems, and one activity involves discussing food chains, the following question-answer-evaluate behavioral combination might exist in this context:

[0096] Question: Teacher: What is a food chain? Why is it so important in an ecosystem?

[0097] Response: Students answer questions about food chains and provide examples.

[0098] Evaluation: Teacher: Very good, you explained it very clearly, and you have a good grasp of the concept of the food chain!

[0099] It's understandable that not all questions, answers, and evaluations are related to the classroom content. In order to filter out effective questions and answers, it is necessary to analyze the relevance between the topic content and the questions and answers based on the obtained topic content, and then filter out effective questions, effective answers, and effective evaluations.

[0100] In some embodiments, a Finite State Machine (FSM) is used to extract the "question-answer-evaluation" behavior combinations in each activity, and then combined with the topic content to filter out the effective "question-answer-evaluation" combinations. A Finite State Machine is a mathematical model used to describe the transitions between different states of a system. Its specific state transition diagram is shown in Figure 5. The components of a Finite State Machine include states, behaviors, and transitions. Based on the state transitions, the question, answer, and evaluation language that triggers state transitions is extracted as reserve content for question-and-answer classification. The corresponding overall schematic diagram is shown in Figure 5. Figure 6 As shown, the question-and-answer content has been extracted, yielding the corresponding question-and-answer content for each teaching activity, including the "question-answer-evaluation" combination. Besides using the FSM method, other methods such as state transition diagrams and Markov decision processes can also be used to obtain the preset behavioral combinations.

[0101] Next, in order to filter out the valid question and answer content, such as Figure 7 As shown, based on interactive topics, the capabilities of LLM can help analyze the relevance between the topic and the question-and-answer content, thereby filtering out effective questions, effective answers, and effective evaluations.

[0102] Effective questioning refers to questions that guide student thinking, promote discussion, and are closely related to the teaching content. Effective answers refer to responses that accurately address questions, demonstrating students' understanding and application of knowledge. Effective evaluation refers to feedback from teachers or students on questions and answers, aimed at promoting learning and improving understanding. To obtain effective questions, answers, and evaluations based on the topic content, the following steps can be taken: First, clearly define the topic content, i.e., clarify the specific content of the activity theme, segment theme, and classroom theme; then, input the obtained question-and-answer content and topic content into an LLM (Lesson Learning Model), requesting the LLM to evaluate the relevance of each question, answer, and evaluation to the topic; then, set a relevance threshold based on the LLM analysis results, for example, content with a relevance score higher than 0.7 is considered effective. Finally, filter out effective questions, answers, and evaluations based on the threshold.

[0103] In this embodiment, by filtering and analyzing effective question-and-answer evaluation behavior combinations, classroom data can be classified more accurately, ensuring the efficiency and accuracy of the classification results.

[0104] S16. Classify the target behavior combinations to obtain the classification results corresponding to the classroom data.

[0105] In this embodiment, a preset classification standard is first obtained. The classification standard for teacher questions includes: teacher questions can be classified according to two methods: Bloom's classification and the Four Ws classification. Bloom's classification divides questions into six levels, including memory, understanding, application, analysis, evaluation, and creation. The Four Ws classification focuses on the content type of the question, dividing it into four categories: what, why, how, and if. Student answers can be classified according to their content type, including: explanatory answers, where students provide detailed explanations, demonstrating a deep understanding of the question; direct answers, where students respond simply and directly to the question, usually brief; and no answer, where students do not answer the question, possibly because they are unsure or do not know the answer. Teacher evaluations of student answers can be categorized as follows: simple affirmation, where the teacher simply acknowledges the student's answer; specific affirmation, where the teacher specifically acknowledges the student's answer; encouragement, where the teacher uses positive language to encourage the student to continue their efforts; direct negation, where the teacher explicitly points out the error in the student's answer; and repetition of the question or student's answer, where the teacher repeats the question or student's answer, possibly for emphasis or confirmation.

[0106] The aforementioned classification criteria define clear categories and standards for asking, answering, and evaluating questions. This provides a clear framework for subsequent analysis, enabling the Large Language Model (LLM) to understand and process the input corpus. Therefore, this classification criterion is input into the Large Language Model to provide a classification framework for it.

[0107] Next, the compiled effective question-and-answer / evaluation behavior combinations and corpus related to the topic content are input into the LLM. This input data includes the specific texts of questions, answers, and evaluations. The LLM analyzes and processes the input content according to the classification framework, outputting effective question categories, effective answer categories, and effective evaluation categories. Specifically, the LLM parses the input question, answer, and evaluation texts, identifying keywords, phrases, and sentence structures, and can also determine the speaker's identity for each statement. Then, according to the classification framework, the LLM categorizes the questions, answers, and evaluations separately. For example: Question classification: Based on Bloom's taxonomy and the WHM (What, What, What, What) classification, questions are divided into levels such as memory and comprehension. Answer classification: Student answers are classified as explanatory answers, direct answers, or no answer. Evaluation classification: Teacher evaluations are classified as simple affirmations, targeted affirmations, etc. Next, the LLM generates the classification results for effective questions, effective answers, and effective evaluations based on the analysis results. For example: Effective question classification: Comprehension. Effective answer classification: Explanatory answer. Effective evaluation classification: Motivation.

[0108] In the above process, LLM first performs preliminary classification of the input text content according to preset classification criteria, identifying the basic types of questions, answers, and evaluations. The output of this step is to label all content as questions, answers, or evaluations. Based on the first classification, LLM conducts more detailed analysis and classification of each category (questions, answers, and evaluations) to identify effective questions, effective answers, and effective evaluations. Specifically, this may include: further classifying questions into different types (such as memory, comprehension, etc.) based on their content and form, thus obtaining effective question classifications; analyzing student answers to determine their type (such as explanatory answers, direct answers, etc.), thus obtaining effective answer classifications; and analyzing teacher evaluations to determine their nature (such as simple affirmation, encouragement, etc.), thus obtaining effective evaluation classifications.

[0109] Through this phased classification process, LLM can first identify the basic structure of the text, and then analyze each part in depth to generate more specific classification results. The final output will include structured data of effective questions, effective answers, and effective evaluations.

[0110] Figure 8 This is a schematic diagram illustrating the classification results obtained from classroom data in this embodiment. Specifically, the classification includes: Effective Question Classification: Based on Bloom and the four-way classification method, the classification results of teacher questions are output. Effective Response Classification: The type of student response is output, such as explanatory, direct, or no response. Effective Evaluation Classification: The type of evaluation given by the teacher to the student's response is output.

[0111] By following the steps above, classroom interactions can be systematically categorized and analyzed, resulting in detailed classification results of classroom data. This not only helps in understanding teaching effectiveness but also helps in optimizing teaching strategies and enhancing students' learning experience.

[0112] In some embodiments, the method further includes: analyzing the teaching segments in conjunction with the topic content, and then displaying summary suggestions for the teaching segments. Specifically, the segment topics in the topic content, along with classroom data, can be input into an LLM network. The LLM network can then summarize and provide suggestions for each teaching segment, offering multi-dimensional analysis results for classroom analysis.

[0113] Please see Figure 9 , Figure 9 This is a schematic diagram of a classroom data classification device provided in an embodiment of this application. The device 400 includes:

[0114] Classroom data acquisition module 401 is used to acquire classroom data;

[0115] Data preprocessing module 402 is used to preprocess the classroom data to obtain preprocessed classroom data;

[0116] The classroom structure acquisition module 403 is used to decompose the preprocessed classroom data according to the preset classroom structure model to obtain the teaching links, teaching activities and teaching behaviors corresponding to the preprocessed classroom data.

[0117] The topic acquisition module 404 is used to obtain topic content based on the teaching steps and the teaching activities;

[0118] The target behavior combination determination module 405 is used to obtain a preset behavior combination based on the teaching behavior, and filter the behavior combination according to the theme content to obtain a target behavior combination.

[0119] The classroom data classification module 406 is used to classify the target behavior combination and obtain the classification result corresponding to the classroom data.

[0120] The classroom data classification device 400 described above can be a software module. The software module includes several instructions, which are stored in a memory. The processor can access the memory and call the instructions to execute them in order to complete the classroom data classification method described in the above embodiments.

[0121] In some embodiments, the classroom data classification device 400 can also be constructed from hardware devices. For example, the classroom data classification device 400 can be constructed from one or more chips, and the chips can work in coordination to complete the classroom data classification method described in the various embodiments. As another example, the classroom data classification device 400 can also be constructed from various logic devices, such as general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), microcontrollers, ARM (Acorn RISC Machine) or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination of these components.

[0122] It should be noted that the classroom data classification device 400 described above can execute the classroom data classification method for electronic devices provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in the embodiments of the classroom data classification device 400 can be found in the classroom data classification method for electronic devices provided in the embodiments of this application.

[0123] Please see Figure 10 , Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device 500 includes one or more processors 501 and a memory 502. The memory 502 is connected to one or more processors 501, for example, via a bus.

[0124] Processor 501 is configured to support the electronic device 500 in performing the corresponding functions in the methods described in the above method embodiments. Processor 501 may be a central processing unit (CPU), a network processor (NP), a hardware chip, or any combination thereof. The aforementioned hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0125] Memory 502 is used to store program code, etc. Memory 502 may include volatile memory (VM), such as random access memory (RAM); memory 502 may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); memory 502 may also include combinations of the above types of memory.

[0126] The memory 502 can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the classroom data classification method in the embodiments of this application. The processor 501 executes various functional applications and data processing of the classroom data classification method and classroom data classification device by running the non-volatile software programs, instructions, and modules stored in the memory 502, that is, it realizes the functions of each module or unit of the classroom data classification method and classroom data classification device provided in the above method embodiments.

[0127] The memory 502 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function. The data storage area may store data created based on the use of the classroom data sorting device, etc. In some embodiments, the memory may include memory remotely located relative to the processor 501, and this remote memory may be connected to the classroom data sorting device via a network.

[0128] The one or more modules are stored in the memory 502. When executed by the one or more processors 501, they execute the classroom data classification method in any of the above method embodiments. For example, they execute the method steps described in the above method embodiments to realize the functions of the modules described in the above device embodiments.

[0129] The electronic device in this application embodiment may specifically be an ultra-mobile personal computer device, a server or server cluster, etc.

[0130] This application provides a non-volatile computer-readable storage medium storing computer-executable instructions that are executed by one or more processors, for example... Figure 10 One of the processors 501 can enable the one or more processors to execute the classroom data classification method in any of the above method embodiments.

[0131] This application provides a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium. The computer program includes program instructions that, when executed by the electronic device, enable the electronic device to perform the classroom data classification method in any of the above method embodiments.

[0132] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0133] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A classroom data classification method, characterized in that, include: Acquire classroom data; The classroom data is preprocessed to obtain preprocessed classroom data; Based on the preset classroom structure model, the preprocessed classroom data is decomposed to obtain the teaching links, teaching activities and teaching behaviors corresponding to the preprocessed classroom data; The thematic content is obtained based on the teaching steps and activities described above; Based on the teaching behavior, a preset combination of behaviors is obtained, and the combination of behaviors is filtered according to the topic content to obtain a target combination of behaviors; The target behavior combinations are classified to obtain the classification results corresponding to the classroom data.

2. The method according to claim 1, characterized in that, The preprocessing of the classroom data to obtain preprocessed classroom data includes: Obtain text data based on the classroom data; The text data is processed, including grammar checking, homophone handling, context understanding and optimization, proper noun and terminology recognition, to obtain processed text data; The speaker's identity is determined in the processed text data to obtain preprocessed classroom data, which is optimized text content containing the speaker's identity.

3. The method according to claim 2, characterized in that, The step of breaking down the preprocessed classroom data according to a preset classroom structure model to obtain the teaching segments, teaching activities, and teaching behaviors corresponding to the preprocessed classroom data includes: The optimized text content containing the speaker's identity is input into a preset classroom structure model. The classroom structure model analyzes the optimized text content containing the speaker's identity based on the structure and pattern it has learned, so as to identify the teaching segment. Based on the identified teaching segments, identify the teaching activities contained in each teaching segment; Based on the identified teaching activities, the content and characteristics of the teaching activities are analyzed to identify the teaching behaviors in each teaching activity.

4. The method according to claim 3, characterized in that, The optimized text content containing the speaker's identity is input into a preset classroom structure model. The classroom structure model analyzes the optimized text content containing the speaker's identity based on its learned structures and patterns to identify teaching segments, including: The optimized text content containing the speaker's identity is prepared as data. The data preparation includes text formatting, timestamp sorting, and data cleaning to obtain text content that meets the input requirements of the preset classroom structure model. Load the preset classroom structure model, and input the text content and timestamp information that meet the input requirements into the classroom structure model; Feature extraction is performed on the text content that meets the input requirements to identify key behaviors, and the location of the key behaviors in the classroom data is determined based on the timestamp information; The key behaviors whose locations have been determined are classified and categorized into preset teaching segments, thereby obtaining teaching segments containing specific content.

5. The method according to claim 2, characterized in that, The classroom data also includes video data from the classroom process, and determining the speaker's identity in the processed text data specifically involves: After performing visual analysis on the video data, the students' action data during the classroom process is obtained; By combining the student's action data, the speaker's identity in the processed text data is determined.

6. The method according to claim 1, characterized in that, The process of obtaining thematic content based on the teaching steps and activities includes: Text summarization is performed on the interactive content of the teaching activities to obtain the activity theme of each teaching activity; Based on the activity theme, obtain the theme corresponding to each teaching segment; The theme of the entire lesson is obtained based on the theme of the aforementioned segments; The activity theme, the session theme, and the classroom theme constitute the theme content.

7. The method according to claim 1, characterized in that, The method further includes: Based on the aforementioned theme and content, the teaching process is analyzed, and a summary and recommendations for the teaching process are presented.

8. The method according to any one of claims 1 to 7, characterized in that, The step of obtaining a preset combination of behaviors based on the teaching behaviors, and filtering the combination of behaviors based on the topic content to obtain a target combination of behaviors includes: Based on the teaching behaviors, a question-answer-evaluation behavior combination is obtained; wherein, the question-answer-evaluation behavior combination refers to a sequence of behaviors that includes questioning, answering, and evaluating, extracted from the teaching activities for the topic content; The question-and-answer evaluation behavior combination and the topic content are input into a large language model. The large language model outputs an effective question-and-answer evaluation behavior combination as the target behavior combination. The effective question-and-answer evaluation behavior combination refers to a behavior combination with a score greater than a threshold.

9. The method according to claim 8, characterized in that, The process of classifying the target behavior combinations to obtain the classification results corresponding to the classroom data includes: Obtain the preset classification criteria; The preset classification criteria are input into the large language model to provide a classification framework for the large language model; The effective question-answer-evaluation behavior combination and the corpus corresponding to the topic content are input into the large language model, so that the large language model outputs effective question classification, effective answer classification and effective evaluation classification according to the classification framework.

10. An electronic device, characterized in that, include: A memory and a processor, the memory being connected to the processor, the processor being configured to execute one or more computer programs stored in the memory, the processor, when executing the one or more computer programs, causing the electronic device to implement the classroom data classification method as described in any one of claims 1-9.

Images