Classroom quality evaluation method, computer device, and storage medium

Abstract

The application provides a classroom quality evaluation method, a computer device and a storage medium. The method comprises the following steps: dividing a classroom text according to a classroom teaching event to obtain a plurality of classroom text segments, wherein the classroom text is a text obtained by performing automatic speech recognition on a classroom audio; extracting a classroom text segment belonging to a target evaluation dimension from the plurality of classroom text segments to obtain a target classroom text segment; obtaining an evaluation reference text matched with the target classroom text segment from an evaluation reference database, wherein the evaluation reference text comprises an evaluation guide text and a standard teaching reference text; inputting the target classroom text segment and the evaluation reference text into a preset first large model to perform classroom quality evaluation, and obtaining a quality evaluation result corresponding to the target evaluation dimension; and outputting a classroom quality evaluation text corresponding to the target evaluation dimension according to the quality evaluation result corresponding to the target evaluation dimension. The technical solution improves the classroom evaluation efficiency.

Description

Technical Field

[0001] This application relates to the field of teaching evaluation technology, and in particular to classroom quality evaluation methods, computer equipment, and storage media. Background Technology

[0002] Teaching evaluation is a student-centered classroom assessment activity aimed at improving teaching quality through observation and in-depth analysis of the teaching process. For students, classroom evaluation guides learning direction, clarifies learning goals and requirements, stimulates learning interest and initiative, increases classroom participation, and ensures learning quality. At the teaching management level, classroom evaluation provides a basis for teaching decisions, helps teaching management departments allocate resources rationally and optimize curriculum design, is a key means of monitoring teaching quality, and serves as an important reference for teacher evaluation.

[0003] Traditional classroom assessment usually relies on manual observation and evaluation, which is highly subjective and inefficient. Summary of the Invention

[0004] This application provides classroom quality assessment methods, computer equipment, and storage media, with the aim of improving the efficiency of classroom assessment.

[0005] Firstly, a method for evaluating classroom quality is provided, including:

[0006] The classroom text is segmented according to classroom teaching events to obtain multiple classroom text fragments. The classroom text is obtained by performing automatic speech recognition (ASR) on the classroom audio.

[0007] From the multiple classroom text fragments, extract the classroom text fragments that belong to the target evaluation dimension to obtain the target classroom text fragments;

[0008] From the lesson evaluation reference database, obtain lesson evaluation reference texts that match the target classroom text fragment. The lesson evaluation reference texts include lesson evaluation guidance texts and standard teaching reference texts.

[0009] The target classroom text fragment and the evaluation reference text are input into the preset first large model to evaluate classroom quality, and the quality evaluation results corresponding to the target evaluation dimension are obtained.

[0010] Based on the quality evaluation results corresponding to the target evaluation dimensions, output the classroom quality evaluation text corresponding to the target evaluation dimensions.

[0011] In this technical solution, the classroom audio is first segmented according to classroom teaching events, and the resulting classroom text is obtained through ASR (Automatic Reflection and Evaluation). Multiple classroom text fragments are then extracted from these fragments to identify the target evaluation dimension. Next, evaluation reference texts matching the target text fragments are retrieved from an evaluation reference database. The target text fragments and their matching reference texts are then input into a pre-set first model for classroom quality evaluation, yielding the quality evaluation results corresponding to the target evaluation dimension. Finally, based on these results, the corresponding classroom quality evaluation text is output, achieving automatic output of the classroom quality evaluation text without manual observation or assessment, thus improving classroom evaluation efficiency. The target classroom text fragments belonging to the target evaluation dimensions, along with matching evaluation reference texts, are input into the large model. The fragmented classroom texts provide the large model with clear analytical objectives, preventing it from analyzing irrelevant content and outputting invalid or redundant information. The evaluation reference texts include evaluation guidance texts and standard teaching reference texts. The evaluation guidance texts provide scientific evaluation guidance for the large model, while the standard teaching reference texts provide standard examples. This allows the large model to output more credible classroom quality evaluation results that align with the target evaluation dimensions, improving the accuracy and objectivity of classroom evaluations. Furthermore, inputting standard teaching reference texts into the large model enhances the practicality of the classroom quality evaluation results, giving them practical value.

[0012] In conjunction with the first aspect, in one possible implementation, the step of outputting the classroom quality evaluation text corresponding to the target evaluation dimension based on the quality evaluation result corresponding to the target evaluation dimension includes: performing a reliability verification on the quality evaluation result corresponding to the target evaluation dimension; and if the verification passes, outputting the quality evaluation result corresponding to the target evaluation dimension as the classroom quality evaluation text corresponding to the target evaluation dimension.

[0013] If the quality evaluation results output by the large model pass the reliability verification, outputting the quality evaluation results of the large model as the corresponding classroom quality evaluation text for the evaluation dimension helps to ensure the accuracy of the classroom quality evaluation text.

[0014] In conjunction with the first aspect, in one possible implementation, the reliability verification of the quality evaluation results corresponding to the target evaluation dimension includes: inputting the quality evaluation results corresponding to the target evaluation dimension and the target classroom text fragment into a preset second large model for reliability verification, and obtaining the verification result.

[0015] Inputting the quality evaluation results output by the model and the target classroom text fragments back into the large model for reliability verification helps to ensure the reliability of the classroom quality evaluation text.

[0016] In conjunction with the first aspect, in one possible implementation, the step of inputting the quality evaluation result corresponding to the target evaluation dimension and the target classroom text fragment into a preset second large model for reliability verification to obtain the verification result includes: generating a first reasoning guidance prompt based on the quality evaluation result corresponding to the target evaluation dimension, the target classroom text fragment, and a first task instruction text, wherein the first task instruction text is used to instruct the large model to output the verification result and the judgment reason corresponding to the verification result based on the quality evaluation result corresponding to the target evaluation dimension and the target classroom text fragment; and inputting the first reasoning guidance prompt into the second large model for logical reasoning to obtain the verification result.

[0017] Based on the quality evaluation results corresponding to the target lesson evaluation dimensions, the target classroom text fragments, and the first task instruction text, a first reasoning guidance instruction is generated, which can instruct the large model to output the verification results and judgment reasons, thereby improving the reliability of the verification.

[0018] In conjunction with the first aspect, in one possible implementation, obtaining the evaluation reference text matching the target classroom text fragment from the evaluation reference database includes: obtaining standard teaching reference text with the course tag of the target course tag from the evaluation reference database, as the standard teaching reference text matching the target classroom text fragment, wherein the target course tag is used to indicate the subject and / or grade level and / or knowledge chapter to which the target classroom text fragment belongs; and obtaining evaluation guidance text with a similarity greater than a similarity threshold to the target classroom text fragment from the evaluation reference database, as the evaluation guidance text matching the target classroom text fragment.

[0019] By obtaining standard teaching reference texts with the same course tags as the corresponding classroom text fragments, and by obtaining evaluation guidance texts with a high degree of similarity to the target classroom text fragments, it is beneficial to obtain evaluation reference texts with a high degree of correlation with the classroom text fragments, thereby helping the large model output more accurate quality evaluation results.

[0020] In conjunction with the first aspect, in one possible implementation, each of the plurality of classroom text fragments has a content tag, the content tag being used to indicate the content category to which the classroom text fragment belongs; the step of extracting classroom text fragments belonging to the target evaluation dimension from the plurality of classroom text fragments to obtain the target classroom text fragment includes: extracting classroom text fragments with content tags as target content tags from the plurality of classroom text fragments to obtain the target classroom text fragment, wherein the target content tag is the content tag corresponding to the target evaluation dimension.

[0021] By extracting classroom text fragments whose content tags are the same as those corresponding to the target evaluation dimensions, we can make the target classroom content fragments strongly correlated with the evaluation dimensions, thus providing a clear analysis object for the large model.

[0022] In conjunction with the first aspect, in one possible implementation, the step of inputting the target classroom text fragment and the evaluation reference text into a preset first large model for classroom quality evaluation to obtain the quality evaluation result corresponding to the target evaluation dimension includes: generating a second reasoning guidance prompt based on the target classroom text fragment, the evaluation reference text, and the second task instruction text, wherein the second task instruction text is used to instruct the large model to use the evaluation reference text as the evaluation basis and, in conjunction with the target classroom text, output a quality evaluation result corresponding to the target evaluation dimension and containing the evaluation basis; and inputting the second reasoning guidance prompt into the first large model for logical reasoning to obtain the quality evaluation result corresponding to the target evaluation dimension.

[0023] Based on the target classroom text fragment, the evaluation reference text, and the second task instruction text, a second reasoning guidance instruction is generated. This instruction can guide the large model to output quality evaluation results containing the evaluation criteria, which helps the large model output reliable quality evaluation results and improves the accuracy of classroom evaluation.

[0024] In conjunction with the first aspect, in one possible implementation, before inputting the target classroom text fragment and the evaluation reference text into a preset first large model for classroom quality evaluation to obtain the quality evaluation result corresponding to the target evaluation dimension, the method further includes: performing classroom behavior recognition on the classroom video corresponding to the classroom text to obtain classroom behavior data, wherein the classroom behavior data is used to describe the classroom behavior in the classroom video; the step of inputting the target classroom text fragment and the evaluation reference text into the preset first large model for classroom quality evaluation to obtain the quality evaluation result corresponding to the target evaluation dimension includes: inputting the target classroom text fragment, the evaluation reference text, and the classroom behavior data into the first large model for classroom quality evaluation to obtain the quality evaluation result corresponding to the target evaluation dimension.

[0025] By inputting classroom behavior data, along with target classroom text fragments and evaluation reference texts, into a large model for classroom quality evaluation, the classroom behavior data can supplement the missing information in the classroom text fragments, enrich the input information, and help improve the accuracy of the quality evaluation results output by the large model.

[0026] Secondly, a classroom quality assessment device is provided, comprising:

[0027] The text segmentation module is used to segment classroom text according to classroom teaching events to obtain multiple classroom text fragments. The classroom text is the text obtained by automatic speech recognition of classroom audio.

[0028] The extraction module is used to extract classroom text fragments belonging to the target evaluation dimension from the multiple classroom text fragments to obtain the target classroom text fragments;

[0029] The standard text acquisition module is used to acquire lesson evaluation reference texts that match the target classroom text fragments from the lesson evaluation reference database. The lesson evaluation reference texts include lesson evaluation guidance texts and standard teaching reference texts.

[0030] The classroom quality evaluation module is used to input the target classroom text fragment and the evaluation reference text into a preset first large model to evaluate the classroom quality and obtain the quality evaluation results corresponding to the target evaluation dimension.

[0031] The text output module is used to output the classroom quality evaluation text corresponding to the target evaluation dimension based on the quality evaluation results corresponding to the target evaluation dimension.

[0032] Thirdly, a computer device is provided, including 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, wherein when the processor executes the one or more computer programs, the computer device performs the classroom quality assessment method described in the first aspect.

[0033] Fourthly, a computer-readable storage medium is provided, which stores a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform the classroom quality evaluation method of the first aspect.

[0034] This application can achieve the following technical effects: It inputs target classroom text fragments belonging to the target evaluation dimension, along with matching evaluation reference texts, into a large model. The fragmented classroom texts provide the large model with clear analytical objectives, preventing it from analyzing irrelevant content and outputting invalid or redundant information. The evaluation reference texts include evaluation guidance texts and standard teaching reference texts. The evaluation guidance texts provide scientific evaluation guidance for the large model, while the standard teaching reference texts provide standard examples. This allows the large model to output more credible classroom quality evaluation results that conform to the target evaluation dimension, improving the accuracy and objectivity of classroom evaluation. Attached Figure Description

[0035] 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.

[0036] Figure 1 A flowchart illustrating a classroom quality evaluation method provided in this application embodiment;

[0037] Figure 2 A schematic diagram illustrating the composition of a reasoning guidance prompt for classroom quality assessment provided in an embodiment of this application;

[0038] Figure 3 A flowchart illustrating another classroom quality assessment method provided in this application embodiment;

[0039] Figure 4 A schematic diagram illustrating the composition of reasoning guidance prompts required for classroom quality assessment and reliability verification provided in the embodiments of this application;

[0040] Figure 5 This is a schematic diagram of the structure of a classroom quality evaluation device provided in an embodiment of this application;

[0041] Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0042] 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.

[0043] 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.

[0044] The technical solution of this application is applicable to classroom assessment scenarios. In classroom assessment scenarios, it is necessary to evaluate classroom quality from multiple dimensions, such as teaching objectives, teaching methods, teaching process, and teaching effectiveness.

[0045] Traditional lesson evaluation methods usually rely on manual assessment, and the evaluation results are greatly affected by human subjective factors, which makes them less objective. In addition, they require a lot of time for manual observation and evaluation, resulting in low evaluation efficiency.

[0046] In some technical solutions, classroom quality evaluation is based on automated systems. These automated systems are statistical models based on keyword recognition or behavior frequency, or behavioral analysis models based on deep learning. However, these analysis systems have difficulty in deeply understanding the teaching context in a classroom setting, and they are also difficult to adapt to the characteristics of the subject and specific evaluation standards. As a result, the output evaluation content may not reflect the actual situation.

[0047] In view of this, this application proposes a classroom quality evaluation scheme, constructing an automatic evaluation system consisting of a multimodal event extraction module, an educational knowledge enhancement module, and a large-scale model. The multimodal event extraction module is used to segment classroom text according to classroom teaching events, and extract classroom text fragments belonging to the target evaluation dimension from the segmented classroom text fragments as target classroom text fragments to be input into the large-scale model. The educational knowledge enhancement module is used to obtain evaluation reference texts matching the target classroom text fragments from the evaluation reference database as classroom evaluation reference data to be input into the large-scale model. The large-scale model is used to output the classroom evaluation results, with the target classroom text fragments replacing the entire text. Using standard teaching reference data as the analysis object of the large model helps clarify its analytical objectives. This data provides scientific guidance and standard examples for classroom evaluation, standardizing the model's output and preventing factual errors, unfounded opinions, conceptual confusion, or non-compliance with teaching evaluation standards. This improves the accuracy, objectivity, and relevance of classroom evaluations. Furthermore, inputting standard teaching reference texts into the large model enhances the practicality of its output, making the evaluation results more valuable. Automated evaluation systems also improve efficiency. Additionally, this application uses a verification module to further validate the reliability of the large model's output. If the verification passes, the large model's output is used as the classroom quality evaluation text, further enhancing its reliability.

[0048] The technical solution of this application is described in detail below.

[0049] See Figure 1 , Figure 1 This is a flowchart illustrating a classroom quality evaluation method provided in an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps:

[0050] S101, divide the classroom text according to classroom teaching events to obtain multiple classroom text fragments.

[0051] Here, classroom text refers to the text used to represent the dialogue content in the classroom. Classroom text is obtained by automatic speech recognition of classroom audio.

[0052] The process involves acquiring classroom audio, performing ASR (Automatic Recognition) on the audio to obtain classroom dialogue text, and then preprocessing this text to obtain the classroom text. Text preprocessing includes removing interjections, repetitive words, and words irrelevant to the lesson content. Interjections include phrases like "um" and "ah," while irrelevant words include imperative phrases like "class begins" and "get out of class ends." This preprocessing reduces redundant information in the classroom text and improves the model's ability to understand it.

[0053] Segmenting classroom texts into multiple text fragments based on classroom teaching events refers to using classroom teaching events as the basic analytical element to identify and understand the content of the classroom text, dividing the text into multiple text fragments corresponding to each teaching event, such that any two adjacent text fragments correspond to different teaching events. In this application, classroom teaching events are defined based on teaching interaction behaviors (including but not limited to teacher questioning, student answering, teacher feedback, and teacher assignment of classroom tasks) and various teaching tasks (including but not limited to interactive teaching tasks, explanatory teaching tasks, and experiential teaching tasks) under PTA theory. One text fragment reflects one classroom teaching event, and one classroom teaching event corresponds to one teaching behavior or one teaching task. PTA theory refers to the theory that divides the classroom structure into three levels: Procedure, Task, and Action.

[0054] In some embodiments, classroom text is input into a multimodal event extraction module, and the segmentation module within the multimodal event extraction module segments the classroom text into multiple classroom text fragments. The segmentation module within the multimodal event extraction module is an algorithm module based on deep learning. It performs semantic recognition and classification on the classroom text, outputting multiple classroom text fragments and content tags for each fragment. The content tags indicate the content category to which the classroom text belongs.

[0055] In a specific implementation scenario, content can be categorized into three types: teacher-student dialogue, teacher-student interaction, and student learning. If the content label of a classroom text fragment indicates that the content category belongs to teacher-student dialogue, it means that the classroom text fragment reflects the questions and answers between the teacher and students in the classroom. If the content label of a classroom text fragment indicates that the content category belongs to teacher-student interaction, it means that the classroom text fragment reflects the dialogue content during classroom interaction between the teacher and students. If the content label of a classroom text fragment indicates that the content category belongs to student learning, it means that the classroom text fragment reflects the teaching content delivered by the teacher. In this specific implementation scenario, classroom teaching events can be divided into teacher-student dialogue events, teacher-student interaction events, and student learning events. The content labels include three types of content labels. For example, the three content labels are label1, label2, and label3. label1 indicates that the content category belongs to teacher-student dialogue, label2 indicates that the content category belongs to teacher-student interaction, and label3 indicates that the content category belongs to student learning.

[0056] S102, extract classroom text fragments belonging to the target evaluation dimension from multiple classroom text fragments to obtain the target classroom text fragments.

[0057] Here, the evaluation dimensions refer to the dimensions involved in evaluating the quality of classroom teaching. There are usually multiple evaluation dimensions, which are divided based on educational and teaching theories, evaluation standards, teaching practice experience, and evaluation purposes.

[0058] In a specific implementation scenario, multiple evaluation dimensions include eight dimensions: classroom overview, key knowledge points, evaluation of classroom objectives, effectiveness of questioning, suggestions for questioning, evaluation of interaction, suggestions for interaction, and student learning feedback. These eight evaluation dimensions can meet the evaluation requirements of four aspects: student learning, teacher teaching, course nature, and classroom culture. That is, they meet the observation requirements of the LICC observation model and ensure the comprehensiveness of the evaluation dimensions.

[0059] The target evaluation dimension can be any one of multiple evaluation dimensions. For example, if multiple evaluation dimensions include the aforementioned class overview, key knowledge points, evaluation of class objectives, effectiveness of questioning, suggestions for questioning, evaluation of interaction, suggestions for interaction, and student learning feedback, then the target evaluation dimension could be either the class overview, key knowledge points, evaluation of class objectives, effectiveness of questioning, suggestions for questioning, evaluation of interaction, suggestions for interaction, or student learning feedback.

[0060] In one embodiment, the extraction submodule within the multimodal event extraction module extracts classroom text fragments belonging to the target evaluation dimension from multiple classroom text fragments to obtain the target classroom text fragment. Since each classroom text fragment has a content tag, the extraction submodule within the multimodal event extraction module extracts classroom text fragments whose content tag matches the target content tag from the multiple classroom text fragments to obtain the target classroom text fragment.

[0061] The target content tags are the content tags corresponding to the target evaluation dimensions, which are obtained through pre-setting. Taking multiple evaluation dimensions, including the aforementioned class overview, key knowledge points, class objective evaluation, question effectiveness, question suggestions, interaction evaluation, interaction suggestions, and student learning feedback, as an example, the correspondence between the pre-set evaluation dimensions, content tags, and the corresponding content categories is shown in Table 1.

[0062] Evaluation Dimensions Content Tags Content Classification Class Overview label1, label2, label3 Teacher-student dialogue, teacher-student interaction, and student learning Key knowledge points label1, label2, label3 Teacher-student dialogue, teacher-student interaction, and student learning Classroom Objectives Assessment label1 Teacher-student dialogue Question validity label1 Teacher-student dialogue Questions and suggestions label1 Teacher-student dialogue Interactive reviews label2 Teacher-student interaction Interactive suggestions label2 Teacher-student interaction Student Learning Feedback label3 Student learning

[0063] Table 1

[0064] Assuming the target evaluation dimension is the effectiveness of questions, the content label corresponding to the target evaluation dimension is label1. From multiple classroom text fragments, extract the classroom text fragment with content label1 to obtain the target classroom text fragment.

[0065] By extracting classroom text fragments whose content tags are the same as those corresponding to the target evaluation dimensions, we can make the target classroom content fragments strongly correlated with the evaluation dimensions, thus providing a clear analysis object for the large model.

[0066] S103. Obtain the lesson evaluation reference text that matches the target classroom text fragment from the lesson evaluation reference database.

[0067] Here, the reference texts for lesson evaluation include lesson evaluation guidance texts and standard teaching reference texts. Lesson evaluation guidance texts are texts that define the nature, objectives, content framework, academic quality standards, and implementation suggestions of a course. They serve as guidance for classroom evaluation and include national curriculum standards, such as the new curriculum standards. Standard teaching reference texts are texts compiled around a specific textbook, including teaching objectives, instructional design, and teaching suggestions. They provide teachers with reference examples for interactive teaching.

[0068] The lesson evaluation reference database stores lesson evaluation guidance texts and standard teaching reference texts. These texts are broken down into text fragments and stored within the database. The database also stores course tags corresponding to the standard teaching reference texts. These tags indicate the course attributes of the lesson evaluation texts, including subject (including but not limited to Chinese, Mathematics, and English), and / or grade level (including but not limited to primary and junior high school), and / or knowledge chapter.

[0069] The evaluation reference texts that match the target classroom text fragments include evaluation guidance texts that match the target classroom text fragments and standard teaching reference texts that match the target classroom text fragments.

[0070] In some embodiments, the evaluation reference text matching the target classroom text fragment is obtained from the evaluation reference database through the following steps A1-A2:

[0071] A1. Obtain the standard teaching reference text with the target course tag from the course evaluation reference database, and use it as the standard teaching reference text to match the target classroom text fragment.

[0072] Here, the target course label is used to indicate the subject and / or grade level and / or knowledge chapter to which the target classroom text fragment belongs.

[0073] In one feasible implementation, the target classroom text fragment is matched with chapter-related text in a chapter search database to obtain a target course tag. The chapter search database stores multiple chapter names, the subjects and / or grade levels to which the multiple chapter names belong, and the chapter-related text corresponding to each of the multiple chapter names. The chapter-related text is the text associated with the chapter name and used to introduce the knowledge of the chapter. The target course tag includes the chapter name corresponding to the target chapter-related text in the chapter search database, as well as the subject and / or grade level to which the target chapter-related text belongs. The target chapter-related text is the chapter-related text with the highest similarity to the target classroom text fragment.

[0074] Alternatively, the target classroom text fragment can be input into a preset classification model for course tag classification to obtain the target course tag. This application does not limit the specific implementation method for determining the target course tag.

[0075] After obtaining the target course label, the course labels of each standard teaching reference text fragment in the course evaluation reference database are traversed. The standard teaching reference text composed of standard teaching reference text fragments whose course labels are the target course labels is used as the standard teaching reference text that matches the target classroom text fragment.

[0076] A2. From the lesson evaluation reference database, obtain lesson evaluation guidance texts whose similarity to the target classroom text fragment is greater than the similarity threshold, and use them as lesson evaluation guidance texts that match the target classroom text fragment.

[0077] In one feasible implementation, the similarity between the target text fragment and each lesson evaluation guidance text fragment in the lesson evaluation reference database is calculated based on the BM25 algorithm. The lesson evaluation guidance text composed of lesson evaluation text fragments with a similarity greater than the similarity threshold is used as the lesson evaluation guidance text that matches the target classroom text fragment.

[0078] In another feasible implementation, text features of the target classroom text segment can be extracted to obtain target text features. The similarity between the target text features and the text features corresponding to each lesson evaluation guidance text segment in the lesson evaluation reference database can be calculated. Lesson evaluation guidance texts composed of lesson evaluation guidance text segments with similarity greater than a similarity threshold are used as the lesson evaluation guidance texts matched with the target classroom text segment. Specifically, the cosine similarity or Jaccard similarity between the text features and the text features corresponding to each lesson evaluation guidance text segment in the lesson evaluation reference database is calculated to determine the similarity between the target text features and the text features corresponding to each lesson evaluation guidance text segment in the lesson evaluation reference database. This application does not limit the specific implementation method for obtaining lesson evaluation guidance texts with similarity greater than a similarity threshold to the target classroom text segment.

[0079] In steps A1-A2 above, by obtaining standard teaching reference texts with the same course tags as the classroom text fragments, and by obtaining evaluation guidance texts with a high similarity to the target classroom text fragments, it is beneficial to obtain evaluation reference texts with a high degree of correlation with the classroom text fragments, thereby helping the large model output more accurate quality evaluation results.

[0080] S104: Input the target classroom text fragment and the evaluation reference text matching the target classroom text fragment into the preset first large model to evaluate the classroom quality and obtain the quality evaluation results corresponding to the target evaluation dimensions.

[0081] Here, the primary model is an AI model that combines natural language processing technology to understand and process various types of text and images, generating corresponding text descriptions based on user needs and mimicking human reasoning processes. This primary model includes, but is not limited to, Qwen-VL and InternVL2. In this application, the primary model is used to simulate the human reasoning process regarding classroom quality assessment and output quality assessment results.

[0082] In one feasible implementation, the target classroom text fragment and the evaluation reference text matching the target classroom text fragment are input into the preset first large model for classroom quality evaluation through the following steps B1-B2, so as to obtain the quality evaluation results corresponding to the target evaluation dimension:

[0083] B1. Generate second reasoning guidance prompts based on the target classroom text fragment, the evaluation reference text matching the target classroom text fragment, and the second task instruction text.

[0084] The second task instruction text instructs the large model to use the evaluation reference text that matches the target classroom text fragment as the evaluation basis, and to output a quality evaluation result corresponding to the target evaluation dimension, including the evaluation basis, in conjunction with the target classroom text. An example of the second task instruction text is as follows:

[0085] "You are a professional teaching evaluation expert. Please use the following evaluation reference text [xxx] as the basis for your evaluation, and combine it with the classroom text excerpt [yyy] to conduct a classroom quality evaluation in the target evaluation dimensions. You need to strictly follow the evaluation criteria to give evaluation suggestions, and indicate which parts of the evaluation reference text the evaluation suggestions are derived from."

[0086] The second reasoning guidance prompt includes the target classroom text fragment, a lesson evaluation reference text matching the target classroom text fragment, and the second task instruction text. For example, the components of the second reasoning guidance prompt are as follows: Figure 2 As shown.

[0087] B2. Input the second reasoning guidance prompt into the first major model for logical reasoning to obtain the quality evaluation results corresponding to the target evaluation dimensions.

[0088] In steps B1-B2 above, a second reasoning guidance instruction is generated based on the target classroom text fragment, the evaluation reference text, and the second task instruction text. This instruction instructs the large model to output quality evaluation results containing evaluation criteria, which helps the large model output reliable quality evaluation results and improves the accuracy of classroom evaluation.

[0089] S105, based on the quality evaluation results corresponding to the target evaluation dimensions, output the classroom quality evaluation text corresponding to the target evaluation dimensions.

[0090] In one feasible implementation, the quality evaluation results corresponding to the target evaluation dimensions are output as the classroom quality evaluation text corresponding to the target evaluation dimensions.

[0091] In the above Figure 1In the corresponding technical solution, the classroom audio is first segmented according to classroom teaching events and ASR is performed to obtain classroom text, resulting in multiple classroom text fragments. Then, classroom text fragments belonging to the target evaluation dimension are extracted from the multiple classroom text fragments to obtain the target classroom text fragment. Next, evaluation reference texts matching the target classroom text fragments are obtained from the evaluation reference database. Then, the target classroom text fragments and the evaluation reference texts matching the target classroom text fragments are input into the preset first large model for classroom quality evaluation to obtain the quality evaluation results corresponding to the target evaluation dimension. Finally, based on the quality evaluation results corresponding to the target evaluation dimension, the classroom quality evaluation text corresponding to the target evaluation dimension is output, realizing the automatic output of classroom quality evaluation text without manual observation and evaluation, thus improving the efficiency of classroom evaluation. The target classroom text fragments belonging to the target evaluation dimensions, along with matching evaluation reference texts, are input into the large model. The fragmented classroom texts provide the large model with clear analytical objectives, preventing it from analyzing irrelevant content and outputting invalid or redundant information. The evaluation reference texts include evaluation guidance texts and standard teaching reference texts. The evaluation guidance texts provide scientific evaluation guidance for the large model, while the standard teaching reference texts provide standard examples. This allows the large model to output more credible classroom quality evaluation results that align with the target evaluation dimensions, improving the accuracy and objectivity of classroom evaluations. Furthermore, inputting standard teaching reference texts into the large model enhances the practicality of the classroom quality evaluation results, giving them practical value.

[0092] See Figure 3 , Figure 3 A flowchart illustrating another classroom quality assessment method provided in this application embodiment is shown below. Figure 3 As shown, the method includes the following steps:

[0093] S201, divide the classroom text according to classroom teaching events to obtain multiple classroom text fragments.

[0094] S202, extract classroom text fragments belonging to the target evaluation dimension from multiple classroom text fragments to obtain the target classroom text fragments.

[0095] S203. Obtain the lesson evaluation reference text that matches the target classroom text fragment from the lesson evaluation reference database.

[0096] For the specific implementation of steps S201-S203, please refer to the description of steps S101-S103 above, which will not be repeated here.

[0097] S204: Perform classroom behavior recognition on the classroom video corresponding to the classroom text to obtain classroom behavior data.

[0098] Here, the classroom video corresponding to the classroom text refers to the classroom video belonging to the same course as the classroom audio corresponding to the classroom text.

[0099] Classroom behavior data describes classroom behaviors in the corresponding classroom videos. This data includes classroom behavior text and / or classroom behavior segments. Classroom behavior text refers to text describing classroom behaviors, including but not limited to texts such as "group discussion," "going to the front," "practice," "reading aloud," and "experiment." Classroom behavior segments refer to video clips corresponding to specific classroom behaviors, such as video clips of students going to the front, students discussing, and students practicing.

[0100] The process involves inputting the classroom video corresponding to the classroom text into a pre-defined visual model for classroom behavior reasoning and recognition, thereby obtaining classroom behavior data. The visual model is an artificial intelligence model that integrates computer vision technology and large-scale deep learning. Based on deep learning architectures such as convolutional neural networks and recurrent neural networks (Transformers), the visual model extracts and recognizes features from the classroom video and outputs classroom behavior data.

[0101] S205: Input the target classroom text fragment, the evaluation reference text matching the target classroom text fragment, and classroom behavior data into the preset first model to evaluate classroom quality and obtain the quality evaluation results corresponding to the target evaluation dimensions.

[0102] For an introduction to the first major model, please refer to the description of step S104 above.

[0103] In one feasible implementation, the target classroom text fragment, the evaluation reference text matching the target classroom text fragment, and classroom behavior data are input into a preset first large model for classroom quality evaluation through the following steps C1-C2, thereby obtaining the quality evaluation results corresponding to the target evaluation dimensions:

[0104] C1. Generate third reasoning guidance prompts based on the target classroom text fragment, the evaluation reference text matching the target classroom text fragment, classroom behavior data, and the third task instruction text.

[0105] Here, the third task instruction text instructs the large model to use the lesson evaluation reference text as the evaluation basis, and to output a quality evaluation result corresponding to the target lesson evaluation dimension, including the evaluation basis, by combining the target classroom text and classroom behavior data. An example of the third task instruction text is as follows:

[0106] "You are a professional teaching evaluation expert. Please use the following evaluation reference text [xxx] as the basis for your evaluation, and combine it with classroom text excerpts [yyy] and classroom behavior data [zzz] to conduct a classroom quality evaluation in the target evaluation dimensions. You must strictly follow the evaluation criteria to provide evaluation suggestions, and indicate which parts of the evaluation reference text the evaluation suggestions are derived from."

[0107] The third reasoning guidance prompt includes the target classroom text fragment, classroom behavior data, evaluation reference text matching the target classroom text fragment, and the third task instruction text. For example, the components of the third reasoning guidance prompt are as follows: Figure 4 As shown in T1.

[0108] C2. Input the third reasoning guidance prompt into the first major model for logical reasoning to obtain the quality evaluation results corresponding to the target evaluation dimensions.

[0109] In steps C1-C2 above, classroom behavior data, along with target classroom text fragments and evaluation reference texts, are input into the large model for classroom quality evaluation. Classroom behavior data can fill in the missing information in the classroom text fragments, enrich the input information, and help improve the accuracy of the quality evaluation results output by the large model.

[0110] S206, perform reliability verification on the quality evaluation results corresponding to the target evaluation dimensions output by the first major model.

[0111] Because the output of the large model is uncertain, there may be factual errors, unfounded evaluation opinions, conceptual confusion, or non-compliance with teaching evaluation standards. Therefore, the reliability of the quality evaluation results corresponding to the target evaluation dimensions output by the first large model is verified in order to correct the output results of the first large model.

[0112] In some embodiments, the quality evaluation results corresponding to the target lesson evaluation dimensions and the target classroom text fragments are input into a pre-defined second large model for reliability verification, and the verification results are obtained. The second large model and the first large model can be the same large model, or the second large model and the first large model can be different large models.

[0113] Inputting the quality evaluation results output by the model and the target classroom text fragments back into the large model for reliability verification helps to ensure the reliability of the classroom quality evaluation text.

[0114] Specifically, the quality evaluation results corresponding to the target lesson evaluation dimensions and the target classroom text fragments are input into the pre-set second large model through the following steps D1-D2 for reliability verification, and the verification results are obtained:

[0115] D1. Generate the first reasoning guidance prompt based on the quality evaluation results corresponding to the target lesson evaluation dimensions, the target classroom text fragments, and the first task instruction text.

[0116] Here, the first task instruction text instructs the large model to output verification results and corresponding reasoning based on the quality evaluation results corresponding to the target lesson evaluation dimensions and the target classroom text fragments. An example of the first task instruction text is as follows:

[0117] "The quality evaluation result [uuu] corresponding to the target lesson evaluation dimension is the evaluation suggestion and the evaluation basis of the evaluation suggestion output by evaluating the target classroom text fragment [yyy] according to the target lesson evaluation dimension. Please judge whether the quality evaluation result [uuu] corresponding to the target lesson evaluation dimension is accurate as a professional classroom evaluator. If it is accurate, please output yes and give the reason for the judgment. If it is inaccurate, please output no and give the reason for the judgment."

[0118] The first reasoning guidance prompt includes the quality evaluation results corresponding to the target evaluation dimensions, the target classroom text excerpt, and the first task instruction text. For example, the components of the first reasoning guidance prompt are as follows: Figure 4 As shown in T2.

[0119] D2. Input the first reasoning prompt into the second model for logical reasoning and obtain the verification result.

[0120] In steps D1-D2 above, based on the quality evaluation results corresponding to the target lesson evaluation dimensions, the target classroom text fragments, and the first task instruction text, a first reasoning guidance instruction is generated, which can instruct the large model to output the verification results and judgment reasons, thereby improving the reliability of the verification.

[0121] S207. If the verification passes, the quality evaluation results corresponding to the target evaluation dimension output by the first major model will be used as the classroom quality evaluation text corresponding to the target evaluation dimension.

[0122] In the above Figure 3 In the corresponding technical solution, after inputting the target classroom text fragment, the evaluation reference text matching the target classroom text fragment, and classroom behavior data into the preset first large model for classroom quality evaluation, and obtaining the quality evaluation results corresponding to the target evaluation dimensions, the reliability of the quality evaluation results corresponding to the target evaluation dimensions output by the first large model is corrected. If the verification passes, the quality evaluation results corresponding to the target evaluation dimensions output by the first large model are used as the classroom quality evaluation text corresponding to the target evaluation dimensions, which helps to ensure the accuracy of the classroom quality evaluation text.

[0123] The method of this application has been described above; the apparatus of this application will be described below.

[0124] See Figure 5 , Figure 5 This is a schematic diagram of the classroom quality assessment device provided in the embodiments of this application, such as... Figure 5 As shown, the classroom quality assessment device 30 includes:

[0125] The text segmentation module 301 is used to segment classroom text according to classroom teaching events to obtain multiple classroom text fragments, wherein the classroom text is the text obtained by automatic speech recognition of classroom audio.

[0126] The extraction module 302 is used to extract classroom text fragments belonging to the target evaluation dimension from the plurality of classroom text fragments to obtain the target classroom text fragments;

[0127] The standard text acquisition module 303 is used to acquire, from the lesson evaluation reference database, a lesson evaluation reference text that matches the target classroom text fragment, wherein the lesson evaluation reference text includes lesson evaluation guidance text and standard teaching reference text;

[0128] The classroom quality evaluation module 304 is used to input the target classroom text fragment and the evaluation reference text into a preset first large model to evaluate the classroom quality and obtain the quality evaluation results corresponding to the target evaluation dimension.

[0129] The text output module 305 is used to output the classroom quality evaluation text corresponding to the target evaluation dimension based on the quality evaluation results corresponding to the target evaluation dimension.

[0130] It should be noted that the classroom quality evaluation device 30 described above can execute the classroom quality evaluation method 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 can be found in the classroom quality evaluation method provided in the embodiments of this application.

[0131] See Figure 6 , Figure 6 This is a schematic diagram of the structure of a computer device 40 provided in an embodiment of this application. The computer device 40 includes a processor 401 and a memory 402. The memory 402 is connected to the processor 401, for example, via a bus.

[0132] Processor 401 is configured to support the computer device 40 in performing the corresponding functions in the methods described in the above method embodiments. Processor 401 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.

[0133] Memory 402 is used to store program code, etc. Memory 402 may include volatile memory (VM), such as random access memory (RAM); memory 402 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 402 may also include combinations of the above types of memory.

[0134] The memory 402 is used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the classroom quality evaluation method in the embodiments of this application. The processor executes various functional applications and data processing of the classroom quality evaluation method by running the non-volatile software programs, instructions, and modules stored in the memory, thereby realizing the functions of the classroom quality evaluation method provided in the above method embodiments.

[0135] The memory 402 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 quality assessment device. In some embodiments, the memory may include memory remotely located relative to the processor, which can be connected to the classroom quality assessment device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0136] The one or more modules are stored in the memory. When executed by the one or more processors, they perform the classroom quality evaluation method in any of the above method embodiments. For example, they perform the method steps described in the above method embodiments to realize the functions of the modules described in the above device embodiments.

[0137] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the method described in the foregoing embodiments.

[0138] 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 associated 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.

[0139] 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 quality evaluation method, characterized in that, include: The classroom text is segmented according to classroom teaching events to obtain multiple classroom text fragments. The classroom text is obtained by automatic speech recognition of classroom audio. From the multiple classroom text fragments, extract the classroom text fragments that belong to the target evaluation dimension to obtain the target classroom text fragments; From the lesson evaluation reference database, obtain lesson evaluation reference texts that match the target classroom text fragment. The lesson evaluation reference texts include lesson evaluation guidance texts and standard teaching reference texts. The target classroom text fragment and the evaluation reference text are input into the preset first large model to evaluate classroom quality, and the quality evaluation results corresponding to the target evaluation dimension are obtained. Based on the quality evaluation results corresponding to the target evaluation dimensions, output the classroom quality evaluation text corresponding to the target evaluation dimensions.

2. The method according to claim 1, characterized in that, The step of outputting the classroom quality evaluation text corresponding to the target evaluation dimension based on the quality evaluation results of the target evaluation dimension includes: The reliability of the quality evaluation results corresponding to the target evaluation dimensions is verified. If the verification passes, the quality evaluation results corresponding to the target evaluation dimension are output as the classroom quality evaluation text corresponding to the target evaluation dimension.

3. The method according to claim 2, characterized in that, The reliability verification of the quality evaluation results corresponding to the target evaluation dimensions includes: The quality evaluation results corresponding to the target lesson evaluation dimensions and the target classroom text fragments are input into the preset second large model for reliability verification, and the verification results are obtained.

4. The method according to claim 3, characterized in that, The process of inputting the quality evaluation results corresponding to the target lesson evaluation dimensions and the target classroom text fragments into a pre-set second large model for reliability verification, and obtaining the verification results, includes: Based on the quality evaluation results corresponding to the target evaluation dimensions, the target classroom text fragment, and the first task instruction text, a first reasoning guidance prompt is generated. The first task instruction text is used to instruct the large model to output the verification results and the judgment reasons corresponding to the verification results based on the quality evaluation results corresponding to the target evaluation dimensions and the target classroom text fragment. The first reasoning guidance prompt is input into the second large model for logical reasoning to obtain the verification result.

5. The method according to claim 1, characterized in that, The step of obtaining the lesson evaluation reference text matching the target classroom text fragment from the lesson evaluation reference database includes: From the course evaluation reference database, obtain the standard teaching reference text with the course tag of the target course tag, as the standard teaching reference text that matches the target classroom text fragment. The target course tag is used to indicate the subject and / or grade level and / or knowledge chapter to which the target classroom text fragment belongs. From the evaluation reference database, evaluation guidance texts with a similarity greater than a similarity threshold to the target classroom text segment are obtained and used as evaluation guidance texts that match the target classroom text segment.

6. The method according to any one of claims 1-5, characterized in that, Each of the plurality of classroom text fragments has a content tag, which is used to indicate the content category to which the classroom text fragment belongs; The step of extracting classroom text fragments belonging to the target evaluation dimension from the plurality of classroom text fragments to obtain the target classroom text fragment includes: From the multiple classroom text fragments, extract the classroom text fragments whose content tags are the target content tags to obtain the target classroom text fragments, where the target content tags are the content tags corresponding to the target evaluation dimension.

7. The method according to any one of claims 1-5, characterized in that, The step of inputting the target classroom text fragment and the evaluation reference text into a preset first large model for classroom quality evaluation, and obtaining the quality evaluation results corresponding to the target evaluation dimension, includes: Based on the target classroom text fragment, the evaluation reference text, and the second task instruction text, a second reasoning guidance prompt is generated. The second task instruction text is used to instruct the large model to use the evaluation reference text as the evaluation basis and, in conjunction with the target classroom text, output a quality evaluation result corresponding to the target evaluation dimension and containing the evaluation basis. The second reasoning guidance prompt is input into the first large model for logical reasoning to obtain the quality evaluation result corresponding to the target evaluation dimension.

8. The method according to any one of claims 1-5, characterized in that, Before inputting the target classroom text fragment and the evaluation reference text into a preset first large model for classroom quality evaluation and obtaining the quality evaluation result corresponding to the target evaluation dimension, the method further includes: Classroom behavior recognition is performed on the classroom video corresponding to the classroom text to obtain classroom behavior data, which is used to describe the classroom behavior in the classroom video. The step of inputting the target classroom text fragment and the evaluation reference text into a preset first large model for classroom quality evaluation, and obtaining the quality evaluation results corresponding to the target evaluation dimension, includes: The target classroom text fragment, the evaluation reference text, and the classroom behavior data are input into the first large model to evaluate classroom quality, and the quality evaluation results corresponding to the target evaluation dimension are obtained.

9. A computer device, characterized in that, The device includes 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, and the processor, when executing the one or more computer programs, causing the computer device to perform the method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1-8.

Images