Teaching case evaluation and analysis system and method

Abstract

The invention relates to the field of teaching case evaluation, and discloses a teaching case evaluation and analysis system and method, which realize systematic management and evaluation of a large number of teaching plans through cooperative work of a plurality of modules. Wherein the teaching plan information acquisition module is responsible for acquiring multimedia data and copywriting data of a teaching plan to be evaluated; the keyword extraction module extracts keywords for the copywriting data; the grouping module groups and ranks all the teaching plans to be evaluated according to the extracted keywords; and the evaluation module carries out teaching plan evaluation by means of an evaluation model according to grouping and sorting results, and finally obtains a teaching plan evaluation analysis result. All the modules are tightly matched, efficient processing of teaching case evaluation and analysis is achieved, and powerful support is provided for teaching quality improvement.

Description

Technical Field

[0001] The present invention relates to the field of teaching case evaluation, and in particular to a teaching case evaluation and analysis system and method. Background Art

[0002] Lesson plans are the core component of classroom teaching and the ultimate embodiment of teacher preparation. They have an important impact on the quality of education and student learning outcomes. With the rise of advanced technologies such as big data and artificial intelligence, the production of lesson plans has gradually become faster and more convenient. Big data technology collects and stores massive teaching case data. By integrating data generated by various teaching links in the school, such as classroom performance records, homework completion, and test scores, it can build a comprehensive and rich teaching case data set, providing a solid foundation for in-depth analysis of teaching cases. Natural language processing (NLP) technology in artificial intelligence technology can efficiently analyze the text content in teaching cases, and can also automatically extract key information in the cases, such as teaching objectives, teaching methods, student feedback, etc. By pre-setting generation rules, users only need to enter information including target teaching materials, target courses, class time constraints, etc., and the system can generate course lesson plans.

[0003] Although the automatic generation of lesson plans reduces the burden of lesson plan design for teachers, the quality of a large number of lesson plans is uneven. The existing lesson plan evaluation mainly relies on manual work. Experts read and analyze the lesson plans, score them based on their subjective experience, and then comprehensively process the scores of multiple experts to obtain the final evaluation of the lesson plans. This method is not only inefficient, but also easily affected by subjective factors, resulting in certain deviations in the accuracy and objectivity of the evaluation results. Lesson plan evaluation has many meanings. It can not only ensure the scientificity and rationality of teaching objectives, but also optimize the selection and organization of teaching content, improve the effectiveness of teaching methods, ensure the rational use of teaching resources, and promote the professional growth of teachers, which ultimately translates into an overall improvement in teaching quality. Therefore, it is particularly necessary to develop a teaching case evaluation and analysis system and method. Summary of the invention

[0004] The present invention aims to provide a teaching case evaluation and analysis system and method, which can systematically manage and evaluate a large number of teaching plans through the collaborative work of keyword extraction, grouping sorting, and feature extraction modules, thereby improving the efficiency of teaching case evaluation and analysis.

[0005] In order to achieve the above object, the present invention adopts the following technical scheme:

[0006] A teaching case evaluation and analysis system, comprising:

[0007] The lesson plan information acquisition module is used to obtain the associated data of the lesson plan to be evaluated, the associated data includes multimedia data and text data, and the multimedia data includes audio data and video data;

[0008] Keyword extraction module, used to extract keywords from copywriting data;

[0009] A grouping module is used to group and sort all lesson plans to be evaluated based on the extracted keywords;

[0010] The evaluation module is used to evaluate the teaching plan based on the evaluation model according to the grouping and sorting results, and obtain the teaching plan evaluation analysis results.

[0011] The principles and advantages of this solution are: in actual application, in the actual teaching plan process, the copy data is usually the main teaching material, and the multimedia data is the auxiliary teaching material. Acquiring multimedia data and copy data facilitates comprehensive evaluation of the lesson plan; the keyword extraction module processes the copy data, and extracts keywords that can represent the core content and key information of the lesson plan from the copy, which helps to quickly classify and understand the lesson plan. The grouping module groups and sorts all the lesson plans to be evaluated based on the extracted keywords, making the evaluation process more organized and targeted. The evaluation module outputs the lesson plan evaluation analysis results through the evaluation model according to the results of grouping and sorting, providing a reference for the improvement and optimization of teaching quality. This application can improve the comprehensiveness and efficiency of lesson plan evaluation.

[0012] Preferably, as an improvement, the evaluation model adopts a random forest learning model.

[0013] Technical effect: It can effectively handle complex relationships in large amounts of data, and the results are highly accurate.

[0014] Preferably, as an improvement, the evaluation module further includes:

[0015] The evaluation model creation submodule is used to collect labeled teaching plan samples, extract features from the teaching plan samples and convert them into numerical feature vectors to obtain a data set, and use the data set to train a random forest model to obtain an evaluation model;

[0016] The evaluation result acquisition submodule is used to apply the trained evaluation model to the teaching plan evaluation to obtain the corresponding evaluation results.

[0017] Technical effect: By training the evaluation model, it is easy to improve the accuracy, efficiency and consistency of the evaluation.

[0018] Preferably, as an improvement, the evaluation module further includes a multi-level fusion feature extraction submodule for extracting features during the creation of the evaluation model and the application of the evaluation model; the multi-level fusion feature extraction submodule includes:

[0019] A first feature extraction molecular module, used for performing first feature extraction based on the copywriting data;

[0020] A second feature extraction molecular module, used for performing second feature extraction based on audio data;

[0021] A third feature extraction molecular module, used for performing third feature extraction based on video data;

[0022] The comparative analysis molecule module is used to determine whether the first feature extracted from the copywriting data is missing, and if so, to obtain the missing feature;

[0023] The fusion strategy acquisition molecular module is used to obtain the fusion strategy according to the missing degree of the missing features. If the missing degree exceeds the threshold, the feature fusion extraction is performed through the full fusion molecular module. If the missing degree does not exceed the threshold, the feature fusion extraction is performed through the multi-level point-to-point fusion molecular module.

[0024] The fully fused molecular module is used to convert audio and video information into text information, and integrate the text information with the copy data to obtain the integrated copy data, perform the first feature extraction based on the integrated copy data, and then obtain the missing feature through the comparative analysis molecular module; determine the missing degree of the missing feature, if the missing degree of the missing feature meets the threshold, perform feature fusion through the multi-level point-to-point fusion molecular module; if the missing degree of the missing feature does not meet the threshold, perform feature extraction through the second feature extraction molecular module and the third feature extraction molecular module to obtain the final extracted feature;

[0025] The multi-level point-to-point fusion molecular module is used for multi-level analysis in the order of text, audio, and video. After each analysis, the missing features are analyzed again through the comparative analysis module. If there are no missing features, the analysis ends; if there are missing features, the next level of feature extraction is entered.

[0026] Technical effect: As the difficulty of feature extraction of text, audio, and video gradually increases, the above module can help reduce the difficulty of feature extraction and reduce resource waste.

[0027] Preferably, as an improvement, the keyword extraction module includes:

[0028] The sub-heading extraction sub-module is used to extract keywords from each sub-heading of the copywriting data;

[0029] The text keyword extraction submodule is used to extract keywords from the text of the copy data;

[0030] The keyword screening submodule is used to weight the extracted keywords and screen out valid keywords based on the weighting results.

[0031] Technical effect: The content in the lesson plan is diverse, and the layout is extracted through keywords, which makes it easier to obtain more effective keywords and quickly classify the lesson plans to be evaluated through keywords.

[0032] Preferably, as an improvement, the keyword screening submodule includes:

[0033] The knowledge graph creation molecule module is used to evaluate the objectives based on the teaching plan content and create a knowledge graph;

[0034] The weighting module is used to match keywords with knowledge graph keywords and assign values ​​to keywords based on the degree of matching;

[0035] The molecule screening module is used to screen out effective keywords according to a preset threshold.

[0036] Technical effect: Different subjects and chapters have different evaluation objectives. Therefore, creating a knowledge graph based on the evaluation objectives and then screening keywords can improve the effectiveness of using keywords and lay the foundation for subsequent effective grouping.

[0037] Preferably, as an improvement, the grouping module includes:

[0038] The association analysis submodule is used to analyze the similarity of effective keywords through cluster analysis method;

[0039] The grouping submodule is used to group the lesson plans to be evaluated according to their similarities;

[0040] The sorting submodule is used to sort the grouped lesson plans within and between groups.

[0041] Technical effect: By grouping and sorting, numerous lesson plans can be summarized and compared, making the evaluation more contrasting and the key points more prominent and intuitive.

[0042] Preferably, as an improvement, it also includes an expert verification module for randomly selecting evaluation lesson plans for expert verification, obtaining expert evaluation analysis results, and comparing and analyzing whether the expert evaluation analysis results and the model evaluation analysis results exceed a threshold. When the threshold is exceeded, the evaluation model is optimized.

[0043] Technical effect: It is easy to improve the accuracy of the evaluation model.

[0044] Preferably, as an improvement, it also includes an abnormality assessment module for assessing whether the teaching plan is abnormal based on abnormality assessment parameters, and issuing an early warning when the teaching plan is abnormal.

[0045] Technical effect: By conducting abnormal evaluation of lesson plans, it not only promotes teachers to learn from excellent lesson plans and make professional progress, but also can timely discover education loopholes, encourage teachers to make rectifications, and control the overall education quality.

[0046] It also includes a teaching case evaluation and analysis method, which is applied to a teaching case evaluation and analysis system. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic structural diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following is further described in detail through specific implementation methods:

[0049] The embodiment is basically as shown in the attached Figure 1 As shown:

[0050] A teaching case evaluation and analysis system, comprising:

[0051] The lesson plan information acquisition module is used to obtain the associated data of the lesson plan to be evaluated, and the associated data includes multimedia data and text data; the text data includes document data, ppt data, and classroom test data, and the multimedia data includes video data and audio data of teaching. Acquiring multimedia data and text data facilitates comprehensive evaluation of the lesson plan.

[0052] The keyword extraction module is used to extract keywords from the copy data; in the teaching process, the copy data is the main teaching material, and keywords representing the core content and key information of the teaching plan can be extracted from the copy, so as to facilitate the rapid grouping of the teaching plan by keywords. The keyword extraction module includes a subtitle extraction submodule, a text keyword extraction submodule and a keyword screening submodule.

[0053] The subheading extraction submodule is used to extract keywords from each subheading of the copy data; the keywords in each subheading of the copy data reflect the main structure and theme distribution of the lesson plan. Specifically, the subheadings are extracted through formats such as bold, underline, and specific fonts, and subheading structures (such as "1. Title", "One. Title", "[Title]"), and then the natural language processing technology is used to split the subheadings into multiple words, remove stop words, and leave the keywords. In this embodiment, the natural language processing technology uses a word segmentation algorithm.

[0054] The text keyword extraction submodule is used to extract keywords from the text of the copywriting data; the keywords in the text help to further explore the key information in the lesson plan and assist in information association. Specifically, the TF-IDF (term frequency-inverse document frequency) algorithm is used to screen out stop words and common words, calculate the frequency of each word in the text, and use the words with higher frequency as keywords. Consider the frequency of the words in a single section and the distribution in the entire lesson plan set, taking into account both local importance and global importance, so as to extract more representative keywords.

[0055] The keyword screening submodule is used to weight the extracted keywords and screen out effective keywords based on the weighting results. By weighting and screening the extracted keywords, low-value keywords can be excluded, and a more representative and effective keyword set can be obtained, thereby improving the accuracy and efficiency of subsequent information processing and decision-making; according to different weight settings and screening criteria, the pertinence and applicability of keywords in specific scenarios of teaching plan evaluation can be met. The keyword screening submodule includes a knowledge graph creation molecule module, a weighting molecule module, and a screening molecule module.

[0056] The knowledge graph creates a molecular module, which is used to create a knowledge graph according to the evaluation objectives of the teaching plan content; specifically, the evaluation objectives of the teaching plan are obtained, and information is extracted according to the evaluation objectives. Information extraction includes entity recognition and relationship extraction. Entity recognition is to extract key entities from the teaching plan content, such as subject concepts, characters, events, theorems, etc. For example, in the physics teaching plan, "Newton's Second Law", "mass", "acceleration", etc. are all entities; relationship extraction is to determine the relationship between entities, such as "belong to", "include", "cause and effect", "association" and other relationships. In this embodiment, a graph database knowledge graph model is selected to store and manage knowledge graphs with nodes representing entities and edges representing relationships. The extracted entities and relationships are classified and organized to form a hierarchical structure, and the attributes of various entities and constraints on relationships are defined. When knowledge is extracted from multiple teaching plans and there are different expressions of the same entity, entity alignment is performed; when there is a conflict in knowledge from different sources, it needs to be resolved. In this embodiment, a threshold judgment is also performed on the scale of the knowledge graph. When the scale exceeds the threshold range, that is, the knowledge graph is large in scale and highly complex, the Neo4j graph database is used for storage to efficiently process graph structure data and support complex graph query and analysis operations. When the scale does not exceed the threshold range, a relational database is used for storage, and entities and relationships are stored in different tables respectively.

[0057] The weighted numerator module is used to match keywords with knowledge graph keywords and assign values ​​to keywords based on the degree of matching; specifically, keyword matching is performed through the edit distance algorithm to obtain an initial matching degree. In this embodiment, it is also considered that the knowledge graph contains not only keywords (entities) but also the relationships between entities. Therefore, in addition to comparing the keywords themselves, the structural information of the knowledge graph is also used to correct the matching degree. If two keywords are connected by a shorter path in the knowledge graph and the path relationship is closely related, the initial matching degree is corrected by a preset constant coefficient to improve the matching degree.

[0058] The screening molecule module is used to screen out effective keywords according to preset thresholds, improve the effectiveness of using keywords, and lay the foundation for subsequent effective grouping.

[0059] The grouping module is used to group and sort all the teaching plans to be evaluated based on the extracted keywords, summarize and compare the numerous teaching plans, so that the evaluation is more contrasting and the key points are more prominent and intuitive. The grouping module includes a correlation analysis submodule, a grouping submodule and a sorting submodule.

[0060] The association analysis submodule is used to analyze the similarity of effective keywords through clustering analysis methods; clustering analysis aims to divide the samples in the effective keyword data set into different clusters based on similarity criteria, so that the similarity of samples in the same cluster is high and the similarity of samples between different clusters is low. The word vector model is used to convert keywords into numerical vectors, and the keywords are mapped to the feature space so that semantically similar keywords are close in the vector space. The vectorized keywords are input into the clustering algorithm for clustering operations, completing the cluster allocation of keywords and realizing the aggregation of similar keywords.

[0061] The grouping submodule is used to group the lesson plans to be evaluated according to their similarities. Based on the keyword clustering results obtained by the association analysis submodule, the lesson plans to be evaluated are grouped according to the clusters to which the keywords they contain belong. If most of the keywords contained in two lesson plans belong to the same cluster, it indicates that they have high similarity in content and are classified into the same group.

[0062] The sorting submodule is used to sort the grouped lesson plans within the group and between groups. The intra-group sorting is intended to sort the lesson plans within the same group in terms of importance based on the degree of fit between the lesson plans and the keyword features of the group. The inter-group sorting is based on the overall features of the lesson plans within the group. In this embodiment, the overall features are keyword coverage, and different groups are sorted in terms of importance. In this way, the relative positions and importance of different lesson plans under the similarity dimension are clearly displayed.

[0063] The evaluation module is used to evaluate the teaching plan based on the evaluation model according to the grouping and sorting results, and obtain the teaching plan evaluation analysis results. The evaluation model adopts the random forest learning model. The evaluation module includes an evaluation model creation submodule and an evaluation result acquisition submodule. The evaluation model creation submodule is used to collect labeled teaching plan samples, extract features from the teaching plan samples and convert the features into numerical feature vectors to obtain a data set, use the data set to train the random forest model, and obtain an evaluation model; the evaluation result acquisition submodule is used to apply the trained evaluation model to the teaching plan evaluation to obtain the corresponding evaluation results.

[0064] The evaluation module also includes a multi-level fusion feature extraction submodule, which is used for feature extraction in the evaluation model creation and evaluation model application process; the features are determined according to the teaching plan evaluation requirements, including multi-dimensional features such as teaching plan content and teaching plan integrity. The multi-level fusion feature extraction submodule includes: a first feature extraction molecular module, a second feature extraction molecular module, a third feature extraction molecular module, a comparative analysis molecular module, a fusion strategy acquisition molecular module, a full fusion molecular module, and a multi-level point-to-point fusion molecular module.

[0065] The first feature extraction molecular module is used to extract the first feature based on the text data; the second feature extraction molecular module is used to extract the second feature based on the audio data; the third feature extraction molecular module is used to extract the third feature based on the video data; the comparative analysis molecular module is used to determine whether the first feature extracted from the text data is missing. If it is missing, the missing feature is obtained. If it is not missing, the first feature is directly used as the final extracted feature; the fusion strategy acquisition molecular module is used to obtain the fusion strategy according to the missing degree of the missing feature. If the missing degree exceeds the threshold, the feature fusion extraction is performed through the full fusion molecular module. If the missing degree does not exceed the threshold, the feature fusion extraction is performed through the multi-level point-to-point fusion molecular module. The fully fused molecular module is used to convert audio and video information into text information, and integrate the text information with the copy data to obtain the integrated copy data, perform the first feature extraction based on the integrated copy data, and then obtain the missing features through the comparative analysis molecular module; determine the missing degree of the missing features, if the missing degree of the missing features meets the threshold, perform feature fusion through the multi-level point-to-point fusion molecular module; if the missing degree of the missing features does not meet the threshold, perform feature extraction through the second feature extraction molecular module and the third feature extraction molecular module to obtain the final extracted features. The multi-level point-to-point fusion molecular module is used to perform multi-level analysis in the order of copy, audio, and video, and analyze the missing features again through the comparative analysis module after each analysis. If there are no missing features, the analysis ends; if there are missing features, the next level of feature extraction is entered. This embodiment also includes a time inference submodule, which is used to predict the time interval of the audio media corresponding to the missing feature part based on the text structure and recorded content and based on the trained prediction model, so as to facilitate the capture of audio for feature extraction; if the current analysis level is audio, the capture interval of the video is determined based on the text and audio, and the feature of the captured part is extracted, and then added to the extracted feature to form the final feature.

[0066] It also includes an expert verification module, which is used to randomly select evaluation lesson plans for expert verification, obtain expert evaluation analysis results, and compare and analyze whether the expert evaluation analysis results and the model evaluation analysis results exceed the threshold. If the threshold is exceeded, the evaluation model is optimized. Specifically,

[0067] Cases to be evaluated were randomly selected, and authoritative experts were invited according to the number of cases. Each group of 5 experts selected the same cases. According to the expert grouping and the number of interval cases, experts were organized to conduct reviews, and the expert scoring data was obtained to obtain the original average score of each case. The expert scoring model was obtained based on the regression analysis of the original average score of each case and the expert scoring. The expert scoring model was used to eliminate the influence of expert differences on the average score, and the case score after removing the expert influence was obtained. The relative results were weighted summed to obtain the average score of each case. The expert scoring correction model was obtained based on the average score after the first adjustment and the regression analysis of the expert scoring. The expert scoring model was used to eliminate the influence of expert differences on the average score again, and the final average score after the second adjustment was obtained. The relationship between the rank increment and the basic rank was determined by the regression analysis method, and then the influence of differences between different experts on the ranking was eliminated, achieving a more accurate and comprehensive case evaluation. Among them, the method for obtaining the rank increment after removing the basic influence is as follows: suppose there are m cases, n experts, i is the case or experience number, j is the expert number, X is the average score after the first adjustment, and x is the average score after the second adjustment. ij is the original score of expert j on case i or experience i = 1, 2, ..., m; j = 1, 2, ..., n; remove the highest score and the lowest score of case i to get the weighted original average score Y i =(∑X ij -X imax -X imin ) / (n-2); use X ij The original score of expert j on case i or experience i is the independent variable, and the original mean score of case i is Y i As the dependent variable, the expert scoring model was obtained by regression analysis: K ij =A j X ij +B j ; Use the expert scoring model to get the fitting score K for each case ij , remove the highest and lowest scores of the same case and calculate the weighted average score: M i =(∑K ij -K imax -K imin ) / (n-2); with X ij is the independent variable, M i As the dependent variable, regression analysis was used to fit the expert score correction model: H ij =a j X ij +b j ; Use expert scoring to correct the model and get the fitting score H for each case ij , remove the highest and lowest scores of the same case and calculate the final weighted average score: Q i =(∑H ij -H imax -H imin) / (n-2); get the final weighted average score Q i Optimizing the evaluation model based on the expert verification results facilitates improving the accuracy of the evaluation model.

[0068] It also includes an abnormal evaluation module, which is used to evaluate whether the teaching plan is abnormal based on abnormal evaluation parameters, and issue an early warning when the teaching plan is abnormal. The abnormal evaluation parameters include but are not limited to target achievement deviation, content obsolescence, lack of method diversity, and content accuracy errors. The difference rate between the expected teaching objectives of the teaching plan and the actual student learning outcomes is the target achievement deviation. For example, the teaching plan sets that students should master 80% of specific knowledge points after the course, but the actual test finds that only 40% have been mastered, then an abnormal warning is issued; content accuracy errors such as in the history teaching plan, if the time, characters, causal relationships, etc. of historical events are incorrectly stated, it is a content abnormality; content obsolescence such as in the computer science teaching plan, if the focus is still on teaching obsolete programming languages ​​or technologies, it is judged that the content is obsolete, which is an abnormal situation. Lack of method diversity is such as the types of teaching methods used in the statistical teaching plan. If the entire teaching plan relies only on a single lecture method and lacks other methods such as group discussion and practical operation, the lower limit of the types of teaching methods is set. If it is lower than this limit, it is judged as abnormal and needs to be warned. By conducting abnormal evaluation of lesson plans, it is not only possible to promote teachers to learn from excellent lesson plans and make professional progress, but also to discover educational loopholes in a timely manner, encourage teachers to make corrections, and control the overall quality of education.

[0069] It also includes a teaching case evaluation and analysis method, which is applied to a teaching case evaluation and analysis system.

[0070] The above is only an embodiment of the present invention, and the common knowledge such as the known specific technical solutions and / or characteristics in the solution is not described in detail here. It should be pointed out that for those skilled in the art, without departing from the technical solution of the present invention, several modifications and improvements can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A teaching case evaluation and analysis system, characterized in that: include: The lesson plan information acquisition module is used to obtain the associated data of the lesson plan to be evaluated, the associated data includes multimedia data and text data, and the multimedia data includes audio data and video data; Keyword extraction module, used to extract keywords from copywriting data; A grouping module is used to group and sort all lesson plans to be evaluated based on the extracted keywords; The evaluation module is used to evaluate the teaching plan based on the evaluation model according to the grouping and sorting results, and obtain the teaching plan evaluation analysis results.

2. A teaching case evaluation and analysis system according to claim 1, characterized in that: The evaluation model adopts a random forest learning model.

3. A teaching case evaluation and analysis system according to claim 2, characterized in that ,, the evaluation module also includes: The evaluation model creation submodule is used to collect labeled teaching plan samples, extract features from the teaching plan samples and convert them into numerical feature vectors to obtain a data set, and use the data set to train a random forest model to obtain an evaluation model; The evaluation result acquisition submodule is used to apply the trained evaluation model to the teaching plan evaluation to obtain the corresponding evaluation results.

4. A teaching case evaluation and analysis system according to claim 3, characterized in that: The evaluation module also includes a multi-level fusion feature extraction submodule, which is used for feature extraction during the evaluation model creation and evaluation model application process; the multi-level fusion feature extraction submodule includes: A first feature extraction molecular module, used for performing first feature extraction based on the copywriting data; A second feature extraction molecular module, used for performing second feature extraction based on audio data; A third feature extraction molecular module, used for performing third feature extraction based on video data; The comparative analysis molecule module is used to determine whether the first feature extracted from the copywriting data is missing, and if so, to obtain the missing feature; The fusion strategy acquisition molecular module is used to obtain the fusion strategy according to the missing degree of the missing features. If the missing degree exceeds the threshold, the feature fusion extraction is performed through the full fusion molecular module. If the missing degree does not exceed the threshold, the feature fusion extraction is performed through the multi-level point-to-point fusion molecular module. The fully fused molecular module is used to convert audio and video information into text information, and integrate the text information with the copy data to obtain the integrated copy data, perform the first feature extraction based on the integrated copy data, and then obtain the missing feature through the comparative analysis molecular module; determine the missing degree of the missing feature, if the missing degree of the missing feature meets the threshold, perform feature fusion through the multi-level point-to-point fusion molecular module; if the missing degree of the missing feature does not meet the threshold, perform feature extraction through the second feature extraction molecular module and the third feature extraction molecular module to obtain the final extracted feature; The multi-level point-to-point fusion molecular module is used for multi-level analysis in the order of text, audio, and video. After each analysis, the missing features are analyzed again through the comparative analysis module. If there are no missing features, the analysis ends; if there are missing features, the next level of feature extraction is entered.

5. A teaching case evaluation and analysis system according to claim 1, characterized in that: The keyword extraction module comprises: The sub-heading extraction sub-module is used to extract keywords from each sub-heading of the copywriting data; The text keyword extraction submodule is used to extract keywords from the text of the copy data; The keyword screening submodule is used to weight the extracted keywords and screen out valid keywords based on the weighting results.

6. A teaching case evaluation and analysis system according to claim 5, characterized in that: The keyword screening submodule includes: The knowledge graph creation molecule module is used to evaluate the objectives based on the teaching plan content and create a knowledge graph; The weighting module is used to match keywords with knowledge graph keywords and assign values ​​to keywords based on the degree of matching; The molecule screening module is used to screen out effective keywords according to a preset threshold.

7. A teaching case evaluation and analysis system according to claim 1, characterized in that: The grouping module comprises: The association analysis submodule is used to analyze the similarity of effective keywords through cluster analysis method; The grouping submodule is used to group the lesson plans to be evaluated according to their similarities; The sorting submodule is used to sort the grouped lesson plans within and between groups.

8. The teaching case evaluation and analysis system according to claim 1, characterized in that: It also includes an expert verification module, which is used to randomly select evaluation lesson plans for expert verification, obtain expert evaluation analysis results, and compare and analyze whether the expert evaluation analysis results and the model evaluation analysis results exceed the threshold. When the threshold is exceeded, the evaluation model is optimized.

9. The teaching case evaluation and analysis system according to claim 1, characterized in that: It also includes an abnormality assessment module, which is used to assess whether the teaching plan is abnormal based on abnormality assessment parameters, and issue an early warning when the teaching plan is abnormal.

10. A teaching case evaluation and analysis method, characterized in that: Applicable to a teaching case evaluation and analysis system as described in any one of claims 1-9.

Images