Teaching quality comprehensive evaluation system and method based on data analysis
By introducing social network comments and deep learning models into teaching quality evaluation, the problem of bias in student satisfaction survey results has been solved, achieving a more accurate and fair evaluation of teaching quality and promoting teaching improvement and management decisions.
Patent Information
- Application Number
- CN202510929090.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-07
- Publication Date
- 2025-12-05
AI Technical Summary
In the current teaching quality evaluation, the results of student satisfaction surveys are biased due to pressure from teachers or concerns from students, resulting in low reliability and validity of the evaluation results, which affects the pertinence and effectiveness of teaching improvement.
By acquiring students' comments on social media platforms, target comments are extracted using named entity recognition and deep learning models. Combined with teaching evaluation questionnaire scores, a multi-source information correction method is adopted to evaluate teaching quality, including quantifying teacher-student relationships and emotional tendencies, and dynamically correcting questionnaire scores.
It improves the fairness, transparency, and credibility of teaching quality evaluation, maintains stability under high-risk conditions, provides precise guidance for teaching improvement and a basis for management decisions, reduces evaluation costs, and protects students' freedom of anonymous expression.
Smart Images
Figure CN121073264A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of teaching quality evaluation, and particularly relates to a teaching quality comprehensive evaluation system and method based on data analysis. BACKGROUND
[0002] In the teaching quality evaluation system, student satisfaction as a core indicator directly reflects the degree of fit between teaching effectiveness and student needs. By conducting regular satisfaction surveys, understanding students' real feelings about course content, teaching methods, teacher-student interaction and learning resources, etc., important feedback can be provided to managers and teachers.
[0003] Currently, student questionnaire surveys are often used as the main means to measure satisfaction in teaching quality evaluation. However, due to various objective and subjective factors, such as teachers requiring students to give high scores, students being reluctant to answer truthfully due to concerns, etc., the survey results often deviate from the students' real feelings. Such bias not only weakens the reliability and validity of the satisfaction evaluation, but also affects the targeting and implementation effect of subsequent teaching improvement initiatives. SUMMARY
[0004] The purpose of the present application is to provide a teaching quality comprehensive evaluation system and method based on data analysis, which solves the above technical problems.
[0005] The purpose of the present application can be achieved by the following technical solutions: A teaching quality comprehensive evaluation method based on data analysis, comprising the following steps: Obtain comments posted by students on a social network platform, extract target comments therefrom, the target comments being comments corresponding to the same teacher that meet predetermined requirements, determine a first proportion, the first proportion being used to quantify the relationship between the overall students and the teacher; Obtain the labels of the target comments based on a pre-constructed label output model, the labels including positive and negative, and calculate the proportion of the number of target comments with positive labels to the total number of target comments, denoted as a second proportion; Obtain the answer results of the evaluation questionnaire, obtain the questionnaire scores based on the answer results, and obtain the median B of the questionnaire scores; Correct the median B based on the first proportion and the second proportion to obtain the standard score F of the teacher, and evaluate the teaching quality of the teacher according to the standard score.
[0006] As a further scheme of the present application, obtaining the target comments comprises: Obtain the content on the social network platform corresponding to the comments, extract the teacher's name in the content, extract the characters in the teacher's name, combine the characters to obtain a plurality of combined words, and take the single character and the teacher's name as combined words; Adding a preset supplementary word before and / or after the first character and / or the last character of the single combined word to obtain a plurality of target words; Comments corresponding to the same teacher and existing the target word and / or the teacher name are taken as target comments.
[0007] As a further scheme of the present application, the target proportion comprises: A number A1 of target comments existing the target word is obtained, and a number A2 of target comments is obtained; A target proportion A=A1 / A2 is calculated.
[0008] As a further scheme of the present application, the label of the target comment comprises: A database is established, and the database stores comments based on artificial labeling labels; A label output model is established based on deep learning, the label output model is trained and verified based on the database, the target comment is input into the trained and verified label output model, and a label of the target comment is output.
[0009] As a further scheme of the present application, the standard score comprises: The options are valued; The values of the options in the answer results are accumulated to obtain a questionnaire score; A standard score is calculated k1 and k2 represent the first coefficient and the second coefficient respectively.
[0010] As a further scheme of the present application, before the standard score is obtained, the method further comprises: If the average value of the questionnaire score is less than a preset average value threshold and / or the variance is greater than a preset variance threshold, the correction is not performed, and the median of the questionnaire score is taken as the standard score.
[0011] As a further scheme of the present application, the teaching quality of the teacher is evaluated according to the standard score, which comprises: Score thresholds F1 and F2 are set, and F1<F2; When F<F1, the teaching quality is determined to be poor; When F1≤F<F2, the teaching quality is determined to be general; When F2≤F, the teaching quality is determined to be good.
[0012] A teaching quality comprehensive evaluation system based on data analysis comprises: A first correction module: obtaining comments published by students on a social network platform, extracting target comments therefrom, the target comments being comments corresponding to a same teacher and satisfying a preset requirement, determining a first proportion, the first proportion being used to quantify the relationship between the overall students and the teacher; The second correction module: based on the pre-constructed label output model, the label of the target comment is obtained, the label includes positive and negative, and the proportion of the number of target comments with positive label to the total number of target comments is calculated as a second proportion; The evaluation module: obtaining the answer result of the teaching evaluation questionnaire, obtaining the questionnaire score based on the answer result, and obtaining the median B of the questionnaire score; The standard score F of the teacher is obtained by correcting the median B based on the first proportion and the second proportion, and the teaching quality of the teacher is evaluated according to the standard score.
[0013] The beneficial effects of the present application are as follows: The present application discards the single evaluation idea of relying only on questionnaire scores, introduces student comments in the real context of social networks, quantifies the intimacy of teacher-student interaction with target word proportion, and automatically identifies comment tendency through deep learning model, dynamically corrects with questionnaire median, and fully offsets the scoring distortion caused by teacher pressure or student concerns. The complementarity of multi-source information makes the evaluation result closer to the real experience of students, can maintain stability in high-risk situations, and takes into account subjective satisfaction and objective feedback when evaluating good and bad grades, which not only provides accurate guidance for teaching improvement, but also provides reliable basis for management decision-making, significantly improves the fairness, transparency and credibility of teaching quality evaluation. In addition, the present method can track the performance of teachers for a long time, update the teaching improvement effect in time through continuous data, and avoid the contingency caused by one-time evaluation; the automatic processing process reduces manual intervention, reduces evaluation cost, and at the same time protects the freedom of students' anonymous expression, further consolidates the objectivity and fairness of the evaluation system. BRIEF DESCRIPTION OF DRAWINGS
[0014] The present application will be further described below in conjunction with the accompanying drawings.
[0015] Figure 1 is a flowchart of a teaching quality comprehensive evaluation method based on data analysis. DETAILED DESCRIPTION
[0016] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.
[0017] Please refer to Figure 1 The present application is a teaching quality comprehensive evaluation system and method based on data analysis, which comprises the following steps: Obtaining the comments published by students on the social network platform, extracting the target comments therefrom, and calculating the target proportion; In a preferred embodiment, obtaining target comments and target proportion comprises: Calling the social network open interface to capture the post content and comments associated with the course keywords within a time window, using a named entity recognition model to locate the teacher's name in the text, for example, identifying the name "XYZ", then splitting the name into single characters "X", "Y", "Z", and generating "XY", "YZ", "XZ" and other connected combination words according to the adjacent splicing rule, and at the same time, the complete name "XYZ" and each single character are also included in the combination word set; For each single character combination word in the set, introduce common affectionate or respectful words, the rule is to add a prefix before the first character or a suffix after the last character, such as adding "A" "old" before the single character "X" to get "AX" "old X", or adding "brother" "sister" "teacher" after the character to get "X brother" "X sister" "X teacher", and the same processing is also done for double character combination "XY" words, forming a target word list covering common nicknames, honorifics and full names; Then use Boolean matching to search the comment content, mark the comments that contain target words or complete names and have a unique correspondence with the teacher entity as target comments, record the comments matched to target words as A1, and all target comments as A2, and finally output the target proportion A by dividing A1 by A2; It is worth noting that by generating diversified target words and matching in comments, the students' colloquial address and nicknames for teachers can be captured, which often appear when students feel close or appreciate, so the target proportion can indirectly reflect the teacher-student relationship. Compared with the practice of only retrieving formal names, this strategy significantly reduces the omission rate and has stronger perception ability for informal expressions; In the subsequent evaluation process, the first proportion serves as a correction factor for the questionnaire score, helping the comprehensive model balance the students' true attitude in the anonymous social context and the explicit feedback in the questionnaire, thereby improving the credibility and sensitivity of the final teaching quality score and making the evaluation results more in line with the students' true feelings.
[0018] Based on the pre-constructed label output model, obtain the label of the target comment, and calculate the second proportion; In a specific embodiment, based on the historical course comment, the original text is collected, and sentences such as "The teacher's explanation is very clear" and "The homework is too much" are used as basic samples. Artificially label each sentence as positive or negative according to teaching satisfaction, and construct a corpus with sentiment labels. Then, the corpus is processed by sentence segmentation, word segmentation and denoising, and a pre-trained language model such as Chinese BERT is used for vectorization. The samples are randomly divided into training set, validation set and test set. Through fine-tuning, a label output model that can distinguish positive and negative emotions is obtained, and the hyperparameters and loss function are iteratively adjusted on the validation set until convergence. After the model training is completed, input the target comment such as "X teacher's class is humorous" or "X teacher's answering is delayed" into the model, the model outputs the corresponding positive or negative label, and then the ratio of the number of positive labels to the total number of target comments is obtained as the second ratio as the input of subsequent evaluation; It can be understood that the system automatically captures the subtle emotional tendency in the language of students, and uses a deep learning model to replace manual judgment piece by piece, which not only saves manpower but also maintains consistency. At the same time, through the supervised learning of large-scale labeled data, the model maintains high recognition ability for colloquial and slang expressions, avoiding simple emotional dictionary missed detection and misjudgment. After the introduction of the second ratio, an independent and real-time student emotion perspective can be provided when the questionnaire score is disturbed externally, and the first ratio is used to dynamically correct the median of the questionnaire, so that the comprehensive evaluation is more comprehensive and closer to the real experience of students, thereby providing a more reliable basis for teaching improvement and decision-making. According to the evaluation questionnaire answered by the student, the questionnaire score is obtained, the median of the questionnaire score is obtained to reflect the majority of the situation, and the median B is corrected according to the first ratio and the second ratio; In another preferred embodiment, through the interface of the teaching affairs platform, all student evaluation questionnaires of the teacher in the current term are called, the answers of each question of each questionnaire are mapped to a preset score table, for example, "very agree" in a five-level scale is assigned the highest score, and "very disagree" is assigned the lowest score, and the scores of all questions in the same questionnaire are summed to obtain the total score of the questionnaire; When a plurality of questionnaires of the same teacher are aggregated, the total scores are sorted from small to large, and the value located in the center of the sequence is taken as the median B of the questionnaire score to reflect the majority of the student's answered questionnaire, and the first ratio (reflecting the frequency of nickname or endearment) and the second ratio (reflecting the proportion of positive emotional labels) obtained in the previous stage are called, and the standard score F is calculated; It should be noted that the design principle of the formula is to take the robust representative value B of the questionnaire score as the evaluation benchmark, and integrate the two independent indicators of the social network into a single correction coefficient through multiplication. Using Instead of linear weighting, it can ensure that when k1 or k2 is zero, it directly degenerates to pure questionnaire score, avoiding excessive fluctuations caused by invalid or extreme network signals; by choosing the product instead of the sum, the two social indicators will only significantly increase the score when they are both positive, and if they are opposite, the correction range will be suppressed, thereby achieving the "common approval only for bonus" effect of logical harmony; The questionnaire result is fused with two social network indexes, complementary information sources are introduced when the questionnaire is affected by favor or resistance, the extreme value interference is reduced by correcting the median instead of the average, the standard score retains the subjective overall judgment of the questionnaire on the teaching satisfaction, and the real attitude signal of the student in the natural context is absorbed, and finally the comprehensive conclusion of the evaluation system is more stable, more reliable, and closer to the actual classroom, and more objective and fair decision reference is provided for teaching improvement, resource allocation and teacher incentive. It is worth noting that if the average of the questionnaire score is less than the preset average threshold and / or the variance is greater than the preset variance threshold, no correction is performed, and the median of the questionnaire score is used as the standard score. When the overall average score is obviously low or the distribution variance is too large, it usually means that the students are generally dissatisfied or the opinions are extremely polarized, and at this time, if the correction term generated by the social network is still used to adjust the median, the serious problems revealed by the questionnaire may be covered up, or the influence of a small amount of noise information on the result may be amplified; therefore, directly using the median as the standard score can make the evaluation sensitive to abnormal situations, and ensure that the score can truly reflect the outstanding contradictions existing in the classroom. The teaching quality of the teacher is evaluated according to the standard score. In another preferred embodiment of the present application, the evaluation of the teaching quality of the teacher according to the standard score comprises: The score thresholds F1 and F2 are set, and F1<F2. When F<F1, the teaching quality is determined to be poor. When F1≤F<F2, the teaching quality is determined to be general. When F2≤F, the teaching quality is determined to be good.
[0019] A teaching quality comprehensive evaluation system based on data analysis, characterized by comprising: The first correction module: obtaining the comments published by the students on the social network platform, extracting the target comments, the target comments are comments corresponding to the same teacher and meeting the preset requirements, determining a first proportion, the first proportion is used to quantify the relationship between the overall students and the teacher; The second correction module: obtaining the label of the target comment based on the pre-constructed label output model, the label includes positive and negative, and the proportion of the number of target comments with positive labels to the total number of target comments is calculated as a second proportion; The evaluation module: obtaining the answer result of the evaluation questionnaire, obtaining the questionnaire score based on the answer result, and obtaining the median B of the questionnaire score; The median B is corrected based on the first proportion and the second proportion, to obtain the standard score F of the teacher, and the teaching quality of the teacher is evaluated according to the standard score.
[0020] The above has been described in detail one embodiment of the present application, but the content is only the preferred embodiment of the present application, cannot be considered for limiting the scope of the present application. Any equivalent changes and improvements made in the scope of the present application, should still belong to the scope of the present application.
Claims
1. A data analysis-based comprehensive evaluation method for teaching quality, characterized in that, The method comprises the following steps: obtaining comments published by students on a social network platform, extracting target comments meeting preset requirements corresponding to the same teacher from the comments, determining a first proportion for quantifying the relationship between the overall students and the teacher; obtaining labels of the target comments based on a pre-constructed label output model, the labels including positive and negative, and calculating a second proportion of the number of target comments with positive labels to the total number of target comments; obtaining the answer results of the evaluation questionnaire, obtaining a questionnaire score based on the answer results, and obtaining a median B of the questionnaire score; correcting the median B based on the first proportion and the second proportion to obtain a standard score F of the teacher, and evaluating the teaching quality of the teacher according to the standard score. 2.The method of claim 1, wherein, The target comments are obtained by: obtaining content on the social network platform corresponding to the comments, extracting the teacher's name from the content, extracting characters in the teacher's name, combining the characters to obtain a plurality of combined words, and taking the single character and the teacher's name as combined words; adding a preset supplementary word before the first character and / or after the last character of the single combined word to obtain a plurality of target words; taking the comments corresponding to the same teacher and containing the target words and / or the teacher's name as the target comments. 3.The method of claim 2, wherein, The target proportion is obtained by: obtaining the number A1 of target comments containing the target words, and obtaining the number A2 of target comments; calculating the target proportion A = A1 / A2.
4. The teaching quality comprehensive evaluation method based on data analysis according to claim 1, characterized in that, The labels of the target comments are obtained by: establishing a database, and storing comments with labels annotated by artificial labeling in the database; establishing a label output model based on deep learning, training and verifying the label output model based on the database, inputting the target comments into the trained and verified label output model, and outputting the labels of the target comments.
5. The method of claim 1, wherein the method further comprises: The standard score is obtained by: assigning values to the options; adding up the values of the options in the answer results to obtain the questionnaire score; Computing the standard score k1, k2 represent a first coefficient and a second coefficient, respectively.
6. The method of claim 1, wherein the method further comprises: Before obtaining the standard score, the method further comprises: if the average value of the questionnaire score is less than a preset average threshold value and / or the variance is greater than a preset variance threshold value, the correction is not performed, and the median of the questionnaire score is taken as the standard score.
7. The method according to claim 1, wherein, The evaluation of the teaching quality of the teacher according to the standard score comprises: setting score thresholds F1 and F2, and F1 < F2; when F < F1, the teaching quality is determined to be poor; when F1 ≤ F < F2, the teaching quality is determined to be general; when F2 ≤ F, the teaching quality is determined to be good.
8. A data analysis-based comprehensive evaluation system for teaching quality, characterized by, The method comprises: a first correction module: obtaining comments published by students on a social network platform, extracting target comments meeting preset requirements corresponding to the same teacher from the comments, determining a first proportion for quantifying the relationship between the overall students and the teacher; a second correction module: obtaining labels of the target comments based on a pre-constructed label output model, the labels including positive and negative, and calculating a second proportion of the number of target comments with positive labels to the total number of target comments; an evaluation module: obtaining the answer results of the evaluation questionnaire, obtaining a questionnaire score based on the answer results, and obtaining a median B of the questionnaire score; The standard score F of the teacher is obtained based on the first proportion and the second proportion correction of the median B, and the teaching quality of the teacher is evaluated according to the standard score.