AI-based teaching quality evaluation method

Abstract

The invention discloses an AI-based teaching quality evaluation method, which relates to the technical field of teaching quality management, and comprises the steps of historical data collection and preprocessing, sound model establishment, content model establishment, knowledge point distribution model establishment and speech recognition and teaching quality evaluation. According to the method, a sound model, a content model and a knowledge point distribution model are respectively established, the similarity between two factors of sound and teaching content of teacher teaching and excellent teachers is analyzed through an AI algorithm, comprehensive evaluation of teacher teaching is finally obtained, and through comprehensive analysis of the models, the teaching quality of teachers can be quantitatively evaluated. Therefore, a basis for improving the teaching method is provided for teachers. In addition, the device can give specific improvement suggestions according to evaluation results, and helps teachers to obtain better effects in future teaching activities.

Description

Technical Field

[0001] The present invention relates to the technical field of teaching quality management, and in particular to a teaching quality evaluation method based on AI. Background Art

[0003] In the process of teachers' teaching activities, due to the significant differences in the backgrounds of students and the diversity and complexity of teaching content, it is often difficult to accurately evaluate and judge the teaching quality of teachers. In addition, there is no intuitive improvement guidance after the evaluation of teachers' teaching quality, which makes it difficult to improve teachers' teaching quality. For this reason, we propose an AI-based teaching quality evaluation method. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that in the process of teachers' teaching activities, due to the significant differences in the backgrounds of student groups and the diversity and complexity of teaching content, it is often difficult to accurately evaluate and judge the teaching quality of teachers, which leads to the easy burial of talents and affects the fairness of education. In addition, there is no intuitive improvement guidance after the evaluation of teachers' teaching quality, which makes it difficult to improve teachers' teaching quality. The present invention provides an AI-based teaching quality evaluation method.

[0005] In order to achieve the above-mentioned purpose, the present invention specifically adopts the following technical solutions:

[0006] An AI-based teaching quality evaluation method includes the following steps:

[0007] S1. Historical data collection and preprocessing: collect excellent course videos to ensure high video quality without obvious noise interference, use video processing software to extract audio tracks and convert them into standard formats;

[0008] S2. Establish a sound model, apply audio processing technology, extract the spectral characteristics of the teacher's lecture sound, use AI algorithm to establish an excellent teacher's sound model, and identify and simulate the pitch curve of excellent teachers;

[0009] S3. Establish a content model, use natural language processing technology to extract key knowledge points in teaching content, use AI algorithms to analyze the relevance of keywords and teaching textbook content, and establish an excellent teacher content model;

[0010] S4. Establish a knowledge point distribution model, use natural language processing technology to extract key knowledge points in the teaching content, and obtain the time points where the key knowledge points are located throughout the course, use the AI ​​algorithm model to analyze the time distribution of key knowledge points, and establish a knowledge point distribution model for excellent teachers;

[0011] S5. Speech recognition and teaching quality assessment: Use speech recognition technology to transcribe the teacher's lecture content into audio, use audio processing technology and natural language processing technology to extract the teacher's lecture sound, key knowledge points and key knowledge points time distribution, use AI algorithm to analyze its similarity with the sound model, content model and knowledge point distribution model of excellent teachers, and take the average value as the evaluation indicator of teaching quality.

[0012] Furthermore, in S2, Fourier transform is used to extract spectrum features. First, the audio signal is collected to obtain the signal representation f(t) in the time domain. Then, the continuous Fourier transform formula is applied to calculate the spectrum representation F(ω) of the signal. The formula is as follows:

[0013] F(ω)=∫_{-∞}^{∞}f(t)e^{-iωt}dt

[0014] Among them, F(ω) represents the spectrum of the signal, ω is the angular frequency, i is the imaginary unit, and e is the base of the natural logarithm. By calculating this integral, we can obtain the amplitude and phase information of the signal at different frequencies, that is, the spectrum characteristics.

[0015] Furthermore, the establishment of the excellent teacher voice model uses the Hidden Markov Model (HMM) algorithm for modeling and analysis. In the process of establishing the voice model, the spectrum feature is used as input data, and the voice curve formula is as follows:

[0016] P(O|λ)=∑_{Q}P(O|Q,λ)P(Q|λ)

[0017] Where P(O|Q,λ) is the probability of observing the sequence O given the state sequence Q and the model parameter λ.

[0018] Furthermore, when the excellent teacher content model is established, the TF-IDF (Term Frequency-Inverse Document Frequency) algorithm is used to determine the importance of key knowledge points in the text. The higher the TF-IDF value, the greater the importance of the word in the document. The calculation formula is:

[0019] \text{TF-IDF}(t,d)=\text{TF}(t,d)\times\text{IDF}(t)

[0020] \text{TF}(t,d): the term frequency of word (t) in document (d), that is, the number of times word (t) appears in document (d) divided by the total number of all words in document (d), \text{IDF}(t): the inverse document frequency of word (t), calculated as (\log\left(\frac{N}{\text{df}(t)}\right)), where (N) is the number of all documents, and (\text{df}(t)) is the number of documents containing word (t).

[0021] Furthermore, the supervisor of the knowledge point distribution model uses regression analysis to build a model, and based on the results of the regression analysis, explains the specific impact of key knowledge points and the time period of class on the teaching effect, as well as the interaction between them. The calculation formula is:

[0022] y=\beta_0+\beta_1x_1+\beta_2x_2+\epsilon

[0023] Among them, (y) is the teaching effect (dependent variable), (x_1): TF-IDF value of key knowledge points (independent variable), (x_2): class time (independent variable), (\beta_0): intercept term, (\beta_1,\beta_2): regression coefficient, indicating the influence of independent variable on dependent variable, (\epsilon): error term.

[0024] Furthermore, the similarity between the teacher's lecture content and the excellent teacher model is calculated using cosine similarity, and the calculation formula is:

[0025] \text{cosine_similarity}(A,B)=\frac{A\cdot B}{|A||B|}

[0026] Among them, A and B represent two vectors respectively, A·B represents the dot product of vectors A and B, that is, the sum of the products of corresponding elements of the two vectors, ||A|| and ||B|| represent the Euclidean norm (length) of vectors A and B respectively, that is, the square root of the sum of the squares of each element of the vector.

[0027] Furthermore, the teaching quality evaluation standard is that an average similarity of 90% or more is excellent, an average similarity of 80%-89% is good, an average similarity of 70%-79% is fair, and an average similarity of less than 70% is poor.

[0028] The beneficial effects of the present invention are as follows:

[0029] 1. The present invention extracts the teaching course characteristics of excellent teachers through AI algorithms, establishes sound models, content models and knowledge point distribution models respectively, and then uses AI algorithms to analyze the similarities between the two major factors of teachers' teaching, namely, sound and teaching content, and excellent teachers, and finally obtains a comprehensive evaluation of teachers' teaching. Specifically, the sound model focuses on the voice characteristics of teachers, such as speaking speed, pitch, volume, etc., while the content model focuses on the depth and breadth of teaching content, and the knowledge point distribution model analyzes the coverage and distribution patterns of knowledge points in the teaching content. Through the comprehensive analysis of these models, the teaching quality of teachers can be quantitatively evaluated, thereby providing teachers with a basis for improving teaching methods. In addition, the device can also give specific improvement suggestions based on the evaluation results to help teachers achieve better results in future teaching activities.

[0030] 2. The present invention can obtain a value between -1 and 1 by calculating the cosine similarity between the teacher's lecture content and the excellent teacher model. The closer the value is to 1, the higher the similarity is. The similarity value reflects the closeness between the teacher's lecture content and the excellent teacher model in terms of sound characteristics, content structure and knowledge point distribution, so it can be used as an evaluation indicator of teaching quality. By analyzing the distribution of these similarity values, the strengths and weaknesses in teaching can be identified. For example, if the similarity between a teacher's lecture content on a specific knowledge point and the excellent teacher model is generally low, then this may indicate that the teacher needs further training and improvement in the teaching of these knowledge points. On the contrary, if the similarity is generally high, it means that the teacher has done well in these aspects and can serve as a role model for other teachers to learn. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a specific operation flow chart of the teaching quality evaluation method of the present invention. DETAILED DESCRIPTION

[0032] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0033] See also Figure 1 The present invention provides an AI-based teaching quality evaluation method, comprising the following steps:

[0034] S1. Historical data collection and preprocessing: collect excellent course videos to ensure high video quality without obvious noise interference, use video processing software to extract audio tracks and convert them into standard formats;

[0035] S2. Establish a sound model, apply audio processing technology, extract the spectral characteristics of the teacher's lecture sound, use AI algorithm to establish an excellent teacher's sound model, and identify and simulate the pitch curve of excellent teachers;

[0036] S3. Establish a content model, use natural language processing technology to extract key knowledge points in teaching content, use AI algorithms to analyze the relevance of keywords and teaching textbook content, and establish an excellent teacher content model;

[0037] S4. Establish a knowledge point distribution model, use natural language processing technology to extract key knowledge points in the teaching content, and obtain the time points where the key knowledge points are located throughout the course. Use the AI ​​algorithm model to analyze the time distribution of key knowledge points, and establish a knowledge point distribution model for excellent teachers; by analyzing the structure and content distribution of teachers' lectures, determine the position of the teaching focus in the course time. The evaluation criteria such as the frequency of the teaching focus and the rationality of the position can be obtained. Since students' attention changes as the course progresses, the corresponding relationship between the teaching content distributed at different time points and the distribution of the time points when students' attention is concentrated will greatly affect the students' learning effect. Therefore, by analyzing the time distribution of key knowledge points, the course structure can be optimized to ensure that the most important teaching content is arranged during the period when students' attention is most concentrated. In addition, the knowledge point distribution model can also help teachers understand which knowledge points require more explanation time and which knowledge points students may have already mastered, so as to adjust teaching strategies and improve teaching efficiency.

[0038] S5. Speech recognition and teaching quality assessment: Use speech recognition technology to transcribe the teacher's lecture content into audio, use audio processing technology and natural language processing technology to extract the teacher's lecture sound, key knowledge points and key knowledge points time distribution, use AI algorithm to analyze its similarity with the sound model, content model and knowledge point distribution model of excellent teachers, and take the average value as the evaluation indicator of teaching quality.

[0039] In this embodiment, preferably, in S2, Fourier transform is used to extract spectral features. First, the audio signal is collected to obtain the signal representation f(t) in the time domain. Then, the continuous Fourier transform formula is applied to calculate the spectral representation F(ω) of the signal. The formula is as follows:

[0040] F(ω)=∫_{-∞}^{∞}f(t)e^{-iωt}dt

[0041] Among them, F(ω) represents the spectrum of the signal, ω is the angular frequency, i is the imaginary unit, and e is the base of the natural logarithm. By calculating this integral, we can get the amplitude and phase information of the signal at different frequencies, that is, the spectrum characteristics. These features play a vital role in the establishment of sound models. They not only reflect the basic physical properties of audio signals, but can also be used for further audio analysis and processing tasks, such as feature extraction, signal recognition, noise elimination, etc.

[0042] In this embodiment, preferably, the establishment of the excellent teacher voice model uses a Hidden Markov Model (HMM) algorithm for modeling and analysis. In the process of establishing the voice model, the spectrum feature is used as input data, and the voice curve formula is as follows:

[0043] P(O|λ)=∑_{Q}P(O|Q,λ)P(Q|λ)

[0044] Where P(O|Q,λ) is the probability of observing the sequence O given the state sequence Q and the model parameter λ.

[0045] In the process of establishing the sound model, the spectral features are used as input data for modeling and analysis through algorithms such as the Hidden Markov Model (HMM). The HMM can identify and simulate the teacher's pitch curve, thereby achieving an accurate description and evaluation of the teacher's voice characteristics.

[0046] In this embodiment, preferably, when the excellent teacher content model is established, the TF-IDF (TermFrequency-Inverse Document Frequency) algorithm is used to determine the importance of key knowledge points in the text. The higher the TF-IDF value, the greater the importance of the word in the document. The calculation formula is:

[0047] \text{TF-IDF}(t,d)=\text{TF}(t,d)\times\text{IDF}(t)

[0048] \text{TF}(t,d): the term frequency of word (t) in document (d), that is, the number of times word (t) appears in document (d) divided by the total number of all words in document (d), \text{IDF}(t): the inverse document frequency of word (t), calculated as (\log\left(\frac{N}{\text{df}(t)}\right)), where (N) is the number of all documents, and (\text{df}(t)) is the number of documents containing word (t).

[0049] Before searching for keywords, we first clean the collected text data, including removing stop words and punctuation marks, performing stem extraction or word form restoration, etc., to reduce interference and ensure the accuracy of the results.

[0050] In this embodiment, preferably, the knowledge point distribution model supervisor uses regression analysis to build a model, and based on the results of the regression analysis, explains the specific impact of key knowledge points and class time periods on the teaching effect, as well as the interaction between them, and the calculation formula is:

[0051] y=\beta_0+\beta_1x_1+\beta_2x_2+\epsilon

[0052] Among them, (y) is the teaching effect (dependent variable), (x_1): TF-IDF value of key knowledge points (independent variable), (x_2): class time (independent variable), (\beta_0): intercept term, (\beta_1,\beta_2): regression coefficient, indicating the influence of independent variable on dependent variable, (\epsilon): error term.

[0053] Use regression analysis (such as linear regression or logistic regression) to evaluate the impact of key knowledge point distribution on teaching effectiveness. Build a model with teaching effectiveness as the dependent variable, and the TF-IDF value of key knowledge points and class duration as independent variables to analyze their contribution to teaching effectiveness.

[0054] In the process of model construction, we first identified the TF-IDF value of key knowledge points and the length of class time as the main independent variables, because they are the key factors affecting teaching effectiveness. By collecting a large amount of teaching data, we standardized these data to ensure the accuracy and generalization ability of the model. Then, we used regression analysis techniques, especially linear regression models, to quantify the relationship between these independent variables and teaching effectiveness. Through the size and direction of the regression coefficient, we can judge the specific degree of influence of each independent variable on the teaching effect. For example, a large positive regression coefficient indicates that the TF-IDF value of the knowledge point is positively correlated with the teaching effect, that is, the higher the frequency of occurrence of the knowledge point, the better the teaching effect. Similarly, the regression coefficient of the length of class time can help us understand the impact of the length of class time on the teaching effect. In this way, teachers and educational institutions can better understand how to optimize course content and time allocation to improve teaching effectiveness.

[0055] In this embodiment, preferably, the similarity between the teacher's lecture content and the excellent teacher model is calculated using cosine similarity, and the calculation formula is:

[0056] \text{cosine_similarity}(A,B)=\frac{A\cdot B}{|A||B|}

[0057] Among them, A and B represent two vectors respectively, A·B represents the dot product of vectors A and B, that is, the sum of the products of corresponding elements of the two vectors, ||A|| and ||B|| represent the Euclidean norm (length) of vectors A and B respectively, that is, the square root of the sum of the squares of each element of the vector.

[0058] By calculating the cosine similarity between the teacher's lecture content and the excellent teacher model, a value between -1 and 1 can be obtained. The closer the value is to 1, the higher the similarity. This similarity value reflects the degree of similarity between the teacher's lecture content and the excellent teacher model in terms of sound characteristics, content structure, and knowledge point distribution, and can thus be used as an evaluation indicator of teaching quality.

[0059] By analyzing the distribution of these similarity values, we can identify strengths and weaknesses in teaching. For example, if a teacher's lecture content on a specific knowledge point is generally less similar to that of an excellent teacher model, then this may indicate that the teacher needs further training and improvement in teaching these knowledge points. On the contrary, if the similarity is generally high, it means that the teacher has done a good job in these aspects and can serve as a role model for other teachers to learn from.

[0060] In addition, by regularly calculating and comparing the similarity between teachers’ lecture content and excellent teacher models, educational institutions can monitor the changing trend of teaching quality and adjust teaching strategies and methods in a timely manner. For example, if the similarity is found to be decreasing, it may be necessary to update the teaching content or retrain teachers to ensure the continuous improvement of teaching effectiveness.

[0061] In this embodiment, preferably, the teaching quality evaluation standard is that an average similarity of 90% or more is excellent, an average similarity of 80%-89% is good, an average similarity of 70%-79% is fair, and an average similarity below 70% is poor.

[0062] In practical applications, by setting such evaluation criteria, educational institutions can more intuitively evaluate teachers' teaching effectiveness. Teachers can understand their own teaching performance based on their own scores and make targeted improvements. At the same time, these standards also provide a clear basis for comparison between teachers, which helps to stimulate healthy competition among teachers and promote the improvement of overall teaching quality. In addition, through regular evaluation, educational administrators can track teachers' teaching effectiveness, promptly discover and solve problems that arise in the teaching process, and ensure the achievement of teaching goals.

[0063] The working principle and use process of the present invention: When the device is used, the teaching course characteristics of excellent teachers are extracted through the AI ​​algorithm, and the sound model, content model and knowledge point distribution model are established respectively. Then, the AI ​​algorithm is used to analyze the similarity between the two major factors of teacher teaching, namely, sound and teaching content, and the excellent teachers, and finally a comprehensive evaluation of the teacher's teaching is obtained. Specifically, the sound model focuses on the teacher's voice characteristics, such as speaking speed, pitch, volume, etc., while the content model focuses on the depth and breadth of the teaching content, and the knowledge point distribution model analyzes the coverage and distribution pattern of knowledge points in the teaching content. Through the comprehensive analysis of these models, the teaching quality of teachers can be quantitatively evaluated, thereby providing teachers with a basis for improving teaching methods. In addition, the device can also give specific improvement suggestions based on the evaluation results to help teachers achieve better results in future teaching activities.

[0064] In addition, by comparing the teacher's voice features with the standard model, it is possible to evaluate whether the teacher's speaking speed, intonation, emotional expression, etc. meet the teaching requirements, thereby providing teachers with directions for improvement. Combining the teacher's voice features and teaching content, the AI ​​algorithm can further analyze the teacher's teaching style, such as whether it is lively and interesting, whether it is clear and organized, and whether it can effectively attract students' attention. By evaluating the teacher's voice performance when explaining different knowledge points, it can be found which parts of the teacher may need to be strengthened and which parts have been done well, thereby providing teachers with personalized teaching feedback. Ultimately, the method provided by the present invention can provide education managers with an objective and comprehensive teacher teaching effectiveness evaluation tool to help them make more reasonable teaching decisions and teacher training plans.

[0065] In the process of teaching quality evaluation, we can not only collect valuable feedback data, but also use this data to continuously optimize and improve the sound model, content model and knowledge point distribution model. By adopting scientific methods such as cross-validation, we can comprehensively evaluate the performance of these models and find out their shortcomings. On this basis, we further adjust and optimize the algorithm parameters in order to achieve higher accuracy and efficiency. In this way, we can not only ensure the continuous improvement of teaching quality, but also continuously improve and optimize each link in the teaching process, and ultimately maximize the teaching effect.

[0066] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A teaching quality evaluation method based on AI, characterized in that: It includes the following steps: S1. Historical data collection and preprocessing: collect excellent course videos to ensure high video quality without obvious noise interference, use video processing software to extract audio tracks and convert them into standard formats; S2. Establish a sound model, apply audio processing technology, extract the spectral characteristics of the teacher's lecture sound, use AI algorithm to establish an excellent teacher's sound model, and identify and simulate the pitch curve of excellent teachers; S3. Establish a content model, use natural language processing technology to extract key knowledge points in teaching content, use AI algorithms to analyze the relevance of keywords and teaching textbook content, and establish an excellent teacher content model; S4. Establish a knowledge point distribution model, use natural language processing technology to extract key knowledge points in the teaching content, and obtain the time points where the key knowledge points are located throughout the course, use the AI ​​algorithm model to analyze the time distribution of key knowledge points, and establish a knowledge point distribution model for excellent teachers; S5. Speech recognition and teaching quality assessment: Use speech recognition technology to transcribe the teacher's lecture content into audio, use audio processing technology and natural language processing technology to extract the teacher's lecture sound, key knowledge points and key knowledge points time distribution, use AI algorithm to analyze its similarity with the sound model, content model and knowledge point distribution model of excellent teachers, and take the average value as the evaluation indicator of teaching quality.

2. The AI-based teaching quality evaluation method according to claim 1, characterized in that: In S2, Fourier transform is used to extract spectral features. First, the audio signal is collected to obtain the signal representation f(t) in the time domain. Then, the continuous Fourier transform formula is applied to calculate the spectral representation F(ω) of the signal. The formula is as follows: F(ω)=∫_{-∞}^{∞}f(t)e^{-iωt}dt Among them, F(ω) represents the spectrum of the signal, ω is the angular frequency, i is the imaginary unit, and e is the base of the natural logarithm. By calculating this integral, we can obtain the amplitude and phase information of the signal at different frequencies, that is, the spectrum characteristics.

3. The AI-based teaching quality evaluation method according to claim 1, characterized in that: The establishment of the excellent teacher voice model uses the Hidden Markov Model (HMM) algorithm for modeling and analysis. In the process of establishing the voice model, the spectrum characteristics are used as input data, and the voice curve formula is as follows: P(O|λ)=∑_{Q}P(O|Q,λ)P(Q|λ) Where P(O|Q,λ) is the probability of observing the sequence O given the state sequence Q and the model parameter λ.

4. The AI-based teaching quality evaluation method according to claim 1, characterized in that: When the excellent teacher content model is established, TF-IDF (Term Frequency-Inverse Document The TF-IDF algorithm is used to determine the importance of key knowledge points in the text. The higher the TF-IDF value, the greater the importance of the word in the document. The calculation formula is: \text{TF-IDF}(t,d)=\text{TF}(t,d)\times\text{IDF}(t) \text{TF}(t,d): Term Frequency of word (t) in document (d), that is, the number of times word (t) appears in document (d) divided by the total number of words in document (d), \text{IDF}(t): Inverse Document Frequency of word (t), calculated as (\log\left(\frac{N}{\text{df}(t)}\right)), where (N) is the number of all documents and (\text{df}(t)) is the number of documents containing word (t).

5. The AI-based teaching quality evaluation method according to claim 1, characterized in that: The knowledge point distribution model supervisor uses regression analysis to build a model. Based on the results of regression analysis, the supervisor explains the specific impact of key knowledge points and class time periods on teaching effectiveness, as well as the interaction between them. The calculation formula is: y=\beta_0+\beta_1x_1+\beta_2x_2+\epsilon Among them, (y) is the teaching effect (dependent variable), (x_1): TF-IDF value of key knowledge points (independent variable), (x_2): class time (independent variable), (\beta_0): intercept term, (\beta_1,\beta_2): regression coefficient, indicating the influence of independent variable on dependent variable, (\epsilon): error term.

6. The AI-based teaching quality evaluation method according to claim 1, characterized in that: The similarity between the teacher's lecture content and the excellent teacher model is calculated using cosine similarity, and the calculation formula is: \text{cosine_similarity}(A,B)=\frac{A\cdot B}{|A||B|} Among them, A and B represent two vectors respectively, A·B represents the dot product of vectors A and B, that is, the sum of the products of corresponding elements of the two vectors, ||A|| and ||B|| represent the Euclidean norm (length) of vectors A and B respectively, that is, the square root of the sum of the squares of each element of the vector.

7. The AI-based teaching quality evaluation method according to claim 1, characterized in that: The teaching quality evaluation standard is that an average similarity of 90% or more is excellent, an average similarity of 80%-89% is good, an average similarity of 70%-79% is fair, and an average similarity of less than 70% is poor.

Images