Teaching path generation method and system based on large model

Abstract

The invention provides a teaching path generation method and system based on a large model, and the method comprises the steps: obtaining a classroom audio based on a classroom video, and extracting a classroom text in the classroom audio; then, obtaining all knowledge points in a preset course knowledge graph, and performing slicing processing on the classroom text according to each knowledge point by adopting a large language model; finally, according to the starting moment of each slice, the corresponding knowledge points are connected in a directed mode to generate a teaching path. By adopting the automatic analysis function of the large model, the teaching path can be quickly and accurately provided, the time required for teaching evaluation is remarkably shortened, and the working efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent teaching, and particularly relates to a teaching path generation method and system based on a large model. Background Art

[0002] In recent years, with the rapid development of artificial intelligence technology, especially the breakthroughs in the field of large language models, artificial intelligence has demonstrated powerful capabilities in semantic understanding, content generation, etc. At the same time, the education field has been continuously exploring how to use these advanced technologies to improve teaching methods, optimize teaching paths, and enhance learning effects. Traditional classroom teaching mainly relies on teachers' lectures and personal experience, while modern education pays more attention to the design and implementation of teaching paths. How to effectively describe teachers' teaching behaviors and present them in the form of teaching paths, so as to scientifically evaluate their teaching behaviors, is an important issue in current educational research.

[0003] As an important educational resource, classroom recording and broadcasting videos can completely record the teaching process, covering voice, images, and dynamic interaction information in the classroom. However, due to the lack of effective indexing and analysis means, these valuable resources are often not fully utilized. Current educational technology systems mainly rely on static materials such as lesson plans, students' homework, or exercises to extract knowledge points and construct teaching paths based on these knowledge points. This method can provide teaching guidance and evaluation to a certain extent, but essentially it can only indirectly reflect the situation of teaching activities. Lesson plans usually represent the teaching content pre-planned by teachers, while students' homework and exercises more reflect students' understanding and application of specific knowledge points, and cannot comprehensively capture the actual teaching process and dynamic interaction in the classroom.

[0004] Traditional classroom teaching is the most core link in the entire educational process. It is not only a process of knowledge transfer, but also an important place for communication, discussion, and instant feedback between teachers and students. In the classroom environment, teachers can dynamically adjust the teaching method, speed, and depth according to students' reactions, and introduce additional examples or exercises in a timely manner to strengthen understanding. Therefore, the actual situation of classroom teaching is directly related to the quality of teaching and the effectiveness of learning outcomes.

[0005] As a technical means to completely record classroom activities, recording and broadcasting videos can theoretically provide first-hand classroom data and be used to analyze the actual situation of teaching. However, most current recording and broadcasting systems, although they can truly record classroom situations, do not fully explore and utilize the value of these data. Summary of the Invention

[0006] The present invention provides a teaching path generation method and system based on a large model to solve the problem that classroom recording and broadcasting videos in the prior art are not fully utilized.

[0007] To solve the above technical problems, the embodiments of the present invention disclose the following technical solutions:

[0008] One aspect of the present invention provides a teaching path generation method based on a large model, including:

[0009] Obtain classroom audio based on the classroom video, and extract classroom text from the classroom audio;

[0010] Obtain all knowledge points in the preset curriculum knowledge graph;

[0011] Use a large language model to slice the classroom text according to each of the knowledge points;

[0012] Generate a teaching path by connecting the corresponding knowledge points in a directed manner according to the start time of each slice.

[0013] Optionally, the obtaining classroom audio based on the classroom video and extracting classroom text from the classroom audio includes:

[0014] Use a preset multimedia processing tool to perform audio-video separation processing on the classroom video, and use the obtained audio data as the classroom audio;

[0015] Use ASR speech recognition technology to convert the speech in the classroom audio into text data as the classroom text.

[0016] Optionally, before performing the step of using ASR speech recognition technology to convert the speech in the classroom audio into text data as the classroom text, the method further includes:

[0017] Use a preset audio denoising tool to perform noise reduction processing on the classroom audio.

[0018] Optionally, the using a large language model to slice the classroom text according to each of the knowledge points includes:

[0019] For each of the knowledge points, the classroom text is sliced in the following manner:

[0020] Use a large language model to determine whether there is text in the classroom text that matches the knowledge point;

[0021] If so, determine that the knowledge point has been explained; slice the classroom text based on the matching text, and record the start time of the slice as the explanation time of the knowledge point;

[0022] If not, determine that the knowledge point has not been explained.

[0023] Optionally, the generating a teaching path by connecting the corresponding knowledge points in a directed manner according to the start time of each slice includes:

[0024] Each knowledge point in the curriculum knowledge graph is used as a node to construct a teaching path. Among them, different identifiers are used to distinguish between the explained knowledge points and the unexplained knowledge points, and the knowledge points explained at adjacent times are connected, with the previously explained knowledge point pointing to the later explained knowledge point.

[0025] Optionally, before performing the step of slicing the classroom text according to each knowledge point using the large language model, the method further includes:

[0026] Performing optimization processing on the classroom text, where the optimization processing at least includes cleaning, standardization, sentence segmentation optimization, grammar correction, and adding timestamps.

[0027] Optionally, the method further includes:

[0028] Marking the starting knowledge point and the ending knowledge point of the course in the teaching path, where the starting knowledge point is the first explained knowledge point in the course, and the ending knowledge point is the last explained knowledge point in the course.

[0029] Optionally, the method further includes:

[0030] For each unexplained knowledge point, it is set in the teaching path in the following manner:

[0031] Obtaining two adjacent knowledge points of the unexplained knowledge point in the curriculum knowledge graph;

[0032] Setting the unexplained knowledge point between the two adjacent knowledge points in the teaching path.

[0033] Another aspect of the present invention provides a teaching path generation system based on a large model, including:

[0034] A classroom text acquisition module, configured to acquire classroom audio based on a classroom video and extract the classroom text from the classroom audio;

[0035] A knowledge point acquisition module, configured to acquire all knowledge points in a preset curriculum knowledge graph;

[0036] A slicing module, configured to slice the classroom text according to each knowledge point using a large language model;

[0037] A teaching path generation module, configured to generate a teaching path by directionally connecting the corresponding knowledge points according to the starting time of each slice.

[0038] A method and system for generating teaching paths based on large models disclosed by the present invention. First, classroom audio is obtained based on classroom videos, and classroom text in the classroom audio is extracted. Then, all knowledge points in a preset curriculum knowledge graph are obtained, and a large language model is used to slice the classroom text according to each knowledge point. Finally, according to the start time of each slice, the corresponding knowledge points are connected in a directed manner to generate a teaching path.

[0039] By adopting the automated analysis function of large models, the present invention can quickly and accurately provide teaching paths, significantly shortening the time required for teaching evaluation and improving work efficiency.

[0040] The section of the invention content is provided to introduce the selection of concepts in a simplified form, which will be further described in the specific implementation manners below. The section of the invention content is not intended to identify the important features or essential features of the present disclosure, nor is it intended to limit the scope of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] By describing the exemplary embodiments of the present disclosure in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present disclosure will become more apparent. Among them, in the exemplary embodiments of the present disclosure, the same reference numerals generally represent the same components.

[0042] Figure 1 It is a schematic flowchart of a method for generating a teaching path based on a large model provided by an embodiment of the present invention;

[0043] Figure 2 It is for an implementation provided by an embodiment of the present invention Figure 1 The schematic flowchart of step S100 in

[0044] Figure 3 It is a schematic structural diagram of a system for generating a teaching path based on a large model provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0045] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0046] As used herein, the term "including" and its variations mean open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "at least partially based on". The term "an example embodiment" and "an embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc. may refer to different or the same objects. There may also be other explicit and implicit definitions hereinafter.

[0047] Figure 1 The flowchart of a teaching path generation method based on a large model provided for an embodiment of the present invention is as Figure 1 shown, and the method includes the following steps:

[0048] Step S100: Obtain classroom audio based on the classroom video, and extract classroom text from the classroom audio.

[0049] Previously, obtain the teaching video of the classroom by using the camera system installed in the classroom. In an embodiment disclosed in the present invention, as Figure 2 shown, step S100 can be implemented in the following manner:

[0050] Step S101: Use a preset multimedia processing tool to perform audio-video separation processing on the classroom video, and use the obtained audio data as the classroom audio.

[0051] Audio data can be separated from the classroom video through a variety of preset multimedia processing tools, such as FFmpeg, Adobe Audition, Praat, etc.

[0052] Step S102: Use ASR (Automatic Speech Recognition) technology to convert the speech in the classroom audio into text data as the classroom text.

[0053] ASR (Automatic Speech Recognition) technology can parse and convert the speech signal of the classroom audio into editable text, so as to locate the time of explaining each knowledge point in the classroom in text form.

[0054] In an embodiment disclosed in the present invention, before performing step S102, use a preset audio denoising tool to perform noise reduction processing on the classroom audio.

[0055] To improve the accuracy of speech recognition, in the embodiments of the present invention, noise reduction processing is performed on the audio to remove background noise and make the speech clearer. A preset audio denoising tool can be used, such as the noise reduction filter of Adobe Audition, Audacity, SoX (Sound eXchange), Waves Clarity VX, or FFmpeg.

[0056] The first step of noise reduction is to analyze the type of noise. Common noises in classroom audio include the low-frequency noise of air conditioners and electric fans, the whispering of students, the sound of turning pages, and the ambient sound in the distance. Different noises require different denoising methods. For example, continuous background noise can use spectral denoising methods, while burst noises (such as coughing and the sound of dragging chairs) are suitable for threshold denoising or adaptive filtering methods.

[0057] Step S200: Obtain all the knowledge points in the preset course knowledge graph.

[0058] The course knowledge graph is a knowledge management method used to represent the content structure of a course. It organizes the knowledge points in the course according to hierarchical relationships, logical sequences, and relevance to support intelligent teaching analysis and personalized learning. Therefore, before automatically analyzing the classroom content and constructing the teaching path, it is necessary to first obtain all the knowledge points in the course knowledge graph. In the publicly disclosed embodiments of the present invention, all the knowledge points in the course knowledge graph can be obtained using existing technologies according to the form of the course knowledge graph.

[0059] Step S300: Use a large language model to slice the classroom text according to each knowledge point.

[0060] Since classroom teaching may contain a large amount of freely expressed content, the original text may contain repetitive, colloquial, or content unrelated to the knowledge points. In an embodiment disclosed in the present invention, it is necessary to optimize the classroom text. The optimization process at least includes cleaning, standardization, sentence segmentation optimization, grammar correction, and adding timestamps.

[0061] In an embodiment disclosed in the present invention, step S300 can be implemented in the following manner:

[0062] For each knowledge point, the classroom text is sliced in the following manner:

[0063] Use a large language model to determine whether there is text in the classroom text that matches the knowledge point. If so, it is determined that the knowledge point has been explained; based on the matching text, the classroom text is sliced once, and the start time of this slice is recorded as the explanation time of the knowledge point. If not, it is determined that the knowledge point has not been explained.

[0064] The present invention determines whether the knowledge points in the knowledge graph are involved in the classroom text through the semantic matching technology of the large language model. Specifically, the large language model can adopt the method of text vectorization, such as the embedding models of BERT and GPT, to convert the classroom text and knowledge points into vectors respectively, and calculate the semantic similarity. If the similarity is higher than the set threshold (for example, 0.8), it is considered that the knowledge point has been explained, otherwise it is considered not to have been explained.

[0065] For example, the classroom text is "Plants use sunlight through photosynthesis to synthesize organic matter. This process requires carbon dioxide and water and releases oxygen.", and the knowledge point is "photosynthesis".

[0066] The large language model calculates the cosine similarity through the embedded vectors. If the matching degree is high enough, it can be determined that "photosynthesis" has been explained.

[0067] If a knowledge point is determined to have been explained, the system will perform a slicing process on the classroom text based on the matching text position, and record the start time of the slice as the explanation time of the knowledge point. The output slice format is: knowledge point name_class start time_video slice start time. For example, microorganisms in the atmosphere_2024-11-01-11:05_22:52 means that in the classroom starting at 11:05 on November 1, 2024, starting at the 22nd minute and 52nd second, the knowledge point "microorganisms in the atmosphere" was taught in the classroom.

[0068] This way enables each knowledge point to be accurately located at a specific time point in the classroom video, facilitating the construction of subsequent visual teaching paths.

[0069] If a matching text for a certain knowledge point is not found in the classroom text, it is determined that the knowledge point has not been explained. In the visual teaching path, such knowledge points can be marked in dark gray to prompt the teacher to supplement the explanation in subsequent courses. For example, if a teacher should explain "photosynthesis" and "respiration" in the curriculum plan, but only "photosynthesis" is detected in the classroom text, the system can prompt that "photosynthesis" has been explained and "respiration" has not been explained.

[0070] As the teaching progresses throughout the semester, new slices are generated after each class, and the knowledge points are connected into a teaching path according to the time sequence, helping the teacher review the teaching progress and also providing accurate learning references for teaching supervisors and students.

[0071] Step S400: Generate a teaching path by connecting the corresponding knowledge points in a directed manner according to the start time of each slice.

[0072] Each knowledge point in the course knowledge graph is used as a node to construct a teaching path. Among them, different identifiers are used to distinguish the explained knowledge points and the unexplained knowledge points, and the knowledge points explained at adjacent times are connected, with the previously explained knowledge point pointing to the later explained knowledge point.

[0073] When constructing the teaching path, the system takes each knowledge point in the course knowledge graph as a node and, based on the analysis results of the classroom text, distinguishes between the explained and unexplained knowledge points. The explained knowledge points are usually represented by a highlighted color (such as blue or green), while the unexplained knowledge points are represented by a dark gray color to remind the teacher that this content has not been covered yet.

[0074] The connection relationship between nodes is determined by the chronological order of classroom explanations. For the knowledge points already explained in class, the system sorts them according to their explanation time and establishes a directed edge between adjacent knowledge points, that is, the previously explained knowledge point points to the later explained knowledge point. This way can intuitively reflect the teacher's teaching order and also reveal the logical structure of the course content. For example, if "photosynthesis" is explained first in class and then "cellular respiration" is explained, then a connection from "photosynthesis" to "cellular respiration" will be established in the teaching path.

[0075] The teaching path can not only be used for tracking the teaching progress but also help teachers discover the teaching connection between knowledge points. If a certain knowledge point does not appear in the teaching path, it means that the teacher may have skipped this content or did not explain it according to the course plan. In addition, if a certain knowledge point should be used as a leading content but the actual explanation order is adjusted, for example, "cellular respiration" is explained before "photosynthesis", the system can also mark this anomaly to remind the teacher to adjust the teaching order to be more in line with the logic of the knowledge system.

[0076] As the course progresses, the teaching path will be updated dynamically. After each class is completed, the system will, based on the latest analysis results of the classroom recording, supplement new knowledge points and adjust the existing teaching path structure. This method of tracking knowledge points based on the teaching path can provide a more intuitive analysis of knowledge coverage for teachers, students, and teaching managers, help teachers optimize classroom teaching, improve students' learning efficiency, and ensure the integrity and systematicness of the course.

[0077] In an embodiment disclosed by the present invention, the starting knowledge point and the ending knowledge point of the course are marked in the teaching path. The starting knowledge point is the first knowledge point explained in the course, and the ending knowledge point is the last knowledge point explained in the course.

[0078] In the process of constructing a classroom teaching path, in order to more clearly display the order of explaining knowledge points and the course structure, the embodiment of the present invention marks the starting knowledge point and the ending knowledge point of the course in the teaching path of the teaching path, so that teachers, students and teaching supervisors can intuitively understand the starting point and final destination of the course explanation.

[0079] The starting knowledge point refers to the knowledge point that is explained first during the classroom teaching process, that is, the first knowledge point involved in the entire course timeline. It is usually the basic content of the course and provides necessary leading information for subsequent knowledge points. For example, in the "Cell Biology" course, the teacher may first introduce the "Basic Structure of Cells", then "Basic Structure of Cells" is the starting knowledge point of the course. In the teaching path, the starting knowledge point can be highlighted with special marks (such as bold, enlarged, using specific colors or symbols) to indicate that this knowledge point is the beginning of the entire course content.

[0080] The final knowledge point refers to the last knowledge point explained in the classroom teaching process, that is, the last knowledge point in the explanation sequence. It is usually the summary, application or expansion content of the course, and it takes over the explanation of all the previous knowledge points. For example, in the "Ecology" course, the teacher may discuss "Ecosystem Stability" at the end, then "Ecosystem Stability" is the final knowledge point of this class. Similarly, in the teaching path, a specific color or icon can be used to highlight the final knowledge point to indicate that the knowledge point is the final destination of this class.

[0081] In one embodiment disclosed in the present invention, for each unexplained knowledge point, it is set in the teaching path in the following manner:

[0082] (1) Obtain two adjacent knowledge points of the unexplained knowledge points in the course knowledge graph.

[0083] There is a dependency or order relationship between the knowledge points in the course knowledge graph. By querying the graph, two knowledge points adjacent to the unexplained knowledge point can be obtained. Adjacent knowledge points are knowledge points that are closely related to the unexplained knowledge point.

[0084] (2) Set the unexplained knowledge points between two adjacent knowledge points in the teaching path.

[0085] The unexplained knowledge points are set between the adjacent knowledge points determined in (1). In this way, the unexplained knowledge points will be regarded as a bridge between two adjacent knowledge points, helping to build a more reasonable learning path and ensuring that the new node does not disrupt the order of the original knowledge points. After the unexplained knowledge point is explained, the label of the node corresponding to the knowledge point is changed in the teaching path, and a directed connection relationship between it and other nodes is described.

[0086] As the semester progresses from start to finish, each time a class is completed in a course, it is processed in the above manner. Teachers can conveniently view the teaching path and identify which knowledge points have not been covered yet, allowing them to fill in the gaps. Teaching supervisors can also promptly check the teaching situation of teachers and intervene in a timely manner when necessary.

[0087] Figure 3 The following is a schematic structural diagram of a teaching path generation system based on a large model provided by an embodiment of the present invention, as Figure 3 shown, including the following modules:

[0088] Classroom text acquisition module 1, configured to acquire classroom audio based on a classroom video and extract the classroom text from the classroom audio.

[0089] Knowledge point acquisition module 2, configured to acquire all knowledge points in a preset course knowledge graph.

[0090] Slicing module 3, configured to slice the classroom text according to each knowledge point using a large language model.

[0091] Teaching path generation module 4, configured to generate a teaching path by connecting the corresponding knowledge points in a directed manner according to the start time of each slice.

[0092] The various embodiments of the present disclosure have been described above. The above description is exemplary and not exhaustive, and is also not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art in the technical field without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements to the technologies in the market, or to enable other ordinary skill in the art in the technical field to understand the disclosed embodiments.

Claims

1. A teaching path generation method based on a large model, characterized in that: include: Obtain classroom audio based on classroom video, and extract classroom text from the classroom audio; Get all the knowledge points in the preset course knowledge graph; Use a large language model to slice the classroom text according to each of the knowledge points; According to the starting time of each slice, the corresponding knowledge points are connected in a directed manner to generate a teaching path.

2. The method according to claim 1, characterized in that: The method of obtaining classroom audio based on classroom video and extracting classroom text from the classroom audio includes: Use the preset multimedia processing tools to separate the audio and video of the classroom video, and use the obtained audio data as the classroom audio; ASR speech recognition technology is used to convert the speech in the classroom audio into text data as classroom text.

3. The method according to claim 2, characterized in that Before executing the step of converting the speech in the classroom audio into text data using the ASR speech recognition technology as the classroom text, the method further includes: Use the preset audio denoising tool to reduce noise in classroom audio.

4. The method according to any one of claims 1 to 3, characterized in that: The use of a large language model to slice the classroom text according to each of the knowledge points includes: For each of the above-mentioned knowledge points, the classroom text is sliced ​​in the following way: A large language model is used to determine whether there is text matching the knowledge point in the classroom text; If yes, determine that the knowledge point has been explained; slice the classroom text based on the matched text, and record the starting time of the slice as the explanation time of the knowledge point; If not, it is determined that the knowledge point has not been explained.

5. The method according to claim 4, characterized in that The method of generating a teaching path by connecting the corresponding knowledge points in a directed manner according to the starting time of each slice includes: Each knowledge point in the course knowledge graph is used as a node to construct a teaching path, in which different identifiers are used to distinguish between knowledge points that have been explained and knowledge points that have not been explained, and the knowledge points explained at adjacent times are connected, and the knowledge points explained first point to the knowledge points explained later.

6. The method according to claim 1, characterized in that Before executing the step of using the large language model to slice the classroom text according to each of the knowledge points, the method further includes: The classroom text is optimized, and the optimization process at least includes cleaning, standardization, sentence segmentation optimization, grammar correction and adding timestamp.

7. The method according to claim 5, characterized in that The method further comprises: The starting knowledge point and the ending knowledge point of the course are marked in the teaching path. The starting knowledge point is the first knowledge point explained in the course, and the ending knowledge point is the last knowledge point explained in the course.

8. The method according to claim 5, characterized in that The method further comprises: For each unexplained knowledge point, it is set in the teaching path in the following way: Obtain two adjacent knowledge points of the unexplained knowledge point in the course knowledge graph; The unexplained knowledge point is set between the two adjacent knowledge points in the teaching path.

9. A teaching path generation system based on a large model, characterized in that: include: A classroom text acquisition module is configured to acquire classroom audio based on classroom video and extract classroom text from the classroom audio; A knowledge point acquisition module is configured to acquire all knowledge points in a preset course knowledge graph; A slicing module is configured to slice the classroom text according to each of the knowledge points using a large language model; The teaching path generation module is configured to generate a teaching path by connecting the corresponding knowledge points in a directed manner according to the starting time of each slice.

Images