Educational resource matching management method and system based on intelligent perception

Abstract

The invention discloses an educational resource matching management method and system based on intelligent perception, and belongs to the technical field of educational management. The system comprises an intelligent sensing module, a dynamic analysis module, a resource matching module and a display management module. The intelligent sensing module is used for collecting an educational resource library, a teaching outline of a teacher object in a live broadcast process and a feedback result of each student object; the dynamic analysis module is used for analyzing knowledge points contained in a teaching outline, analyzing live broadcast content and positioning the live broadcast content to a current explanation knowledge point; calculating a difficulty index by combining an educational resource library, setting a feedback frequency, and judging whether to trigger early warning according to a feedback result; the resource matching module analyzes an educational resource library and draws a knowledge point path when triggering early warning, and divides an explanation path ending with a current explanation knowledge point; calculating a matching index of each explanation path, and generating a reference manuscript after weighted summary; and the display management module is used for pushing the reference manuscript to the client of the teacher object for display.

Description

Technical Field

[0001] This invention relates to the field of education management technology, specifically to a method and system for matching and managing educational resources based on intelligent sensing. Background Technology

[0002] With the deep integration of information technology and education, online live-streaming teaching has become a mainstream model of distance education. It breaks down the limitations of time and space, providing a flexible learning environment for teachers and students. However, in the real-time interactive context of live-streaming teaching, how to dynamically perceive teaching effectiveness, adjust teaching strategies in a timely manner, and intelligently match educational resources has become a key challenge in improving the quality of online education.

[0003] Currently, traditional online teaching platforms generally have certain limitations. First, the teaching feedback mechanism is lagging and coarse-grained, relying on post-class assessments and lacking real-time guidance. Real-time feedback data is simple and lacks in-depth analysis, resulting in limited effectiveness of teaching adjustments. Second, the analysis of teaching content lacks intelligent perception and cannot analyze content in a fine-grained manner. It cannot identify specific knowledge points or related historical records, making teaching a "black box" lacking intelligent assistance. Third, resource recommendation methods are static and lack context adaptability. Based on keywords or collaborative filtering, static materials are recommended without considering dynamic contexts such as difficulty, student acceptance, or teacher preferences, resulting in recommendations that are general but lack specificity. Finally, early warning and decision-making support mechanisms are lacking. There is a lack of quantitative analysis and risk warning capabilities; problems are only discovered after they accumulate, missing the opportunity for intervention. Teachers rely on experience, lack data support, and struggle to cope with complex scenarios. In summary, existing online live teaching technologies have significant shortcomings in dynamic perception, real-time analysis, intelligent early warning, and precise resource matching during the teaching process. Therefore, there is an urgent need for a technical solution that can deeply integrate intelligent perception technology with educational resource management to achieve refined and adaptive management of the entire teaching process. Summary of the Invention

[0004] The purpose of this invention is to provide an educational resource matching and management method and system based on intelligent perception, so as to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, this invention provides an educational resource matching and management method based on intelligent perception, comprising:

[0006] S100: Collect educational resource database, as well as activity data from students and teachers during the live stream. Analyze the teaching syllabus from the teacher participants and the feedback from each student participant within the activity data.

[0007] The educational resource database contains various teaching records, each of which includes a recorded video and a teaching syllabus.

[0008] Live streaming refers to a remote classroom teaching model that uses network communication technology to transmit teaching content from teachers and feedback data from students in real time.

[0009] Teachers, acting as the initiators in the live stream, use media devices to demonstrate and present teaching content, and dynamically adjust their teaching strategies based on feedback data from students.

[0010] Students, acting as the receivers in the live stream, receive teaching content and engage in learning activities through the interactive interface. After processing the feedback data, they submit it to the teachers for reference.

[0011] A syllabus is a structured document containing different knowledge points, each with a description. Feedback data refers to the set of binary choices students make regarding their understanding of the teaching content during the learning process.

[0012] Student feedback data is a set of binary choices, which reduces the complexity of data processing, but through batch collection and real-time transmission, it can efficiently reflect the overall learning situation.

[0013] By combining a lightweight data model, we can ensure that excessive latency is not introduced in high-concurrency live streaming scenarios, while providing a standardized data format for subsequent AI analysis.

[0014] Before the live broadcast begins, the teacher prepares a teaching syllabus based on the content of the lesson. And input it into the teaching platform. During the live broadcast, the teacher and students follow the teaching syllabus. Each knowledge point will be explained.

[0015] Ensuring the real-time nature, integrity, and structure of data provides reliable input for subsequent intelligent analysis, thereby supporting the dynamic response capability of the entire system.

[0016] S200: Analyze the knowledge points included in the teaching syllabus, analyze the live broadcast content, and locate the currently explained knowledge point. Calculate the difficulty index using the educational resource database, set the feedback frequency, and determine whether to trigger an alert based on student feedback. Specifically, this includes:

[0017] S201. AI technology is used to analyze the live broadcast content in real time and simultaneously output a text summary. The similarity between each knowledge point in the teaching syllabus and the text summary is calculated in real time, and the knowledge point with the highest similarity is selected as the current knowledge point to be explained. Specifically, this includes:

[0018] S2011. A speech recognition engine is used to convert the speech in the live broadcast content into text. At the same time, OCR technology is used to recognize the images in the live broadcast content and extract the text. Simultaneously, the text corresponding to the speech and images is merged into a text summary and continuously output.

[0019] By combining a speech recognition engine and OCR technology with a deep learning model, it can handle noise and diverse content in live broadcasts, improving the accuracy of summarization.

[0020] S2012, Analysis of Teaching Syllabus For each knowledge point description, calculate the average word count of all knowledge point descriptions. Set the interval duration. It also analyzes the duration and word count of the continuous output of text summaries.

[0021] S2013, will continue to output for a duration greater than And the number of words is greater than The text summaries are packaged into data packets, and the text summaries and teaching syllabi in the data packets are calculated separately. The semantic similarity of the content descriptions of each knowledge point.

[0022] Data packets are packaged one by one. When the packaging conditions are met, the text digest within this period is immediately packaged into a data packet. There is no duplication of text digests in different data packets.

[0023] S2014. Data packets are arranged chronologically, and the knowledge point with the highest semantic similarity is selected and associated with each data packet. The most recently generated data packet is identified, and the knowledge point associated with that data packet is used as the current knowledge point to be explained. .

[0024] The design of data packet packaging conditions avoids frequent calculations and optimizes real-time performance. Semantic similarity calculation adopts models based on word embedding or BERT to ensure semantic consistency of knowledge point associations.

[0025] By generating text summaries through multimodal analysis and calculating semantic similarity with the teaching syllabus, the current knowledge point being explained can be accurately located.

[0026] S202. Select resources from the educational resource database that include the teaching syllabus. The teaching records are used to obtain the recorded videos, identify the location and duration of the currently explained knowledge point, and then calculate the difficulty index. Specifically, this includes:

[0027] S2021, Obtain all documents containing the syllabus. The teaching records are analyzed, and the text summary output under the timeline of the recorded video in each teaching record is analyzed. Data packages are generated sequentially and associated with knowledge points, with all data packages arranged in chronological order.

[0028] S2022, Mark the currently explained knowledge point Analyze the associated data packets, identify the distribution of marked data packets in the data packet sequence corresponding to each teaching record, and substitute them into the formula to calculate the knowledge point being explained. Difficulty level:

[0029] ;

[0030] In the formula, To include the syllabus The total number of all teaching records For the first The number of all tagged data packets in each teaching record; and It is a constant, and .

[0031] parameter and The design reflects the weighting difference between labeled and unlabeled data packets, expanding the definition of difficulty by considering not only the explanation time of the knowledge point itself but also the impact of repeated explanations, thus more comprehensively reflecting the complexity and depth of the knowledge point. This design avoids the one-sidedness of relying solely on frequency, making the difficulty assessment more closely aligned with actual teaching scenarios.

[0032] and The first The first teaching record The duration of continuous output of each tagged data packet and the number of words in the text digest. and The first The total duration of continuous output of all data packets in the teaching record and the total number of words in the text summary.

[0033] and The first The first teaching record The total duration of continuous output of all untagged packets and the total number of words in the text digest between the first tagged packet and the next tagged packet.

[0034] The difficulty index is used to assess the teaching difficulty of a certain knowledge point by analyzing the duration and text volume of data packets directly related to the knowledge point in historical teaching records, as well as the overall impact of the content between these data packets.

[0035] The formula, combined with weighted calculation, aims to quantify the complexity and depth of knowledge points, reflect the concentration and repetition of explanations, and thus provide a basis for adjusting teaching strategies in the future.

[0036] Calculating the total duration of continuous output of all unlabeled data packets and the total number of words in the text summary between labeled data packets is to demonstrate the importance of repeatedly explained knowledge points. Simply analyzing the total output duration of labeled data packets and the total number of words in the text summary, however, cannot reflect the impact of repeated explanations.

[0037] By analyzing the distribution of knowledge points in historical teaching records, calculations were performed. An index quantifies the difficulty of teaching a knowledge point.

[0038] S203. Set the feedback frequency based on the difficulty index, and collect student feedback during the live stream according to the feedback frequency. Calculate the risk index based on the difficulty index and feedback results, and determine whether to trigger a risk warning based on the risk index. Specifically, this includes:

[0039] S2031, Computation Teaching Syllabus The average difficulty index of all knowledge points Preset reference frequency According to the formula: Set the feedback frequency after rounding the calculation. .

[0040] Feedback frequency Proportional to the difficulty index, the concept of adaptive sampling is extended: high-difficulty knowledge points require more frequent feedback collection to capture potential problems.

[0041] S2032, The teaching platform is based on feedback frequency. The questionnaire is sent to students at regular intervals, and feedback from each student is collected within a preset time period. The feedback includes both negative and positive feedback.

[0042] S2033. Analyze the proportion of negative feedback in each batch of feedback results, and calculate the risk index by substituting it into the formula along with the difficulty index. If the risk index is greater than the threshold, a risk warning is triggered. Risk Index The calculation formula is as follows:

[0043] ;

[0044] In the formula, A constant greater than 1 For the total batch of feedback results, For the first The percentage of negative feedback in the batch feedback results.

[0045] Risk Index The formula introduces logarithmic functions and constants. This smooths out abrupt changes in difficulty and feedback ratio, avoiding false alarms. A weighted average of negative feedback ratios ensures the robustness of the early warning system.

[0046] The risk index is used to determine whether there are risks in the teaching process and whether an early warning needs to be triggered.

[0047] Based on a comparison of the current difficulty of knowledge points with the historical average difficulty, as well as the overall proportion of negative student feedback, a logarithmic function is used to balance abrupt changes.

[0048] The formula's output helps identify teaching obstacles, ensuring that the early warning mechanism is both robust and adaptable to real-time changes, thereby supporting timely intervention.

[0049] A data-driven early warning mechanism is achieved by setting a dynamic feedback frequency and calculating a risk index.

[0050] It enables intelligent monitoring and adaptive adjustment of the teaching process, and timely identification of teaching difficulties and student comprehension obstacles through quantitative indicators, thereby helping teachers dynamically optimize teaching strategies and improve classroom efficiency and interaction quality.

[0051] S300: When an alert is triggered, analyze the teaching records in the educational resource database and draw knowledge point paths, dividing the explanation paths into those ending with the currently explained knowledge point. Calculate the matching index for each explanation path, and generate a reference document after weighted summarization. Specifically, this includes:

[0052] S301. When a risk warning is triggered, obtain the data packet sequence corresponding to each teaching record. Draw the knowledge point path for each teaching record according to the chronological order of the data packets and the associated knowledge points.

[0053] Each knowledge point path consists of interconnected knowledge points that are allowed to appear repeatedly. The same knowledge point may appear again or multiple times after a certain number of knowledge points.

[0054] S302. Divide each knowledge point path into knowledge point categories. For each explanation path ending with a sample path generated from the live stream content, a matching index is calculated between the explanation path and the sample path. Specifically, this includes:

[0055] S3021, Analyzing Knowledge Point Paths Chinese knowledge points The distribution location of each knowledge point Establish end nodes. After merging all consecutive end nodes, calculate the knowledge point paths. Number of lower-end nodes .

[0056] S3022, Following the knowledge point path The initial knowledge point serves as the starting point, and each subsequent node serves as the endpoint, forming a knowledge point path. Divide into There are three explanation paths. And so on, explanation paths are divided for each knowledge point.

[0057] S3023. Based on the temporal changes of the currently explained knowledge points in the live broadcast content, draw sample paths. Substitute these paths into the formula to calculate the matching index between each explanation path and the sample paths. :

[0058] ;

[0059] After simplification, we get:

[0060] ;

[0061] In the formula, This represents the number of knowledge points in the sample path. This refers to the number of identical knowledge points between the explanation path and the sample path. To explain the path and sample path, the first The semantic similarity of text summaries of data packets corresponding to the same knowledge points.

[0062] To explain the path, the first The duration of continuous output for data packets corresponding to the same knowledge point. To explain the path and sample path, the first The maximum duration of continuous output for data packets corresponding to the same knowledge point.

[0063] Matching Index The formula takes into account semantic similarity. and the ratio of continuous output time The multi-dimensional assessment of extended path similarity not only focuses on content matching but also considers the level of detail in the explanation, making the reference path more in line with actual teaching.

[0064] Semantic similarity is based on an NLP model to ensure content relevance. The time-to-study ratio reflects the level of detail in the teaching, ensuring that sample paths with longer durations for the same knowledge point achieve higher matching indices.

[0065] By mapping knowledge point paths and calculating matching indices, we can identify the historical paths most relevant to the current teaching.

[0066] S303. Using the explanation paths with a matching index greater than a threshold as reference paths, extract the text summaries of the data packets corresponding to the terminal nodes of all reference paths, assign weights according to the matching index, and summarize to generate reference texts. Specifically, this includes:

[0067] S3031. Obtain the matching index of each reference path and the text digest of the data packet corresponding to the end node. Normalize the matching indices of all reference paths using a weighted summation method to obtain the weight coefficients. The sum of the weight coefficients of all reference paths is 1.

[0068] S3032. Divide each text summary into entity units, calculate the cumulative score of the weight coefficient of each entity unit in all text summaries, and select entity units whose cumulative scores exceed a preset threshold to form a candidate summary set.

[0069] S3033. Use a text summarization model to remove semantically repetitive entity units from the candidate summary set, adjust the sentence arrangement according to the logical order and add conjunctions, and finally synthesize and generate a new text summary as a reference manuscript.

[0070] Entity unit segmentation and deduplication techniques, based on keyword extraction or sequence models, extend the intelligence of text summarization: by accumulating weight coefficients, high-frequency and important content is retained first.

[0071] Semantic deduplication and logical adjustments ensure the coherence and readability of the text. The final generated content not only reduces redundancy but also conforms to teaching logic, thus improving the teaching support effect.

[0072] Through weighted summarization and text processing, a concise and coherent reference document is generated.

[0073] It provides personalized teaching assistance by matching historical paths to generate targeted reference content, helping teachers quickly adjust their teaching methods, reduce teaching burden, and enhance teaching adaptability and effectiveness.

[0074] S400: Push the reference text to the teacher's client for display, providing the teacher with reference text assistance during the live broadcast.

[0075] Low-latency communication protocols are used to ensure timely delivery of documents, and UI design is combined to provide non-intrusive prompts, avoiding disruption to the teaching process.

[0076] The present invention also provides an education resource matching and management system based on intelligent sensing, including an intelligent sensing module, a dynamic analysis module, a resource matching module, and a display management module.

[0077] The intelligent sensing module is used to collect educational resource data, as well as the teaching syllabus of the teacher and the feedback results of each student during the live broadcast.

[0078] Collect data from the educational resource database and the activities of teachers and students during live broadcasts, and analyze the teaching syllabus and student feedback results.

[0079] By collecting and analyzing data in real time, basic inputs are provided to ensure the integrity and timeliness of teaching data, support the accuracy of subsequent analysis, and thus lay the foundation for dynamic teaching adjustments.

[0080] The dynamic analysis module is used to parse the knowledge points contained in the teaching syllabus, analyze the live broadcast content, and locate the currently explained knowledge point. It calculates a difficulty index based on the educational resource database, sets the feedback frequency, and determines whether to trigger an alert based on the feedback results.

[0081] AI technology is used to analyze live content in real time, locate the knowledge point being explained, calculate the knowledge point difficulty index DIF based on the educational resource database, set the feedback frequency PL, and then use the risk index FX to determine whether to trigger an alert.

[0082] It enables intelligent monitoring of the teaching process, automatically identifies teaching difficulties and students' comprehension levels, provides timely warnings of potential problems, and helps teachers optimize teaching strategies and improve classroom efficiency.

[0083] When an alert is triggered, the resource matching module analyzes the educational resource database and draws knowledge point paths, dividing the paths into explanation paths ending with the currently explained knowledge point. It calculates the matching index for each explanation path, weights and summarizes them, and then generates a reference document.

[0084] When an alert is triggered, the system analyzes the knowledge point paths in historical teaching records, divides the explanation paths ending with the current knowledge point, calculates the matching index PP for each path, and generates reference texts through weighted summarization.

[0085] It provides personalized teaching support by matching historical paths to generate targeted reference content, thereby enhancing teaching adaptability and effectiveness and reducing teachers' teaching burden.

[0086] The display management module is used to push reference documents to the teacher's client for display.

[0087] The generated reference documents are pushed to the teacher's client for display. This enables real-time visualization of auxiliary information, helping teachers quickly access references during live broadcasts, improving teaching response speed and interaction quality, and ultimately optimizing the overall teaching experience.

[0088] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0089] Dynamic Perception and Real-Time Decision-Making Throughout the Entire Process: By constructing a complete closed loop from data collection and analysis to intervention, this solution enables dynamic monitoring of the entire teaching process. Compared to existing technologies that rely on lagging and isolated data points, this solution can analyze teaching content and learner feedback in real time, giving the system immediate response and proactive decision-making capabilities, transforming teaching management from passive remediation to proactive optimization.

[0090] Precise mapping of teaching content to knowledge points: This solution breaks through the limitation of treating live broadcast content as a single data stream. Through intelligent semantic analysis technology, it automatically decomposes continuous teaching information and associates it with specific nodes in a structured knowledge system. This capability enables the teaching platform to accurately understand the teacher's current teaching intentions, laying a solid foundation for subsequent resource matching and strategy recommendation, and solving the problems of rough content understanding and weak correlation in traditional methods.

[0091] A data-driven adaptive early warning mechanism: This technical solution integrates content features with learner feedback to construct a multi-dimensional risk assessment model. It not only identifies potential teaching difficulties but also dynamically adjusts monitoring frequency and judgment thresholds based on real-time feedback, thereby achieving more flexible and accurate early warnings and overcoming the shortcomings of rigid early warning mechanisms and high false alarm rates in existing technologies.

[0092] Intelligent teaching strategy generation based on historical paths: When teaching obstacles are identified, the solution does not simply recommend static materials, but rather generates strategic guidance that integrates experiences from multiple successful teaching paths by mining high-quality historical teaching records. This method provides teachers not with a single answer, but with validated and flexibly applicable teaching ideas, greatly enhancing the practicality and effectiveness of decision support. Attached Figure Description

[0093] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0094] Figure 1 This is a flowchart illustrating the educational resource matching and management method based on intelligent sensing according to the present invention.

[0095] Figure 2 This is a schematic diagram of the structure of the intelligent sensing-based educational resource matching and management system of the present invention. Detailed Implementation

[0096] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0097] Example 1: Please refer to Figure 1 This invention provides a method for matching and managing educational resources based on intelligent sensing, comprising:

[0098] S100: Collect educational resource database, as well as activity data from students and teachers during the live stream. Analyze the teaching syllabus from the teacher participants and the feedback from each student participant within the activity data.

[0099] The educational resource database contains various teaching records, each of which includes a recorded video and a teaching syllabus.

[0100] In practice, live streaming refers to a remote classroom teaching model that uses network communication technology to transmit teaching content from teachers and feedback data from students in real time.

[0101] Teachers, acting as the initiators in the live stream, use media devices to demonstrate and present teaching content, and dynamically adjust their teaching strategies based on feedback data from students.

[0102] Students, acting as the receivers in the live stream, receive teaching content and engage in learning activities through the interactive interface. After processing the feedback data, they submit it to the teachers for reference.

[0103] A syllabus is a structured document containing different knowledge points, each with a description. Feedback data refers to the set of binary choices students make regarding their understanding of the teaching content during the learning process.

[0104] Student feedback data is a set of binary choices (such as understanding or not understanding), which reduces the complexity of data processing, but can efficiently reflect the overall learning situation through batch collection and real-time transmission.

[0105] In the specific implementation process, a lightweight data model is used to ensure that excessive latency is not introduced in high-concurrency live streaming scenarios, while providing a standardized data format for subsequent AI analysis.

[0106] Before the live broadcast begins, the teacher prepares a teaching syllabus based on the content of the lesson. And input it into the teaching platform. During the live broadcast, the teacher and students follow the teaching syllabus. Each knowledge point will be explained.

[0107] Ensuring the real-time nature, integrity, and structure of data provides reliable input for subsequent intelligent analysis, thereby supporting the dynamic response capability of the entire system.

[0108] S200: Analyze the knowledge points included in the teaching syllabus, analyze the live broadcast content, and locate the currently explained knowledge point. Calculate the difficulty index using the educational resource database, set the feedback frequency, and determine whether to trigger an alert based on student feedback. Specifically, this includes:

[0109] S201. AI technology is used to analyze the live broadcast content in real time and simultaneously output a text summary. The similarity between each knowledge point in the teaching syllabus and the text summary is calculated in real time, and the knowledge point with the highest similarity is selected as the current knowledge point to be explained. Specifically, this includes:

[0110] S2011. A speech recognition engine is used to convert the speech in the live broadcast content into text. At the same time, OCR technology is used to recognize the images in the live broadcast content and extract the text. Simultaneously, the text corresponding to the speech and images is merged into a text summary and continuously output.

[0111] In practice, the speech recognition engine and OCR technology are combined with deep learning models (such as ASR and visual text recognition) to handle noise and diverse content in live broadcasts and improve the accuracy of summarization.

[0112] S2012, Analysis of Teaching Syllabus For each knowledge point description, calculate the average word count of all knowledge point descriptions. Set the interval duration. It also analyzes the duration and word count of the continuous output of text summaries.

[0113] S2013, will continue to output for a duration greater than And the number of words is greater than The text summaries are packaged into data packets, and the text summaries and teaching syllabi in the data packets are calculated separately. The semantic similarity of the content descriptions of each knowledge point.

[0114] Data packets are packaged one by one, and packaging occurs when the packaging condition is met (continuous output duration is greater than 10 ... And the number of words is greater than At the instant of ( ), the text summary within that time period is immediately packaged into a data packet, and there is no duplication of text summaries in different data packets.

[0115] S2014. Data packets are arranged chronologically, and the knowledge point with the highest semantic similarity is selected and associated with each data packet. The most recently generated data packet is identified, and the knowledge point associated with that data packet is used as the current knowledge point to be explained. .

[0116] The design of data packet packaging conditions (duration and word count threshold) avoids frequent calculations and optimizes real-time performance. Semantic similarity calculation adopts models based on word embedding or BERT to ensure semantic consistency of knowledge point associations.

[0117] In practice, text summaries are generated through multimodal analysis (speech and image) and semantic similarity with the teaching syllabus is calculated to accurately locate the knowledge points being explained.

[0118] S202. Select resources from the educational resource database that include the teaching syllabus. The teaching records are used to obtain the recorded videos, identify the location and duration of the currently explained knowledge point, and then calculate the difficulty index. Specifically, this includes:

[0119] S2021, Obtain all documents containing the syllabus. The teaching records are analyzed, and the text summary output under the timeline of the recorded video in each teaching record is analyzed. Data packages are generated sequentially and associated with knowledge points, with all data packages arranged in chronological order.

[0120] S2022, Mark the currently explained knowledge point Analyze the associated data packets, identify the distribution of marked data packets in the data packet sequence corresponding to each teaching record, and substitute them into the formula to calculate the knowledge point being explained. Difficulty level:

[0121] ;

[0122] In the formula, To include the syllabus The total number of all teaching records For the first The number of all tagged data packets in each teaching record; and It is a constant, and .

[0123] parameter and ( The design reflects the weighting difference between labeled data packets (directly related to knowledge points) and unlabeled data packets (interval content), expanding the definition of difficulty by considering not only the explanation time of the knowledge point itself but also the impact of repeated explanations (such as the total duration and word count of unlabeled data packets), thus more comprehensively reflecting the complexity and depth of teaching. This design avoids the one-sidedness of relying solely on frequency, making the difficulty assessment more closely aligned with actual teaching scenarios.

[0124] and The first The first teaching record The duration of continuous output of each tagged data packet and the number of words in the text digest. and The first The total duration of continuous output of all data packets in the teaching record and the total number of words in the text summary.

[0125] and The first The first teaching record The total duration of continuous output of all untagged packets and the total number of words in the text digest between the first tagged packet and the next tagged packet.

[0126] The difficulty index is used to assess the teaching difficulty of a certain knowledge point. It is determined by analyzing the duration and text volume of data packets (such as explanation segments) directly related to the knowledge point in historical teaching records, as well as the overall impact of the content between these data packets.

[0127] In practice, the formula is combined with weighted calculations to quantify the complexity of knowledge points and the depth of teaching, reflecting the concentration and repetition of the explanation, thereby providing a basis for adjusting teaching strategies in the future.

[0128] Calculating the total duration of continuous output of all unlabeled data packets and the total number of words in the text summary between labeled data packets is to demonstrate the importance of repeatedly explained knowledge points. Simply analyzing the total output duration of labeled data packets and the total number of words in the text summary, however, cannot reflect the impact of repeated explanations.

[0129] By analyzing the distribution of knowledge points in historical teaching records, calculations were performed. An index quantifies the difficulty of teaching a knowledge point.

[0130] S203. Set the feedback frequency based on the difficulty index, and collect student feedback during the live stream according to the feedback frequency. Calculate the risk index based on the difficulty index and feedback results, and determine whether to trigger a risk warning based on the risk index. Specifically, this includes:

[0131] S2031, Computation Teaching Syllabus The average difficulty index of all knowledge points Preset reference frequency According to the formula: Set the feedback frequency after rounding the calculation. .

[0132] In the specific implementation process, feedback frequency Proportional to the difficulty index, the concept of adaptive sampling is extended: high-difficulty knowledge points require more frequent feedback collection to capture potential problems.

[0133] S2032, The teaching platform is based on feedback frequency. The questionnaire is sent to students at regular intervals, and feedback from each student is collected within a preset time period. The feedback includes both negative and positive feedback.

[0134] S2033. Analyze the proportion of negative feedback in each batch of feedback results, and calculate the risk index by substituting it into the formula along with the difficulty index. If the risk index is greater than the threshold, a risk warning is triggered. Risk Index The calculation formula is as follows:

[0135] ;

[0136] In the formula, A constant greater than 1 For the total batch of feedback results, For the first The percentage of negative feedback in the batch feedback results.

[0137] Risk Index The formula introduces logarithmic functions and constants. (>1) Smooths out abrupt changes in difficulty and feedback ratio, avoiding false alarms. The weighted average of negative feedback ratios ensures the robustness of the early warning system.

[0138] The risk index is used to determine whether there are risks in the teaching process and whether an early warning needs to be triggered.

[0139] In the specific implementation process, the difficulty of the current knowledge points is compared with the historical average difficulty, as well as the overall proportion of negative student feedback, and a logarithmic function is used to balance the sudden changes.

[0140] The formula's output helps identify teaching obstacles, ensuring that the early warning mechanism is both robust and adaptable to real-time changes, thereby supporting timely intervention.

[0141] A data-driven early warning mechanism is achieved by setting a dynamic feedback frequency and calculating a risk index.

[0142] It enables intelligent monitoring and adaptive adjustment of the teaching process, and timely identification of teaching difficulties and student comprehension obstacles through quantitative indicators (such as difficulty and risk), thereby helping teachers dynamically optimize teaching strategies and improve classroom efficiency and interaction quality.

[0143] S300: When an alert is triggered, analyze the teaching records in the educational resource database and draw knowledge point paths, dividing the explanation paths into those ending with the currently explained knowledge point. Calculate the matching index for each explanation path, and generate a reference document after weighted summarization. Specifically, this includes:

[0144] S301. When a risk warning is triggered, obtain the data packet sequence corresponding to each teaching record. Draw the knowledge point path for each teaching record according to the chronological order of the data packets and the associated knowledge points.

[0145] In the specific implementation process, the knowledge points in each knowledge point path are connected in a series and are allowed to appear repeatedly. The same knowledge point may appear again or multiple times after a gap of several knowledge points.

[0146] S302. Divide each knowledge point path into knowledge point categories. For each explanation path ending with a sample path generated from the live stream content, a matching index is calculated between the explanation path and the sample path. Specifically, this includes:

[0147] S3021, Analyzing Knowledge Point Paths Chinese knowledge points The distribution location of each knowledge point Establish end nodes. After merging all consecutive end nodes, calculate the knowledge point paths. Number of lower-end nodes .

[0148] S3022, Following the knowledge point path The initial knowledge point serves as the starting point, and each subsequent node serves as the endpoint, forming a knowledge point path. Divide into There are three explanation paths. And so on, explanation paths are divided for each knowledge point.

[0149] S3023. Based on the temporal changes of the currently explained knowledge points in the live broadcast content, draw sample paths. Substitute these paths into the formula to calculate the matching index between each explanation path and the sample paths. :

[0150] ;

[0151] After simplification, we get:

[0152] ;

[0153] In the formula, This represents the number of knowledge points in the sample path. This refers to the number of identical knowledge points between the explanation path and the sample path. To explain the path and sample path, the first The semantic similarity of text summaries of data packets corresponding to the same knowledge points.

[0154] To explain the path, the first The duration of continuous output for data packets corresponding to the same knowledge point. To explain the path and sample path, the first The maximum duration of continuous output for data packets corresponding to the same knowledge point.

[0155] Matching Index The formula takes into account semantic similarity. and the ratio of continuous output time The multi-dimensional assessment of extended path similarity not only focuses on content matching but also considers the level of detail in the explanation, making the reference path more in line with actual teaching.

[0156] In practice, semantic similarity is based on an NLP model to ensure content relevance. The time-based ratio reflects the level of detail in the teaching, ensuring that sample paths with longer durations for the same knowledge point achieve higher matching indices.

[0157] By mapping knowledge point paths and calculating matching indices, we can identify the historical paths most relevant to the current teaching.

[0158] S303. Using the explanation paths with a matching index greater than a threshold as reference paths, extract the text summaries of the data packets corresponding to the terminal nodes of all reference paths, assign weights according to the matching index, and summarize to generate reference texts. Specifically, this includes:

[0159] S3031. Obtain the matching index of each reference path and the text digest of the data packet corresponding to the end node. Normalize the matching indices of all reference paths using a weighted summation method to obtain the weight coefficients. The sum of the weight coefficients of all reference paths is 1.

[0160] S3032. Divide each text summary into entity units, calculate the cumulative score of the weight coefficient of each entity unit in all text summaries, and select entity units whose cumulative scores exceed a preset threshold to form a candidate summary set.

[0161] S3033. Use a text summarization model to remove semantically repetitive entity units from the candidate summary set, adjust the sentence arrangement according to the logical order and add conjunctions, and finally synthesize and generate a new text summary as a reference manuscript.

[0162] In practice, entity unit segmentation and deduplication techniques are based on keyword extraction or sequence models (such as TextRank or BERT), which expand the intelligence of text summarization: by accumulating weight coefficients, high-frequency and important content is retained first.

[0163] Semantic deduplication and logical adjustments ensure the coherence and readability of the text. The final generated content not only reduces redundancy but also conforms to teaching logic, thus improving the teaching support effect.

[0164] Through weighted summarization and text processing, a concise and coherent reference document is generated.

[0165] It provides personalized teaching assistance by matching historical paths to generate targeted reference content, helping teachers quickly adjust their teaching methods, reduce teaching burden, and enhance teaching adaptability and effectiveness.

[0166] S400: Push the reference text to the teacher's client for display, providing the teacher with reference text assistance during the live broadcast.

[0167] In the specific implementation process, low-latency communication protocols (such as WebSocket) are used to ensure that the documents are delivered in time, and UI design (such as pop-ups or sidebars) is combined to achieve non-intrusive prompts and avoid interfering with the teaching process.

[0168] Example 2: Please refer to Figure 2 The present invention also provides an education resource matching and management system based on intelligent perception, including an intelligent perception module, a dynamic analysis module, a resource matching module and a display management module.

[0169] The intelligent sensing module is used to collect educational resource data, as well as the teaching syllabus of the teacher and the feedback results of each student during the live broadcast.

[0170] In the specific implementation process, educational resource databases (such as teaching records containing recorded videos and teaching outlines) and activity data of teachers and students during live broadcasts are collected, and the teaching outlines and student feedback results are analyzed.

[0171] By collecting and analyzing data in real time, basic inputs are provided to ensure the integrity and timeliness of teaching data, support the accuracy of subsequent analysis, and thus lay the foundation for dynamic teaching adjustments.

[0172] The dynamic analysis module is used to parse the knowledge points contained in the teaching syllabus, analyze the live broadcast content, and locate the currently explained knowledge point. It calculates a difficulty index based on the educational resource database, sets the feedback frequency, and determines whether to trigger an alert based on the feedback results.

[0173] In the specific implementation process, AI technologies (such as speech recognition and OCR) are used to analyze the live broadcast content in real time, locate the knowledge point being explained, calculate the knowledge point difficulty index DIF based on the educational resource database, set the feedback frequency PL, and then use the risk index FX to determine whether to trigger an early warning.

[0174] It enables intelligent monitoring of the teaching process, automatically identifies teaching difficulties and students' comprehension levels, provides timely warnings of potential problems, and helps teachers optimize teaching strategies and improve classroom efficiency.

[0175] When an alert is triggered, the resource matching module analyzes the educational resource database and draws knowledge point paths, dividing the paths into explanation paths ending with the currently explained knowledge point. It calculates the matching index for each explanation path, weights and summarizes them, and then generates a reference document.

[0176] In the specific implementation process, when an early warning is triggered, the knowledge point paths in the historical teaching records are analyzed, the explanation paths ending with the current knowledge point are divided, the matching index PP of each path is calculated, and reference texts are generated through weighted summarization.

[0177] It provides personalized teaching support by matching historical paths to generate targeted reference content, thereby enhancing teaching adaptability and effectiveness and reducing teachers' teaching burden.

[0178] The display management module is used to push reference documents to the teacher's client for display.

[0179] The generated reference documents are pushed to the teacher's client for display. This enables real-time visualization of auxiliary information, helping teachers quickly access references during live broadcasts, improving teaching response speed and interaction quality, and ultimately optimizing the overall teaching experience.

[0180] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0181] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent perception-based education resource matching management method, characterized in that: The method comprises: S100, collecting an education resource library, and activity data of student objects and teacher objects in a live broadcast process; parsing a teaching outline of the teacher objects and feedback results of each student object in the activity data; S200, parsing knowledge points contained in the teaching outline, analyzing live broadcast content and positioning to a current explanation knowledge point; combining the education resource library to calculate a difficulty index, setting a feedback frequency, and judging whether to trigger a warning according to the feedback results of the student objects; S300, when the warning is triggered, analyzing teaching records in the education resource library and drawing a knowledge point path, dividing an explanation path ending with the current explanation knowledge point; calculating a matching index of each explanation path, and generating a reference script after weighted summarization; S400, pushing the reference script to a client of the teacher object for display, and providing the reference script auxiliary prompt of the current explanation knowledge point for the teacher object in the live broadcast process. 2.The smart sensing based education resource matching management method according to claim 1, characterized in that: In S100, the education resource library contains different teaching records, and each teaching record contains a recording and broadcasting video and a teaching outline; Live broadcast refers to a remote classroom teaching mode formed by using network communication technology to transmit teaching content of a teacher object and feedback data of student objects in real time; The teacher object is the initiator in the live broadcast, demonstrates teaching and presents teaching content through a media device, and dynamically adjusts a teaching strategy according to feedback data from the student objects; The student object is the receiver in the live broadcast, receives teaching content through an interactive interface and performs learning behavior, generates feedback data, and submits the feedback data to the teacher object for reference; The teaching outline refers to a structural file containing different knowledge points, and each knowledge point contains content description; the feedback data refers to a binary selection made by the student object on whether the student object understands the teaching content in the learning process; Before live broadcast is started, the teacher object produces a teaching outline according to the teaching content of this time and inputs the teaching platform; During the live broadcast process, the teacher object explains each knowledge point according to the teaching outline. ​ 3.The smart sensing based education resource matching management method of claim 2, wherein: S200 comprises: S201, adopt AI technology to analyze live content in real time, and output text summary synchronously; calculate the similarity between each knowledge point in the teaching outline and the text summary in real time, and take the knowledge point with the highest similarity as the current explanation knowledge point ; S202、In the educational resource library, the teaching records containing the teaching outline are screened out , the recorded videos in the teaching records are obtained, the position and time length of the current knowledge point being explained are identified to calculate the difficulty index ; S203, setting a feedback frequency according to the difficulty index, collecting feedback results of the student objects according to the feedback frequency in the live broadcast process; calculating a risk index according to the difficulty index and the feedback results, and judging whether to trigger a risk warning according to the risk index. 4.The smart sensing based education resource matching management method of claim 3, wherein: S201 comprises: S2011, converting voice in the live broadcast content into text by using a voice recognition engine, identifying images in the live broadcast content by using OCR technology, extracting text, merging text corresponding to the voice and the images at the same time into a text summary, and continuously outputting the text summary; S2012、analyze the syllabus the average number of words of all content descriptions of the knowledge points ; set the interval duration and analyze the text summary continuous output duration and the number of words; S2013、the text summary with the duration of continuous output greater than and the number of words greater than is packaged as a data packet, and the semantic similarity between the text summary of the data packet and the content description of each knowledge point in the syllabus is calculated respectively. S2014, the data packets are arranged in time sequence, and each data packet selects a knowledge point with the highest semantic similarity for association; a newly generated data packet is identified, and a knowledge point associated with the data packet is taken as a current explanation knowledge point . 5.The smart sensing based education resource matching management method of claim 4, wherein: S202 comprises: S2021、Get all the teaching records containing the syllabus , analyze the text summary output situation of the timeline of the recorded video in each teaching record; generate data packets in turn and associate knowledge points, all data packets are arranged in chronological order; S2022、Mark the current explanation knowledge point The data packet associated with each teaching record is analyzed to determine the distribution of the marked data packet in the data packet sequence, and the difficulty index of the current explanation knowledge point is calculated by substituting the formula: ​ ； wherein is the number of all teaching records including the syllabus , is the number of all marking data packets in the th teaching record; and are constants, and ; and are the duration of the continuous output of the first teaching record and the number of text abstract words of the first marked data packet, respectively; and are the total duration of the continuous output of all data packets in the first teaching record and the total number of text abstract words, respectively; and are the first teaching record, the total duration of the continuous output of all non-marker data packets and the total number of words of the text summary between the first marker data packet and the next marker data packet. 6.The smart sensing based education resource matching management method of claim 5, wherein: S203 comprises: S2031、calculating the syllabus the average value of the difficulty index of all knowledge points , the preset reference frequency , according to the formula: the set feedback frequency after calculation and rounding ; S2032、the teaching platform pushes the questionnaire to the student object according to the feedback frequency The questionnaire is pushed to the student object at a fixed time, and the feedback results of each student object filling in the questionnaire are collected within a preset time length, the feedback results including negative feedback and positive feedback. S2033、analyze the proportion of negative feedback in each batch of feedback results, combine the difficulty index into the formula to calculate the risk index, and trigger risk warning if the risk index is greater than the threshold value; the risk index The calculation formula is as follows: ； In the formula, is a constant greater than 1, is the total batch of feedback results, is the number of negative feedback in the first batch of feedback results. 7.The smart sensing based education resource matching management method of claim 6, wherein: S300 comprises: S301, when the risk warning is triggered, obtaining a data packet sequence corresponding to each teaching record; drawing a knowledge point path for each teaching record according to a time sequence of the data packets and associated knowledge points; S302, respectively, each knowledge point path is divided into an explanation path ending with a knowledge point S302, respectively, each knowledge point path is divided into an explanation path ending with a knowledge point S302, respectively, each knowledge point path is divided into an explanation path ending with a knowledge point S303, taking an explanation path with a matching index greater than a threshold value as a reference path, extracting a text summary of a data packet corresponding to a terminal node of all reference paths, setting a weight according to the matching index, and generating a reference script after summarization. 8.The smart sensing based education resource matching management method of claim 7, wherein: S302 comprises: S3021、analyze knowledge point path The distribution position of the knowledge point , establish an end node for each knowledge point , and after merging all end nodes with continuous positions, count the number of knowledge point paths under the end node ; S3022、According to the knowledge point path The beginning knowledge point is the starting point, and each end node is the terminal point. The knowledge point path is divided into a lecture path; similarly, a lecture path is divided for each knowledge point path;​ S3023、According to the time sequence change of the current explanation knowledge point in the live broadcast content, a sample path is drawn; the matching index between each explanation path and the sample path is calculated respectively by substituting the formula : ； In the formula, is the number of knowledge points in the sample path, is the number of same knowledge points between the explanation path and the sample path; is the semantic similarity of the text summary of the data packet corresponding to the same knowledge point in the explanation path and the sample path. the maximum value of the duration of the continuous output of the data packet corresponding to the same knowledge point in the lecture path and the sample path the maximum value of the duration of the continuous output of the data packet corresponding to the same knowledge point in the lecture path and the sample path the maximum value of the duration of the continuous output of the data packet corresponding to the same knowledge point in the lecture path and the sample path the maximum value of the duration of the continuous output of the data packet corresponding to the same knowledge point in the lecture path and the sample path 9.The smart sensing based education resource matching management method of claim 7, wherein: S303 comprises: S3031, respectively obtaining a matching index of each reference path and a text summary of a data packet corresponding to a terminal node; performing normalization processing on the matching indexes of all reference paths by using a weighted summation method to obtain a weight coefficient; S3032, divide each text summary into entity units, respectively calculate the weight coefficient accumulation score of each entity unit in all text summaries, select the entity units with accumulation score exceeding the preset threshold, and constitute a candidate summary set; S3033, remove the entity units with semantic repetition in the candidate summary set by using the text summary model, adjust the sentence arrangement according to the logical order and add conjunctions, and finally generate a new text summary as a reference script.

10. An intelligent perception-based education resource matching management system, characterized in that: The system comprises an intelligent sensing module, a dynamic analysis module, a resource matching module and a display management module; The intelligent sensing module is used for collecting the education resource library, the teaching outline of the teacher object and the feedback result of each student object in the live broadcast process; The dynamic analysis module is used for analyzing the knowledge points contained in the teaching outline, analyzing the live broadcast content and positioning to the current explanation knowledge point; combining the education resource library to calculate the difficulty index, setting the feedback frequency and judging whether to trigger the early warning according to the feedback result; When the early warning is triggered, the resource matching module analyzes the education resource library and draws the knowledge point path, and divides the explanation path ending with the current explanation knowledge point; Calculate the matching index of each explanation path, and generate a reference script after weighted summary; The display management module is used for pushing the reference script to the client of the teacher object for display.

Images