Creation method and system for multifunctional AI teaching platform

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a creation method and system for a multifunctional AI teaching platform. The method comprises the following steps: acquiring teaching course data, and constructing a course knowledge graph according to the teaching course data so as to obtain course knowledge graph data; performing curriculum homework feature extraction according to the teaching curriculum data to obtain curriculum homework data; according to the curriculum homework data, carrying out paper incomplete degree detection so as to obtain paper incomplete degree data; performing curriculum homework interaction feature extraction and curriculum homework error feature extraction according to the curriculum homework data so as to obtain curriculum homework interaction data and curriculum homework error data; and performing curriculum homework difficulty analysis according to the curriculum homework interaction data and the curriculum homework error data, thereby obtaining curriculum homework difficulty The intelligent and personalized service capability of the teaching platform is improved based on the artificial intelligence technology.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and particularly relates to a method and system for creating a multi-functional AI teaching platform. Background Art

[0002] In existing methods, the quality assessment of course assignments, especially the detection of paper mutilation, relies on traditional image processing algorithms. Although image processing technology has made certain progress, the detection accuracy for mutilated areas of paper is still not high. Especially when the quality of the paper image is poor or the content of the assignment is complex, traditional image processing methods are difficult to achieve accurate mutilation assessment, resulting in the recommendation system of the teaching platform being unable to accurately reflect the true quality of students' assignments. This assessment error has a negative impact on personalized recommendation and teaching resource allocation, reducing the teaching effect of the system. When constructing a course knowledge graph and personalized teaching resource recommendation in existing systems, although relying on preset algorithms and rules, there is a lack of an effective dynamic update mechanism. In practical applications, as the learning situation of students changes continuously, the original teaching resources and course content need to be adjusted and optimized in a timely manner. However, traditional methods often cannot achieve real-time data update, and the update process is cumbersome and has poor timeliness. This leads to the lag of personalized recommendation and course content update on the teaching platform, being unable to effectively respond to changes in students' needs, and reducing the flexibility and adaptability of the teaching system. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a method and system for creating a multi-functional AI teaching platform to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for creating a multi-functional AI teaching platform includes the following steps: Step S1: Obtain teaching course data, and construct a course knowledge graph based on the teaching course data to obtain course knowledge graph data; Step S2: Extract course assignment features according to the teaching course data to obtain course assignment data; detect the paper mutilation degree according to the course assignment data to obtain paper mutilation degree data; Step S3: Extract course assignment interaction features and course assignment error features according to the course assignment data to obtain course assignment interaction data and course assignment error data; analyze the course assignment difficulty according to the course assignment interaction data and course assignment error data to obtain course assignment difficulty data; Step S4: Recommend personalized teaching resources according to the course assignment difficulty data and the paper mutilation degree data to obtain personalized teaching resource data; Step S5: Update the curriculum knowledge graph data based on the teaching personalized resource data to obtain the updated curriculum knowledge graph data, and create a teaching platform based on the updated curriculum knowledge graph data to obtain the teaching platform data.

[0005] By obtaining teaching curriculum data and constructing a curriculum knowledge graph, the present invention can systematically organize and associate curriculum knowledge points, provide an accurate knowledge structure framework for the teaching platform, and help the platform better understand the curriculum content and students' needs. The construction of the curriculum knowledge graph enables the platform to more efficiently perform knowledge association, query, and push, further enhancing the intelligent allocation of teaching resources and personalized services. By extracting the characteristics of curriculum assignments and detecting the degree of paper mutilation, the integrity of the assignments can be accurately identified. In particular, by improving the accuracy of paper mutilation detection, the evaluation errors caused by poor image quality can be reduced, the reliability of assignment evaluation can be improved, and thus the accuracy of the personalized resource recommendation system can be optimized. During the extraction of curriculum assignment interaction characteristics and error characteristics, the performance and error types of students in their assignments can be effectively captured. By combining interaction data and error data for curriculum assignment difficulty analysis, it helps to identify the difficulties and knowledge blind spots encountered by students during the learning process, thereby supporting the platform to provide more personalized learning suggestions and resources for students. By combining the difficulty data of curriculum assignments with the paper mutilation data for personalized resource recommendation, the platform can push the most suitable learning materials for each student, optimize the learning path, help students overcome the difficulties in learning, and improve the learning effect. Finally, by dynamically updating the curriculum knowledge graph according to the teaching personalized resource data, not only can the real-time update of the platform content and resources be ensured, but also the recommendation system can always follow the changes in students' learning progress and needs, thereby making the platform more flexible and adaptable and enhancing the intelligent level of the teaching platform.

[0006] Optionally, step S1 is specifically as follows: Step S11: Obtain teaching curriculum data and perform knowledge point annotation based on the teaching curriculum data to obtain knowledge point data; Step S12: Perform entity recognition based on the knowledge point data to obtain knowledge point entity data; Step S13: Extract relationships from the knowledge point entity data to obtain knowledge point entity relationship data; Step S14: Construct a curriculum knowledge graph based on the knowledge point entity data and the knowledge point entity relationship data to obtain curriculum knowledge graph data.

[0007] By obtaining teaching course data and performing knowledge point annotation, the present invention effectively assigns a structured identifier to the course content, enabling the knowledge points to be clearly identified and tracked, and thus providing a solid foundation for the subsequent construction of the knowledge graph. By performing entity recognition on the knowledge points, the key teaching elements can be accurately extracted, helping the teaching platform understand the core content of the course and the learning focus of students. Then, by extracting the relationships between the knowledge point entities, the internal connections between the knowledge points are further revealed, enabling the platform to identify and present the logical relationships between the knowledge points, which is crucial for understanding the systematic structure of knowledge. On this basis, the construction of the course knowledge graph combines these knowledge points and relationship data to form a complete course knowledge network, enabling the teaching platform to more intelligently identify the learning progress and knowledge mastery of students, thereby providing a basis for personalized learning recommendations. This structured knowledge graph can not only help the platform update and optimize teaching resources in real time, but also dynamically adjust the course content according to the personalized needs of students, thus enhancing the flexibility and adaptability of the platform. By introducing a dynamic update mechanism, the platform can respond to the changes generated during the learning process of students at any time, ensuring the accurate matching of the recommendation system and the course content, greatly enhancing the intelligent level of the teaching system, and ultimately providing students with a more efficient and accurate learning experience.

[0008] Optionally, step S13 is specifically as follows: Step S131: Format the knowledge point entity data to obtain the formatted knowledge point entity data; Step S132: Extract the syntactic feature according to the knowledge point entity relationship data to obtain the knowledge point entity relationship syntactic feature data; Step S133: Construct a knowledge point entity relationship recognition model according to the knowledge point entity relationship syntactic feature data to obtain the knowledge point entity relationship recognition model; Step S134: Extract entity pairs from the formatted knowledge point entity data according to the knowledge point entity relationship recognition model to obtain entity pair data; Step S135: Judge the relationship of the obtained entity pair data according to the knowledge point entity relationship syntactic feature data to obtain the knowledge point entity relationship data.

[0009] The present invention can ensure that the knowledge point entity information has consistency and structure by formatting the knowledge point entity data, so that the subsequent data processing is more efficient and standardized. This formatting process provides a stable basis for further analyzing and processing the knowledge point entity relationship, so that various types of information can be extracted and applied more accurately. Then, by extracting grammatical features from the knowledge point entity relationship data, the grammatical structure of key information can be identified, and more abundant and in-depth grammatical feature data are provided for subsequent relationship analysis, which helps to improve the accuracy of relationship extraction. By constructing a knowledge point entity relationship recognition model, the platform can automatically identify the complex relationship between knowledge points through a trained model, and further enhance the intelligent level of the platform when processing and understanding the entity relationship in the knowledge graph. This model can automatically extract entity pairs according to the previously extracted features, thereby realizing efficient identification of the specific relationship between knowledge points, and providing accurate data support for subsequent relationship judgment and knowledge graph update. When performing relationship judgment on entity pair data, the platform can comprehensively analyze grammatical features and entity relationship features, thereby deriving more accurate entity relationship data, effectively optimizing the relationship structure in the knowledge graph, and enhancing the response speed and adaptability of the knowledge graph to the personalized learning path of students. The entire process not only optimizes the construction quality of the knowledge graph by gradually refining, identifying and judging the entity relationships of knowledge points, but also can update and adjust the teaching content in real time to meet dynamically changing teaching needs, greatly improving the flexibility, intelligence and personalization of the teaching platform.

[0010] Optionally, step S2 specifically includes: Step S21: extracting course assignment features according to the teaching course data, thereby obtaining course assignment data; Step S22: collecting paper images of the course assignment data, thereby obtaining a course assignment paper image; Step S23: performing grayscale conversion on the homework paper image to obtain a grayscale image of the homework paper; Step S24: performing paper contour edge detection according to the course homework paper grayscale image, thereby obtaining paper contour edge data; Step S25: performing defective area recognition on the paper outline edge data, thereby obtaining paper defective area data; Step S26: Calculate the degree of damage of the grayscale image of the homework paper according to the paper damage area data, so as to obtain paper damage degree data.

[0011] By extracting features from the course assignment data, the present invention can effectively identify key feature information from the original data, laying a foundation for subsequent processing. By collecting the images of course assignment papers, the visual data of the assignments can be obtained, providing intuitive input data for subsequent image processing and analysis. Converting the images of course assignment papers to grayscale can simplify the complexity of the images, thereby reducing the consumption of computing resources and improving the efficiency and accuracy of subsequent image analysis processes. By detecting the edges of the paper contours in the grayscale images, the external shape of the paper can be accurately extracted, helping to identify the boundaries and integrity of the assignments, and thus providing important information for the analysis of the degree of mutilation. Identifying the mutilated areas from the edge data of the paper contours helps to accurately locate and identify the missing areas in the paper, providing efficient and accurate input data for subsequent calculation of the degree of mutilation. By calculating the degree of mutilation of the identified mutilated areas of the paper, specific degree-of-mutilation values can be obtained. This data can be used to evaluate the integrity of the assignments, thus providing more accurate information for the personalized recommendation system of the teaching platform, helping the system to make accurate recommendations based on the quality of the assignments, and optimizing the allocation of teaching resources and the design of personalized learning paths. This series of steps provides more accurate and reliable data support for the quality assessment of course assignments by gradually extracting and analyzing the relevant information of the paper images, thereby enhancing the intelligence, adaptability, and real-time update capabilities of the teaching platform, enabling the platform to respond more effectively to the changing needs of students and improving the teaching effect.

[0012] Optionally, step S25 is specifically as follows: Step S251: Calculate the distances between adjacent contour points of the edge data of the paper contours to obtain adjacent contour point distance data; Step S252: Obtain the safety threshold for the distances between adjacent contour points; Step S253: Detect paper breaks based on the safety threshold for the distances between adjacent contour points for the adjacent contour point distance data to obtain paper break data; Step S254: Classify according to the paper break data to obtain paper tear data and paper corner missing data; Step S255: Locate the paper tear areas in the grayscale image of the course assignment paper according to the paper tear data to obtain paper tear area data; Step S256: Locate the paper corner missing areas in the grayscale image of the course assignment paper according to the paper corner missing data to obtain paper corner missing area data; Step S257: Merge the paper mutilated areas according to the paper tear area data and the paper corner missing area data to obtain paper mutilated area data.

[0013] The present invention can provide accurate spatial data support for subsequent paper break detection by calculating the distance between adjacent contour points. This step can effectively identify the structural discontinuity area in the paper and provide basic data for determining whether the paper is broken. Obtaining the safety threshold of the distance between adjacent contour points provides a standard value for break detection, ensuring the accuracy and consistency of the detection, thereby avoiding the occurrence of false alarms or missed alarms and improving the reliability of the entire detection process. Performing break detection on the data of the distance between adjacent contour points helps to accurately identify whether the paper is broken, and marks it as break data, providing detailed break information for subsequent analysis steps. By classifying the paper break data into types, two different types of damage, paper tearing and corner missing, can be further distinguished, which provides a basis for targeted positioning and repair of damaged areas. According to the paper tear data, the tear area of ​​the course homework paper image is located, and the specific location of the paper tear can be accurately marked, thereby providing more refined damage assessment data for the teaching platform. Similarly, according to the paper corner missing data, the corner missing area of ​​the course homework paper image is located, which helps to accurately identify the location of the corner missing on the paper, ensuring that the teaching platform can perform personalized processing for different types of damage. Finally, the paper torn area data and the paper corner chipped area data are merged to form complete paper defect area data, which provides comprehensive information support for subsequent defect degree calculation and quality assessment, greatly improves the accuracy of defect degree detection and the accuracy of the personalized recommendation system of the teaching platform, ensures the rational allocation of teaching resources, and optimizes the teaching platform's ability to respond to changes in student needs.

[0014] Optionally, step S3 specifically includes: Step S31: extracting course homework interaction features and course homework error features according to the course homework data, thereby obtaining course homework interaction data and course homework error data; Step S32: Performing subject depth analysis on the course assignment interaction data to obtain subject depth data; Step S33: performing error type complexity analysis on the course assignment error data, thereby obtaining error type complexity data; Step S34: Evaluate the difficulty of the course assignment based on the error type complexity data and the subject depth data, thereby obtaining the course assignment difficulty data.

[0015] By extracting interaction features and error features from course assignment data, the present invention can deeply understand the behaviors and error types demonstrated by students in assignments, thereby obtaining more accurate course assignment interaction data and error data. These data provide rich basic information for subsequent analysis, helping to identify students' learning habits, understanding biases, and common error types. On this basis, in-depth subject analysis of the course assignment interaction data can reveal students' mastery levels in specific subject areas, obtain subject-depth data, thereby helping to evaluate students' understanding levels of course content, and further optimizing the allocation of teaching resources. By performing complexity analysis of error types on the course assignment error data, the complexity of students' mistakes in assignments can be evaluated, thereby obtaining error type complexity data, which helps to judge the types of difficulties and learning bottlenecks faced by students. Combining the error type complexity data and the subject-depth data for course assignment difficulty assessment can accurately evaluate the difficulty level of course assignments, provide dynamically updated difficulty indicators for the teaching platform, enabling the system to adjust personalized recommendations and teaching strategies according to students' actual abilities. Such an assessment mechanism effectively improves the accuracy of personalized recommendations on the teaching platform, can respond in real time to changes in students' needs, ensure that the teaching content is consistent with students' learning progress, thereby enhancing the adaptability and flexibility of the teaching system and improving the overall teaching effect.

[0016] Optionally, step S32 is specifically as follows: Step S321: Conduct a statistical analysis of the interaction frequency of the course assignment interaction data to obtain high-frequency course assignment interaction data; Step S322: Extract high-level interaction features based on the high-frequency course assignment interaction data to obtain high-level interaction data; Step S323: Conduct in-depth subject reasoning analysis on the high-level interaction data to obtain subject reasoning depth data; Step S324: Obtain students' course assignment answer data; Step S325: Conduct a logical analysis of the course assignment answers based on the students' course assignment answer data to obtain course assignment answer logic data; Step S326: Conduct subject-depth division of the subject reasoning depth data based on the course assignment answer logic data to obtain subject-depth data.

[0017] By statistically analyzing the interaction frequency of course assignment interaction data, the present invention can identify the course content with a high interaction frequency in the assignment by students, and obtain high-frequency course assignment interaction data. This data can help the system identify the focus points of students on certain knowledge points or tasks, providing valuable clues for subsequent feature extraction. Based on these high-frequency interaction data, advanced-level interaction feature extraction can be carried out to discover the deep interaction patterns of students on specific tasks or knowledge points, obtaining more detailed advanced-level interaction data, which helps to more accurately analyze the learning behaviors and needs of students. Then, in-depth analysis of subject reasoning on these advanced-level interaction data helps to reveal the depth of thinking of students in the process of subject reasoning, thereby obtaining subject reasoning depth data. This analysis can reflect the breadth and depth of thinking of students when solving problems, providing a more comprehensive perspective for evaluating their learning ability. At the same time, obtaining students' course assignment answering data and conducting logical analysis of their course assignment answers can reveal the thinking patterns and decision-making processes of students when answering questions, helping to further understand their thinking patterns and error patterns. Combining the logical data of course assignment answers, subject depth division can be carried out on the subject reasoning depth data, which can better associate the mastery of subject knowledge with the reasoning depth of students, thereby obtaining more detailed subject depth data. These data provide a more accurate basis for personalized recommendation for the system, enabling the teaching platform to dynamically adjust and optimize according to the actual understanding and thinking depth of students, thereby improving the accuracy of personalized teaching resource allocation, enhancing the response ability and adaptability of the teaching system, and thus optimizing the teaching effect.

[0018] Optionally, step S335 is specifically as follows: Mark the answering process according to the students' course assignment answering data, so as to obtain answering process data; Detect answering skips for the answering process data, so as to obtain answering skip data; Obtain the answering reasoning path data of the students' course assignments; Identify abnormal reasoning for the answering process data according to the answering reasoning path data of the students' course assignments, so as to obtain abnormal reasoning data; Integrate the course assignment answering logic according to the answering skip data and the abnormal reasoning data, so as to obtain the course assignment answering logic data.

[0019] By marking the answering process based on the answering data of students' course assignments, the present invention can track and record students' answering behaviors, form complete answering process data, and provide a basis for analyzing students' learning paths and behaviors. By detecting answering skips in the answering process data, it is possible to identify whether students skip steps or certain important steps during the answering process, thereby obtaining answering skip data. These data are very important for understanding whether there are thinking breaks or lack of execution of necessary steps in students' answering process. Obtaining the reasoning path data of students' course assignment answers further reveals the reasoning process when students solve problems, and helps to analyze how students transition from one step to another during the problem-solving process. Based on these reasoning path data, abnormal reasoning identification is performed on the answering process data, and unreasonable or atypical reasoning steps in students' reasoning process can be identified, thereby obtaining abnormal reasoning data. These abnormal reasoning data can help identify students' misunderstandings or wrong thinking routes when solving problems, and provide support for accurately assessing their learning disabilities. Finally, by comprehensively analyzing the answering skip data and abnormal reasoning data, complete course assignment answering logic data can be formed, thereby revealing students' thinking patterns and learning rules during the entire answering process. This comprehensive data provides a clear learning trajectory map for the teaching platform, can more accurately respond to students' needs in personalized recommendation and teaching resource allocation, avoids the untimely update of personalized teaching services caused by the lag of traditional evaluation methods, and thus effectively improves the adaptability and flexibility of the teaching system.

[0020] Optionally, step S33 is specifically as follows: Step S331: Classify the error types of the course assignment error data to obtain the basic course assignment error data and the applied course assignment error data; Step S332: Statistically analyze the component usage errors based on the basic course assignment error data to obtain the component usage error data; Step S333: Statistically analyze the circuit debugging errors based on the applied course assignment error data to obtain the circuit debugging error data; Step S334: Calculate the error complexity based on the component usage error data and the circuit debugging error data to obtain the error type complexity data.

[0021] By classifying the error data of course assignments into different error types, the present invention can divide the errors in students' assignments into basic errors and application errors, which helps to accurately identify the gaps in students' understanding and application of basic knowledge. The basic error data and application error data provide a clear distinction for in-depth analysis of students' weak learning points and contribute to the precise recommendation of personalized teaching resources. According to the basic error data of course assignments, the statistics of component usage errors can identify common problems in students' component usage, and then provide targeted guidance for students to help them master the correct method of component usage. Similarly, the statistics of circuit debugging errors in application error data can reveal the problems encountered by students during circuit debugging and provide data support for improving students' practical operation ability. By calculating the error complexity based on the component usage error data and circuit debugging error data, the difficulty and complexity of the errors made by students can be quantified, thereby evaluating the actual challenge level of students in their assignments and providing an accurate basis for difficulty adjustment for the teaching platform. This series of steps not only helps to evaluate students' error types and ability levels in more detail, but also monitors the specific problems encountered by students in actual assignments in real time, providing strong data support for dynamically updating teaching resources and personalized recommendation systems, thus enhancing the flexibility and adaptability of the teaching system.

[0022] Optionally, this specification also provides a creation system for a multi-functional AI teaching platform, which is used to execute the creation method for a multi-functional AI teaching platform as described above. The creation system for a multi-functional AI teaching platform includes: A course knowledge graph construction module, which is used to obtain teaching course data and construct a course knowledge graph based on the teaching course data, so as to obtain course knowledge graph data; A paper mutilation degree detection module, which is used to extract the characteristics of course assignments based on the teaching course data, so as to obtain course assignment data; and detect the paper mutilation degree according to the course assignment data, so as to obtain paper mutilation degree data; A course assignment difficulty analysis module, which is used to extract the interactive characteristics and error characteristics of course assignments based on the course assignment data, so as to obtain course assignment interactive data and course assignment error data; and analyze the difficulty of course assignments according to the course assignment interactive data and course assignment error data, so as to obtain course assignment difficulty data; A teaching personalized resource recommendation module, which is used to recommend teaching personalized resources according to the course assignment difficulty data and the paper mutilation degree data, so as to obtain teaching personalized resource data; A teaching platform creation module, which is used to update the course knowledge graph data according to the teaching personalized resource data, so as to obtain updated course knowledge graph data, and create a teaching platform according to the updated course knowledge graph data, so as to obtain teaching platform data.

[0023] The creation system for the multi-functional AI teaching platform of the present invention can implement any creation method for the multi-functional AI teaching platform of the present invention, which is a medium for coordinating operations and signal transmission between various modules to complete the creation method for the multi-functional AI teaching platform. The internal modules of the system cooperate with each other, thereby enhancing the intelligence and personalized service capabilities of the teaching platform. Brief Description of the Drawings

[0024] Other features, objects, and advantages of the present invention will become more apparent by reading the detailed description of the non-limiting embodiments with reference to the following drawings: Figure 1 It is a schematic flowchart of the steps of the creation method for the multi-functional AI teaching platform of the present invention; Figure 2 It is a detailed schematic flowchart of step S1 in the present invention; Figure 3 It is a detailed schematic flowchart of step S13 in the present invention.

[0025] The implementation, functional characteristics, and advantages of the object of the present invention will be further described with reference to the embodiments and the drawings. Detailed Embodiments

[0026] The technical method of the present invention patent will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0027] In addition, the drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0028] It should be understood that although terms such as "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0029] To achieve the above object, please refer to Figures 1 to 3 , the present invention provides a method for creating a multi-functional AI teaching platform, and the method includes the following steps: Step S1: Obtain teaching course data, and construct a course knowledge graph according to the teaching course data, so as to obtain course knowledge graph data; In this embodiment, it is necessary to obtain complete teaching course data from a teaching platform or a database, including course content, syllabus, learning objectives, textbooks and teaching materials, etc. After the data is obtained, it is processed to label each knowledge point in the course. Specifically, first, the content of each textbook is analyzed, and each teaching module and the knowledge points it contains are identified according to the course syllabus. Each knowledge point should be specifically labeled for subsequent knowledge point entity recognition and relationship extraction. During the labeling process, the parameter used is "knowledge point category", and this parameter needs to be manually specified according to the course content and the instructions of the teacher. Next, each labeled knowledge point is further processed using regular expression-based entity recognition technology to identify the core entities in the knowledge point. The entities of each knowledge point need to include information such as nouns and terms, and these entities form the basic data of the knowledge graph. Then, according to the context relationship between knowledge points, such as conjunctions and transition words, relationship extraction is performed, and through dependency analysis methods, a knowledge point entity relationship network is constructed, and finally course knowledge graph data is formed as the input for subsequent steps.

[0030] Step S2: Extract course assignment features according to the teaching course data, so as to obtain course assignment data; detect the degree of paper mutilation according to the course assignment data, so as to obtain paper mutilation data; In this embodiment, feature data related to homework is extracted from the teaching course data, specifically including homework type, topic category, homework difficulty, homework submission time, etc. These features can be obtained by text analysis of homework information in the teaching course data. For the image data of the homework, first obtain the image of the homework paper, which can be collected at high resolution by a scanner or camera device, and the resolution should be not less than 300 DPI to ensure that the image is clear. After the image is collected, grayscale conversion is performed to convert the RGB image into a grayscale image. During the processing, a fixed grayscale threshold (for example, a threshold of 128) is used to convert the pixel value of the image into a black and white image. Next, edge detection is performed on the grayscale image, using the Canny edge detection algorithm, and the high and low thresholds of the edge detection are set to 50 and 150 respectively. In the edge detection result, the edge area of ​​the image is analyzed using the contour extraction algorithm to identify the outer contour of the paper. The identified contour is further subjected to regional analysis to extract the incomplete area of ​​the paper. The identification standard of the incomplete area is based on the adjacent distance of the contour points (for example, contour points with a distance less than 5mm are regarded as the same area), combined with the calculated incomplete area data, to generate paper incompleteness data.

[0031] Step S3: extracting interactive features of course assignments and extracting error features of course assignments based on the course assignment data, thereby obtaining interactive data of course assignments and error data of course assignments; performing course assignment difficulty analysis based on the interactive data of course assignments and error data of course assignments, thereby obtaining course assignment difficulty data; In this embodiment, feature extraction is performed on coursework data, which is specifically divided into two parts: interactive features and error features. Interactive feature extraction is achieved by analyzing students' answering behavior and feedback information in homework. The data of each homework includes information such as the time when students answer, whether the answer is correct or not, and the selected answer. These data are used to calculate the average answering time and accuracy rate of each question. Questions with answering time exceeding a predetermined threshold (for example, questions exceeding 30 seconds) are considered to be more difficult questions. Error feature extraction is classified by analyzing students' wrong answers, especially the types of errors (such as calculation errors, misunderstandings, etc.). Each type of error will correspond to different error data, and the error classification standard is set according to the teacher's teaching design. The error classification standard will be based on the type of wrong question, such as multiple-choice questions are classified as "wrong options", and fill-in-the-blank questions are classified as "incomplete answers", and each classification has clear parameter boundaries. Afterwards, the interactive data and error data are used to perform difficulty analysis, and the difficulty threshold is set (such as answering time>40 seconds, error rate>40% for high-difficulty questions) combined with the interactive data (answering time and accuracy) of each question and the error features, and the overall difficulty data of the homework is calculated by weighted average method.

[0032] Step S4: Based on the course assignment difficulty data and the paper mutilation degree data, perform personalized teaching resource recommendation to obtain personalized teaching resource data; In this embodiment, the core of personalized teaching resource recommendation is to combine the assignment difficulty data and the paper mutilation degree data, and analyze and match suitable learning resources by setting thresholds. First, the range of difficulty data is obtained through Step S3. The difficulty data of the assignment is normalized to a value between 0 and 1, where the difficulty value 0 represents the easiest assignment and 1 represents the most difficult assignment. At the same time, the range of the paper mutilation degree data is from 0 to 100, where 0 means the paper is intact and 100 means it is completely missing. Combining these two data, the personalized needs of each student can be matched. If a student has a high assignment difficulty value and a large paper mutilation degree (for example, difficulty value > 0.7, mutilation degree > 50), the resources recommended by the system should focus on auxiliary materials with stronger basicity and higher interpretability, such as basic concept explanations, wrong question sets, etc. The criteria for personalized resource recommendation can be set through a rule engine. For example, set the weighted thresholds of the difficulty value and the mutilation degree, such as difficulty × 0.6 + mutilation degree × 0.4. If the result value is greater than 0.7, recommend teaching resources of advanced difficulty.

[0033] Step S5: Update the course knowledge graph data according to the personalized teaching resource data to obtain updated course knowledge graph data, and create a teaching platform based on the updated course knowledge graph data to obtain teaching platform data.

[0034] In this embodiment, the course knowledge graph is updated according to the personalized resource recommendation result. The update process is to identify the association relationship between the knowledge points in the course and the resources by analyzing the personalized resource data. For example, if a certain type of specific teaching video or reference document is included in the recommended resources for a knowledge point, the knowledge point label of this resource will be added to the course knowledge graph. When updating the knowledge graph, a certain relationship connection threshold needs to be set (such as when the association degree > 0.5, it is considered a substantial update) to ensure the accuracy of the graph. The updated knowledge graph will provide new data support for the teaching platform, ensuring that the platform can dynamically adjust the course content according to the latest knowledge points and resources. Finally, a teaching platform is created based on the updated course knowledge graph. The design of the platform depends on the structured storage of the graph data, and each module in the platform is scheduled and recommended according to the relationship between the knowledge points and resources in the graph data.

[0035] Optionally, Step S1 is specifically: Step S11: Obtain teaching course data, and perform knowledge point annotation according to the teaching course data to obtain knowledge point data; In this embodiment, course data is extracted from a teaching platform or a data source. This data includes course names, syllabi, textbook contents, courseware, lecture notes, and after-class exercises, etc. After the course data is collected, natural language processing (NLP) techniques are used for text preprocessing, including word segmentation, stop word removal, and part-of-speech tagging, etc. Through the word segmentation process of the course content, key terms and core concepts are determined, and these terms are associated with specific knowledge points based on the course syllabus. Each knowledge point should be identified through manual annotation or rule matching. The specific annotation includes course chapters, titles of knowledge points, and related sub-knowledge points. During the annotation process, a term threshold is set. For example, if the term appears more than 3 times, it is considered part of the course knowledge point. During the annotation process, the annotation tool can be carried out by combining regular expressions and dictionary matching. For example, the "welding process" in the course is marked as a knowledge point, which involves three sub-knowledge points: "welding mechanism", "heat input", and "welding defects". Finally, the knowledge point data obtained through this step should include the title, description, and identifier of each knowledge point.

[0036] Step S12: Perform entity recognition based on the knowledge point data to obtain knowledge point entity data; In this embodiment, based on the knowledge point data, entity recognition is performed on the knowledge point text. The goal of entity recognition is to extract specific entity information such as terms, equipment, and processes from the course content. Use a rule-based entity recognition method or a dictionary matching method to analyze the knowledge point data. Entity recognition first matches through a predefined entity dictionary. For example, in the welding field, the dictionary includes terms such as "welding equipment", "welding wire", and "heat source". All terms that appear in the course content are matched with the dictionary. If the term matches successfully, then the term is considered an entity. The threshold during the matching process can be set as the frequency of text appearance. For example, if a certain term appears more than 2 times in the entire course content, then it is considered a valid entity. Entity recognition also needs to be further screened according to the context. Adopt rules based on part-of-speech tagging to remove non-entity words such as verbs and adjectives, and only retain noun-class entities. During the recognition process, dependency parsing technology is used to analyze the sentence structure, identify noun phrases and verb phrases, and combine the context to further extract entities representing specific concepts and objects. Through this step, terms such as "welding process", "robot", and "heat input" in the course are extracted as knowledge point entity data, and finally an entity set containing entities and their attributes is formed.

[0037] Step S13: Extract relationships from the knowledge point entity data to obtain knowledge point entity relationship data; In this embodiment, according to the knowledge point entity data, a relation extraction method is used to mine the relations between entities. The task of relation extraction is to identify the semantic connections between entities. For example, the relation between a "welding robot" and a "welding process" is "execute". Relation extraction first performs text dependency analysis on the entity data, constructs a syntactic tree, and annotates the grammatical relations between entities. Through dependency relation analysis, the relations between entities are identified, such as the subject-predicate relation between verbs or the action relation of the object. At this time, a threshold needs to be set according to the preset relation types (such as "execute", "involve", "provide") for relation screening. For example, if the frequency of a verb in a sentence exceeds a set value (such as more than 3 times), it is considered that the verb represents a certain specific relation. Relation extraction can be carried out based on the method of template matching, that is, some common statement templates are defined in advance, such as "[Entity 1] executes [Entity 2]", and the matching relations are searched in the text. Through this method, relations such as "a welding robot executes a welding process" and "a welding process involves heat input" between knowledge point entities are identified. The finally generated knowledge point entity relation data includes the relation type and relevant description between each pair of entities.

[0038] Step S14: Construct a course knowledge graph based on the knowledge point entity data and the knowledge point entity relation data, so as to obtain course knowledge graph data.

[0039] In this embodiment, the knowledge point entities are used as the nodes in the knowledge graph, and the entity relations are used as the edges connecting the nodes. In graph construction, the nodes represent knowledge point entities, and the edges represent the relations between entities. A graph database (such as Neo4j) is used as the storage and construction tool, and the relation data of the nodes and edges is imported into the database through the Cypher query language. Each knowledge point entity node contains detailed information, such as attributes such as the name, description, and type of the knowledge point, and the relation edge contains the specific relation between entities and its weight information. The weight of the relation can be adjusted according to the frequency of the entity relation or its importance in the context. For example, if the relation between a "welding robot" and a "welding process" appears in multiple course materials, the weight of this relation is higher; otherwise, the weight is lower. After the construction of the nodes and edges is completed, the knowledge graph is optimized through graph algorithms to ensure that the connection relations between entities reflect the logical structure of the course content. The finally generated course knowledge graph data contains complete knowledge point entities and their mutual relations, which serve as the basis for subsequent teaching analysis, personalized recommendation, and other operations.

[0040] Optionally, step S13 is specifically: Step S131: Format the knowledge point entity data to obtain formatted knowledge point entity data; In this embodiment, the knowledge point entity data is formatted. The knowledge point entity data contains multiple attributes, such as knowledge point name, description, type, related knowledge points, etc. The purpose of formatting is to uniformly convert this information into a standardized structure for subsequent processing. It is stored using a structured data format (such as JSON or CSV) to ensure that the fields of each knowledge point entity are clearly defined. The specific steps include: First, perform type conversion on the data according to the field type (such as string, integer, boolean, etc.); then, extract the attributes of each knowledge point entity in the form of columns or key-value pairs. For example, the knowledge point "welding process" contains the following fields: Name: welding process, Description: the interaction between materials and heat sources during the welding process, Type: process. According to actual requirements, assign a suitable format to each field, such as text type, numerical type, or enumeration type. If some fields are missing, they can be filled according to the set rules (such as setting default values or leaving them blank). After formatting, the data is stored in a standardized form for further processing.

[0041] Step S132: Extract syntactic features based on the knowledge point entity relationship data to obtain knowledge point entity relationship syntactic feature data; In this embodiment, the relationship data contains the connections between entities. The task of syntactic feature extraction is to identify and extract the syntactic information that can represent the entity relationships. First, tokenize and perform part-of-speech tagging on the sentences containing entity relationships to ensure the part-of-speech and syntactic structure of each word are clear. Use the dependency parsing method to identify the relationships between words in the sentence through the dependency tree, especially the connections between verbs and nouns. For example, in the sentence "The welding robot executes the welding process", the relationship between the verb "executes" and the noun phrase "welding process" can be obtained through dependency analysis. In the relationship data, extract verbs, nouns, and relationship words (such as "executes", "contains") as syntactic features. The syntactic features of each relationship include information such as verb type, sentence structure, and dependency path. These syntactic features are extracted through a rule engine or manual annotation to ensure the accuracy and consistency of the features. Finally, the syntactic feature data should include the relationship types between each entity pair and their related syntactic information, such as verbs or prepositions, and their relative positions in the sentence.

[0042] Step S133: Construct a knowledge point entity relationship recognition model based on the knowledge point entity relationship syntactic feature data to obtain a knowledge point entity relationship recognition model; In this embodiment, a model structure suitable for relation extraction is selected, such as a rule-based model, a supervised learning model (such as SVM, decision tree), or a deep learning model (such as convolutional neural network, recurrent neural network). If a machine learning-based model is selected, it is necessary to first annotate the syntactic feature data to ensure that the relationship labels of each pair of entities in the dataset are clear. For each entity pair, its relationship type is annotated, such as "execute", "involve", "depend on", etc. After the dataset is constructed, it is divided into a training set, a validation set, and a test set. The training set is used to train the model. The input includes the syntactic features of entity relationships and the annotated relationship types. The model gradually adjusts its parameters by learning the mapping relationship between features and labels until a better prediction accuracy is achieved. During the training process, the best model parameters, such as the learning rate, regularization coefficient, etc., are selected through cross-validation and hyperparameter tuning. After training is completed, the model and its parameters are saved and evaluated on the validation set and the test set to ensure that its generalization ability and accuracy meet the expectations. Finally, the obtained knowledge point entity relationship recognition model can be used for entity pair extraction and relationship determination in the subsequent steps.

[0043] Step S134: Extract entity pairs from the formatted data of knowledge point entities according to the knowledge point entity relationship recognition model, so as to obtain entity pair data; In this embodiment, the formatted entity data is input into the trained relationship recognition model. Each entity data contains multiple attributes, such as name, description, type, etc. The model first extracts potential entity pairs in the text through preprocessing. For example, "welding robot" and "welding process" are recognized as a pair of entity pairs. Then, the model analyzes each pair of entities according to the learned relationship features to determine whether there is a relationship between them. The process of entity pair extraction includes two steps: the first step is entity pair recognition, that is, to determine whether the text contains multiple entities and pair them; the second step is to use the knowledge point entity relationship recognition model to determine the relationship of each pair of entities. For example, the model determines the relationship between "welding robot" and "welding process" as "execute" by analyzing their relationship. For each pair of entities, the model outputs the entity pair and its relationship type. Finally, the result of entity pair extraction is a set containing entity pairs and their relationships.

[0044] Step S135: Make a relationship judgment on the obtained entity pair data according to the knowledge point entity relationship syntactic feature data, so as to obtain knowledge point entity relationship data.

[0045] In this embodiment, the entity pair data consists of each pair of entities and their potential relationships. The next step is to determine the specific relationships between these entity pairs according to predefined relationship rules or through model analysis. This step requires the use of syntactic feature information learned in the relationship extraction model, such as verbs, nouns, and their dependency paths. If the relationship features between entity pairs meet a certain preset threshold (for example, if the occurrence frequency of a certain verb exceeds 3 times, then this verb can be used as a relationship identifier), it is considered that there is a certain semantic relationship between them. For each pair of entities, the model first matches its syntactic features with the relationship patterns in the training data, and then determines the relationship type. During the relationship judgment process, thresholds are used to filter and confirm relationships. For example, if the dependency path of a certain verb is consistent with the dependency path of the "execution" relationship in the training data, it is determined that the relationship between this pair of entities is "execution". Finally, the relationship type of each pair of entities is output to generate complete knowledge point entity relationship data. This data set will contain entity pairs and their relationship types, which are used for subsequent knowledge graph construction and analysis.

[0046] Optionally, step S2 is specifically as follows: Step S21: Extract course assignment features from the teaching course data to obtain course assignment data; In this embodiment, the structure of the teaching course data is analyzed, including information such as assignment topics, submission times, grading criteria, answering times, etc. Then, feature extraction is performed based on this information. For example, the difficulty level of the assignment topics, the knowledge points involved, question types (multiple-choice questions, fill-in-the-blank questions, subjective questions, etc.), the number of words in the questions, and the answer formats. These feature extractions can be performed through regular expressions, text processing methods, or directly extracting keywords from the course design documents. For each assignment data, attribute information such as the knowledge points related to the assignment, assignment category, question type, and difficulty is extracted and formatted into a standardized structure, such as storing it in JSON format, containing key-value pairs like "question type: multiple-choice question", "knowledge point: basic mechanical welding". Finally, the generated course assignment data is structured data for subsequent processing.

[0047] Step S22: Collect paper images of the course assignment data to obtain course assignment paper images; In this embodiment, according to the requirements in the course assignment data, a high-resolution camera or scanner is used to collect the paper image of the course assignment. When selecting a camera, ensure that its resolution is not lower than 300 DPI to ensure the clarity of the image and avoid difficulties in subsequent processing due to too low resolution. During the collection process, ensure uniform lighting to avoid the influence of the shadow of the light source on the image quality. According to the attributes of the course assignment data, set the collection parameters, including exposure time, aperture size, and focus distance, to ensure that the obtained image is not distorted. When collecting the image, it is necessary to ensure that the paper is located in the center of the camera's field of view and is flat, avoiding bending or deformation. After the collection is completed, save the image in a common format, such as TIFF or PNG, for subsequent processing, and ensure that the image size conforms to the actual requirements.

[0048] Step S23: Perform grayscale conversion on the course assignment paper image to obtain the grayscale image of the course assignment paper; In this embodiment, the color image is converted into a grayscale image. This process is achieved by weighted averaging the pixel values of the RGB three channels, where the commonly used weighting coefficients are 0.2989 (red), 0.5870 (green), and 0.1140 (blue). The specific calculation formula is: Gray = 0.2989 * R + 0.5870 * G + 0.1140 * B, where R, G, and B are the red, green, and blue components of each pixel in the image respectively. The converted image only has grayscale information, thus reducing the computational complexity. At this time, the color information in the image is discarded, and only the brightness information of the image is retained. The value of each pixel in the grayscale image is usually between 0 and 255, where 0 represents black and 255 represents white. During the conversion process, the original resolution of the image is maintained and it is saved in a grayscale image format, such as PNG or JPEG format.

[0049] Step S24: Perform paper contour edge detection on the grayscale image of the course assignment paper to obtain the paper contour edge data; In this embodiment, common edge detection methods include Canny edge detection and Sobel operator. The Canny edge detection method is selected here, and its process includes the following steps: first, use a Gaussian filter to smooth the image and remove the noise in the image, usually using a Gaussian kernel of 3x3 or 5x5; then, calculate the gradient value of the image, and use the Sobel operator to calculate the horizontal and vertical gradients of the image; then, apply the non-maximum suppression algorithm to refine the edge to ensure the clarity of the edge; finally, use the double threshold method to connect the edge, set the high threshold and the low threshold (for example, 50 and 100), the edge points below the low threshold are suppressed, the points above the high threshold are retained, and the points between the two are retained based on their connection conditions. Through the above steps, the contour edge data of the course homework paper is obtained, which is represented by the edge lines or edge areas in the image, usually a binary image, the edge part is white, and the rest is black.

[0050] Step S25: performing defective area recognition on the paper outline edge data, thereby obtaining paper defective area data; In this embodiment, continuous regions in the image are identified by analyzing the connected domain of edge data. Morphological operations, especially dilation and erosion operations, are used to enhance edge lines and fill in existing blank areas. The dilation operation uses a 3x3 structural element to expand the edge data to ensure the integrity of the contour; the erosion operation removes small noise areas. After these basic operations are completed, the connected component analysis (CCA) method is used to detect separated areas in the image. If an area is obviously missing or abnormal in shape, it is considered to be a defective area of ​​the paper. For example, if a part of the edge data is missing or discontinuous, and the area of ​​the area is less than a certain threshold (such as 50 pixels), it is marked as a defective area. By analyzing the identification and position of all defective areas, the data of the defective areas of the paper are obtained, and the coordinates or area information of these areas will be used in subsequent steps.

[0051] Step S26: Calculate the degree of damage of the grayscale image of the homework paper according to the paper damage area data, so as to obtain paper damage degree data.

[0052] In this embodiment, the area of ​​the entire paper can be calculated by calculating the number of all pixels in the paper image. Then, the total area of ​​the defective area can be calculated by performing area calculation on the defective area obtained in step S25 to obtain the total number of pixels in all defective areas. Next, the defect degree of the paper is calculated, and the formula is: ; For example, if the area of ​​the entire paper image is 2000 pixels and the total area of ​​the damaged area is 150 pixels, the damage degree of the paper is 7.5%. This damage value reflects the degree of paper damage and can be further analyzed or classified according to preset standards.

[0053] Optionally, step S25 is specifically: Step S251: Calculating the distance between adjacent contour points on the paper contour edge data, thereby obtaining the adjacent contour point distance data; In this embodiment, the coordinates of all edge points in the contour edge data are obtained, and the Euclidean distance between adjacent edge points is calculated in sequence according to the coordinates of the edge points. The Euclidean distance calculation formula is: ; Among them, (x1, y1) and (x2, y2) are the coordinates of two adjacent points respectively. In actual operation, all adjacent edge points in the contour edge data are traversed, their distances are calculated one by one, and the calculation results are saved as an array or list. This method can obtain the distance data of each pair of adjacent contour points. For areas with complex contours, especially paper edges with multiple curves and multiple corners, special attention should be paid to the sorting and connection of edge points to ensure accurate calculation of the distance between adjacent points. The data in this step is used for subsequent detection of whether the paper is broken.

[0054] Step S252: obtaining a safety threshold of the distance between adjacent contour points; In this embodiment, a safety threshold for subsequent paper break detection is obtained. The threshold defines the maximum safety distance between adjacent points on the edge of a normal paper. In practical applications, a suitable threshold is determined by analyzing the contour data of a large number of papers and measuring the distance between adjacent contour points of normal paper. For example, under experimental conditions, if the distance between adjacent contour points of normal paper is generally between 1 and 5 mm, 5 mm can be set as the safety threshold. This threshold is adjusted according to the type of paper material, the thickness of the paper and other physical properties. The safety threshold should be obtained through experimental data or standardized literature, and repeatedly verified in the actual process to ensure its applicability on all paper samples.

[0055] Step S253: performing paper break detection on the adjacent contour point distance data according to the adjacent contour point distance safety threshold, thereby obtaining paper break data; In this embodiment, the distance data of all adjacent contour points are traversed and compared with the safety threshold. If the distance between adjacent points is greater than the safety threshold, it is considered that there is a possibility of paper breakage in this area, and the position of the break point is recorded. Through this process, the area of paper breakage can be extracted from the image. The marking of the break point should be as accurate as possible to ensure that the subsequent processing can accurately identify the break area. For the calculation of the break point, the threshold can be adjusted through experimental data when setting the threshold. Generally, it is recommended to set the maximum allowable distance of breakage to 1.5 times or 2 times the threshold to ensure accuracy.

[0056] Step S254: Classify according to the paper breakage data to obtain paper tear data and paper corner missing data; In this embodiment, for each break area, its morphological features are calculated, such as the length and width of the crack and the shape of the fissure. If the fissure is long and regular, it is classified as a paper tear; if the fissure is small and irregular, it is a paper corner missing. In specific implementation, the ratio of the area and shape of the area can be used for judgment. For the torn area, usually the length of its crack exceeds a certain threshold (such as more than 10 millimeters), and the width of the fissure is large. The corner missing area refers to the area where the fissure is relatively small and local, usually located at the corner of the paper and generally has a small area. In this way, the break types of the paper are divided into two types: tear and corner missing, providing data support for subsequent positioning.

[0057] Step S255: Locate the paper tear area in the grayscale image of the course assignment paper according to the paper tear data to obtain the paper tear area data; In this embodiment, the coordinate data of the paper tear are obtained, including the boundary and center position of the torn area. Then, image processing technology is used for accurate positioning of the torn area. For example, through template matching or edge tracking algorithms, an area similar in size and shape to the torn area is found in the grayscale image of the paper. For the detection of the torn area, image segmentation technology such as threshold segmentation or contour-based segmentation can be used to determine the specific position of the tear. Through these methods, the position of the torn area is marked in the paper image, and the relevant coordinate or area data are output for subsequent analysis and processing.

[0058] Step S256: Locate the paper corner missing area in the grayscale image of the course assignment paper according to the paper corner missing data to obtain the paper corner missing area data; In this embodiment, the coordinates and shape features of the corner - missing area are obtained. The corner - missing area is usually located at the corner of the paper and is relatively small. The corner area can be identified through morphological operations in image processing, especially erosion and dilation. For the identification of the corner - missing area, the minimum area of the region (such as less than 50 square millimeters) can be set as the criterion for the corner - missing area, and the angle of the corner - missing area is usually close to a right angle. By analyzing these features, the specific position of the corner - missing area in the paper image is identified, the corner - missing area is marked, and its relevant coordinate information is obtained.

[0059] Step S257: Merge the torn area data and the corner - missing area data of the paper to obtain the incomplete area data of the paper.

[0060] In this embodiment, the coordinate data of the torn area and the corner - missing area are compared. If the two areas are adjacent or overlapping, they are merged into a unified incomplete area. When merging, geometric calculation methods such as the minimum bounding rectangle or the convex hull algorithm can be used to obtain the minimum boundary of the merged area. During the merging process, it is ensured that the merged area does not contain any errors or overlapping parts, and the shape and size of all incomplete areas should be within a reasonable range. The incomplete area data of the merged paper, including information such as position, shape, and area, will provide key data support for subsequent teaching resource recommendation.

[0061] Optionally, step S3 is specifically as follows: Step S31: Extract the interactive features and error features of the course assignment based on the course assignment data to obtain the course assignment interactive data and the course assignment error data; In this embodiment, the interactive features of the course assignment data are extracted. This process includes analyzing data such as the participation situation, submission frequency, and interaction frequency with teachers or classmates from the assignments submitted by students. Specifically, when implemented, the time points when each student submits the assignment, the assignment modification frequency, the number of interactive comments, and the online communication records with teachers or classmates are statistically analyzed. These data constitute the interactive features of the assignment. By calculating the frequencies or time intervals of these features, the interactive data of each student's assignment is formed. Next, the error features are extracted. The extraction of error features is based on the types of incorrect answers of students in the assignment, such as spelling mistakes, conceptual mistakes, arithmetic mistakes, etc. This step automatically identifies and classifies the error types by comparing with the preset standard answers. The occurrence frequency of each error type, the number of incorrect answers, and their distribution are extracted as error features. By extracting these two types of data for each student's assignment, the course assignment interactive data and the course assignment error data are obtained as the basis for subsequent analysis.

[0062] Step S32: Conduct in - depth subject analysis on the course assignment interactive data to obtain the in - depth subject data; In this embodiment, in-depth analysis at the subject level is carried out according to various indicators in the interaction data (such as homework submission time, interaction frequency, depth of problem discussion, etc.). Through the clustering analysis method, students with a relatively high frequency of homework interaction and in-depth discussion are identified to determine which interactions exhibit a relatively high subject depth. Characteristics such as interaction frequency, technicality and complexity of problem discussion can be used as the measurement criteria for subject depth. In actual operation, when analyzing the content of the interaction, some criteria are needed. For example, the frequency of professional terms appearing in the discussion content, whether students can raise effective questions, and whether in-depth academic discussions are carried out among students. In addition, if some problems involve more complex application scenarios or the integration of interdisciplinary knowledge, it indicates a relatively high subject depth. In this process, a subject depth threshold can be set. For example, when the depth of discussion of a certain homework reaches 80%, it is determined to have a relatively high depth. Finally, based on these analysis results, a subject depth data is constructed as an important reference for the assessment of the difficulty of course homework.

[0063] Step S33: Analyze the complexity of the error types of the course homework error data to obtain error type complexity data; In this embodiment, according to the error feature data, the complexity of each error type is analyzed. For example, for math errors, they are divided into basic calculation errors, formula application errors, problem-solving idea errors, etc., and a complexity score is defined for each error type. Specifically, basic calculation errors can be set as low complexity, and formula application errors and problem-solving idea errors are graded according to the depth of students' understanding of the problem and set as medium or high complexity. In this process, a standardized scoring system needs to be defined according to the type of problem and the severity of the error. For example, the complexity score of basic calculation errors is set as 1, formula application errors are set as 2, and problem-solving idea errors are set as 3. The criteria for error complexity scoring can be obtained through statistics on previous homework data to get a reliable assessment criterion. By marking and analyzing the error types of each student, the error type complexity data of each student is obtained, which is convenient for subsequent assessment of the difficulty of course homework.

[0064] Step S34: Evaluate the difficulty of the course homework according to the error type complexity data and the subject depth data to obtain the course homework difficulty data.

[0065] In this embodiment, the subject depth data and the error type complexity data are combined, and through the weighted average method or multiple regression analysis, the comprehensive difficulty score of each course assignment is obtained. The weights of the subject depth data and the error type complexity data in the difficulty assessment can be adjusted according to the actual course objectives. For example, if the course objective focuses on students' in-depth understanding, the weight of the subject depth data can be set to 60%; if the course objective focuses on students' problem-solving ability, the weight of the error type complexity data can be set to 40%. On this basis, the comprehensive difficulty score of each assignment is calculated by weighting. Specifically, a difficulty scoring standard can be set, for example: an assignment with a difficulty score less than 20 is an easy assignment, 20 to 40 is of medium difficulty, and above 40 is a difficult assignment. Finally, the difficulty data of the course assignments are obtained according to the evaluation results, which is convenient for teachers to make personalized teaching arrangements according to the situations of different students.

[0066] Optionally, step S32 is specifically as follows: Step S321: Conduct an interaction frequency statistics on the course assignment interaction data to obtain high-frequency course assignment interaction data; In this embodiment, the course assignment interaction data is sorted and classified to determine each student's interaction behavior in the assignment. The interaction data includes the number of times a student submits an assignment, the online communication frequency with teachers or classmates, the discussion frequency of assignment questions, etc. By analyzing these behavioral data, the interaction frequency of each student within a specified time is statistically calculated. Specifically, the interaction frequency of each student with teachers or classmates during the assignment completion process is calculated, and the timestamp of the interaction behavior is recorded to ensure the accuracy of the data. For each interaction item, a frequency threshold is set, such as at least 5 interactions per week are considered high-frequency interactions. By comparing the interaction frequency of each student with the threshold, the high-frequency course assignment interaction data is screened out. The high-frequency interaction data of each student is calculated by taking the ratio of its interaction frequency to the total number of times, forming a proportional index. Finally, the obtained high-frequency interaction data can be used for subsequent interaction feature extraction and subject reasoning analysis.

[0067] Step S322: Extract high-level interaction features based on the high-frequency course assignment interaction data to obtain high-level interaction data; In this embodiment, higher-level interaction behaviors involving subject depth, complexity of problem discussion, etc. are screened out from the high-frequency interaction data. During specific operations, by classifying each interaction content, the interaction content related to the depth of subject knowledge is extracted. For example, whether students raise challenging questions and whether they can participate in complex knowledge discussions. For each interaction record, complexity indicators of the interaction are defined and calculated, such as the technical terms involved in the interaction, the depth of academic discussion, etc. By setting criteria, such as if the discussion involves at least 3 technical terms, this interaction is marked as a high-level interaction. In addition, the fineness of the interaction content needs to be analyzed. If the content proposed by students can demonstrate relatively systematic knowledge application or logical reasoning, it is marked as a high-level interaction. Through the above methods, high-level interaction data is extracted from the high-frequency interaction data, providing a basis for subsequent in-depth analysis of subject reasoning.

[0068] Step S323: Conduct in-depth analysis of subject reasoning on the high-level interaction data to obtain subject reasoning depth data; In this embodiment, by analyzing each piece of high-level interaction data, it is checked whether students demonstrate high reasoning ability. For example, during the discussion, whether they can give appropriate examples to support their views, whether they can combine multiple knowledge points for analysis, or whether they can draw conclusions through reasoning. During specific operations, a logical analysis tool is used to evaluate the complexity of the reasoning process for each interaction content. For example, a reasoning depth scoring criterion can be set. If a student can reason from one knowledge point to another complex conclusion, the reasoning depth score is 2. If it involves multiple subjects or cross-field reasoning, the reasoning depth score is 3. Each piece of interaction data is marked according to the score of reasoning complexity, and finally a subject reasoning depth data set is formed. This data set reflects the reasoning ability and subject depth demonstrated by students in their homework, providing a quantitative basis for subsequent difficulty assessment and subject depth division.

[0069] Step S324: Obtain students' course assignment answer data; In this embodiment, the specific answer data of students in course assignments is obtained. The answer data includes information such as the answer content, answer time, answer steps, etc. of each assignment submitted by students. During specific operations, the assignment answers of each student are extracted from the assignment management system to ensure data integrity, and the answer text of the students and the time stamps used for answering are recorded. Each assignment answer should include the answer content, problem-solving ideas, and completeness of answer filling of each question by students. At the same time, the answer time should also be recorded as an important indicator to further provide reference data for subsequent answer logic analysis. To ensure the accuracy of the data, the data can be exported through the assignment system or the answer data can be synchronized in real time through the API interface.

[0070] Step S325: Analyze the answering logic of the course assignment based on the answering data of the student's course assignment, so as to obtain the answering logic data of the course assignment; In this embodiment, analyze the steps and logical relationships in the student's answering process. For example, check whether the student follows the standard calculation steps sequentially when solving a math problem, or whether certain important steps are skipped or omitted during the answering process. At this time, use the step sequence analysis method to judge whether the student follows a reasonable logical structure by recording each step of the student's answer and its sequence. For each question, set the standard answer process, compare the student's problem-solving steps with the standard process, and calculate the correctness and integrity of the answering logic. The logical analysis also includes the evaluation of the student's problem-solving strategy, such as whether there is a behavior of bypassing certain problem-solving steps. Finally, obtain the answering logic data of the course assignment through these analyses for subsequent subject depth division.

[0071] Step S326: Divide the subject reasoning depth data according to the answering logic data of the course assignment, so as to obtain the subject depth data.

[0072] In this embodiment, combined with the logical analysis in the student's answering process, check whether the student shows in-depth subject reasoning during the answering process. By evaluating the student's answering logic, set the grade standard for subject reasoning. For example, if the student's answering process is complete, the reasoning process is rigorous, and can demonstrate in-depth subject understanding, it is classified as a high reasoning depth level. If there are obvious logical jumps or errors in the answer, it is classified as a low reasoning depth level. To concretize this process, certain thresholds can be set based on indicators such as the number of steps in the answer, the correctness of each step, and the interdisciplinary connections in the student's reasoning process. For example, if the steps in the answer are complete and the correct rate is higher than 80%, the reasoning depth is classified as "high", otherwise it is classified as "low". Finally, obtain the subject depth data of each student, providing a basis for course difficulty assessment and teaching optimization.

[0073] Optionally, step S335 is specifically: Mark the answering process according to the answering data of the student's course assignment, so as to obtain the answering process data; In this embodiment, the answer contents submitted by students are collected, including each question in the assignment and their answering situations. Then, time series analysis is performed on the answering process of each student, and the time points and answer contents of each step are recorded. For example, record the time spent by students on each question and whether they follow the standard answering order. Each answering step needs to be detailedly marked in terms of answering time and answer content to ensure integrity. Each link in the answering process can be marked precisely by setting marking points, such as "start answering question 1", "finish answering question 1", etc. This process can also be automated through programming. By importing data into the system, timestamps are automatically recorded for each operation of the student to ensure that each question and its answering steps are clearly marked. Finally, the answering process data with time dimension and step details is formed.

[0074] Answer skipping detection is performed on the answering process data to obtain answer skipping data; In this embodiment, answer skipping means that students skip some steps or problem-solving links during the answering process, resulting in an incoherent answering process. By analyzing each step mark in the answering process, it is detected whether students skip the intermediate steps of a certain question or directly move on to the next question. For example, if a student directly enters the result when doing a math problem without performing step-by-step calculations, this behavior is recorded as answer skipping. In implementation, each answering step can be compared one by one through programming to check whether the time interval between steps is too long or too short, exceeding the set normal range. For example, if there is no response for more than 5 minutes, it can be determined as answer skipping. It can also be determined whether there is an answer skipping phenomenon by calculating the staying time of students on a specific question. If the staying time is abnormally short (below the preset threshold, such as below 10 seconds) or abnormally long (exceeding 10 minutes) and there is no step to move on to the next question, it can also be considered that there is an answer skipping phenomenon. Finally, a set of answer skipping data is obtained, marked as "answer skipping" or "no answer skipping", and classified and summarized for subsequent analysis.

[0075] Obtain the answering reasoning path data of students' course assignments; In this embodiment, the reasoning path refers to the thinking path and problem-solving steps used by students when answering each question. To obtain the reasoning path data, it is necessary to record the thinking process or operation behavior of students in each answering step. In specific operations, the thinking process and methods of students to solve each question can be described by designing questionnaires or interacting with students, or the operation steps selected by them during the answering process can be recorded through automated recording tools. For example, for a complex math problem, the reasoning path may include that the student first selects the known conditions in the question, then applies the formula for derivation, and finally obtains the answer. Through an automated tool, the input and selection of each step of the student's operation are recorded to generate a complete dataset of the answering reasoning path. This dataset includes the reasoning path sequence of each question, such as input data, selected steps, problem-solving ideas, etc., and provides a reference for subsequent abnormal reasoning detection.

[0076] Identify abnormal reasoning in the answering process data based on the reasoning path data of students' course assignment answers, so as to obtain abnormal reasoning data; In this embodiment, by analyzing the rules of the students' reasoning paths, a standard reasoning process is defined. For example, when solving a certain math problem, the standard reasoning path should include steps such as analyzing the known conditions, listing equations, and performing calculations. If there are jumps or behaviors that do not conform to the standard reasoning path in the students' answering process (such as directly giving the answer and omitting the reasoning steps), they can be marked as abnormal reasoning. In specific operations, comparative analysis can be carried out using the reasoning path data to detect whether the reasoning path of each student conforms to the predetermined standard process. By writing a program, each student's reasoning path is compared with the standard path, the difference value is calculated, and a threshold is set. For example, if the deviation between the reasoning path and the standard path is more than 10%, it is considered abnormal reasoning. Finally, a set of abnormal reasoning data is generated, marking which answering processes have irregular reasoning behaviors, and recording the specific time points when they occur and the types of their deviations.

[0077] Integrate the answering logic of the course assignment according to the answering skip-step data and the abnormal reasoning data, so as to obtain the answering logic data of the course assignment.

[0078] In this embodiment, the skip-step data and the abnormal reasoning data are combined with the answer process data to analyze whether there are any illogical or unreasonable situations in the student's answer process. During the specific operation, logical judgment is performed on the student's answer behavior by setting rules. For example, if a student has a skip-step phenomenon or abnormal reasoning behavior, the system will mark this part of the answer as incomplete or incorrect. These abnormal data are separated from the normal data, and the student's answer logic is re-evaluated. By comparing the standard answer process and the actual answer process of each question, a threshold is set to define "qualified" or "abnormal" answer behaviors. If there are abnormalities exceeding the threshold in a certain step or reasoning path, the system will mark it as "logical abnormality". Finally, all the logical abnormality data in the answer process are integrated to obtain the complete answer logic data for the course assignment. These data can be used for further analysis of the student's answer behavior, evaluation of the answer quality, adjustment of teaching strategies, etc.

[0079] Optionally, step S33 is specifically as follows: Step S331: Classify the error types of the course assignment error data to obtain the basic error data of the course assignment and the applied error data of the course assignment; In this embodiment, according to the types of assignment questions, the errors are divided into basic errors and applied errors. Basic errors usually refer to the errors that students ignore basic concepts, formulas or operation steps during the answer process. For example, in circuit design, the connection rules of basic components are ignored, or basic arithmetic errors are made in math problems. Applied errors are the errors that students make when applying basic knowledge. For example, during the actual circuit debugging process, the components are not reasonably selected or the components are not correctly connected, resulting in the circuit not working properly. To make this classification, the system classifies the students' assignments according to the preset error type criteria. By writing programs, the judgment criteria for basic errors and applied errors are set. For example, if a student fails to connect a resistor to the power supply in a circuit assignment, it is determined as a basic error; if a student connects the wrong combination of components, resulting in the circuit not working properly, it is determined as an applied error. Finally, based on the data of each assignment step, all errors are classified and recorded, and the basic error data and the applied error data are generated respectively.

[0080] Step S332: Statistically analyze the component usage errors based on the basic error data of the course assignment to obtain the component usage error data; In this embodiment, it is necessary to extract error items related to the use of components from the basic error data. For example, students may wrongly select components such as capacitors, resistors, and diodes in circuit design, or the connection method is incorrect. By analyzing the basic error data, the system identifies which students have made mistakes in using components, such as connecting a resistor to the wrong circuit node or not connecting components according to the standard diagram. The system will count the types of component errors in each student's assignment, record the types of components, the frequency of incorrect operations, and the locations where these errors occur. During implementation, an error code marking mechanism can be adopted. Each time a component usage error occurs, the program will generate an error code and a corresponding description, recording the occurrence times of each error type. When counting, a threshold can be set. For example, if the component error in each question appears more than twice, it is regarded as a high-frequency error and thus marked as an important error type. Finally, the system will generate detailed component usage error data, including information such as the type of component, the type of error, and the occurrence frequency.

[0081] Step S333: Perform circuit debugging error statistics based on the application error data of the course assignment to obtain circuit debugging error data; In this embodiment, application errors usually occur during circuit debugging. When students fail to effectively use circuit analysis tools or wrongly connect components, the circuit cannot work properly. This step first needs to screen out errors related to circuit debugging from the application error data. Common circuit debugging errors include: not correctly measuring the current and voltage in the circuit, using incorrect debugging tools or test instruments, or ignoring the adjustment of some important parameters in the circuit (such as current limit, voltage distribution, etc.). The system scans the operation records of students in the circuit debugging link, identifies each step in the debugging process and analyzes it. Each operation step will be marked as an error or normal. If a certain step fails to meet the standard setting (for example, the test voltage should be 5V, but it is actually 6V), then this step will be classified as a circuit debugging error. The error frequency and type will be recorded through statistical analysis. The set statistical criteria include the number of error occurrences, the type of error, and the link where the error occurs. The system will summarize the circuit debugging error data, record the occurrence frequency and distribution of the error types for subsequent analysis.

[0082] Step S334: Calculate the error complexity based on the component usage error data and the circuit debugging error data to obtain error type complexity data.

[0083] In this embodiment, the error complexity refers to the situation where the occurrence of an error is not just a simple operational mistake, but involves the interaction of multiple links and multiple factors, resulting in a more complex error. For example, in circuit design, an error that simultaneously involves incorrect component selection, incorrect connection, and debugging error is considered more complex than a single error. First, the system assigns weights to each error type, and basic errors and application errors are classified according to their complexity. Then, by combining and analyzing the incorrect component usage data and circuit debugging error data, the complexity of each error type is calculated. During implementation, a weighted formula can be used to represent the error complexity. For example, different weight values are assigned to basic errors and application errors (e.g., the weight of basic errors is 1, and the weight of application errors is 2). The system calculates the overall error complexity of each student by counting the error types of each student. If a student's assignment involves multiple component errors and circuit debugging errors, these errors will be weighted and combined to finally obtain an error complexity value. This complexity value reflects the overall difficulty or complexity of the errors encountered by the student during the answering process.

[0084] Optionally, this specification also provides a creation system for a multi-functional AI teaching platform, which is used to execute the creation method for a multi-functional AI teaching platform as described above. The creation system for a multi-functional AI teaching platform includes: A curriculum knowledge graph construction module, which is used to obtain teaching curriculum data and construct a curriculum knowledge graph based on the teaching curriculum data, so as to obtain curriculum knowledge graph data; A paper mutilation degree detection module, which is used to extract curriculum assignment feature based on the teaching curriculum data to obtain curriculum assignment data; and detect the paper mutilation degree according to the curriculum assignment data to obtain paper mutilation degree data; A curriculum assignment difficulty analysis module, which is used to extract curriculum assignment interaction features and curriculum assignment error features based on the curriculum assignment data to obtain curriculum assignment interaction data and curriculum assignment error data; and analyze the curriculum assignment difficulty according to the curriculum assignment interaction data and curriculum assignment error data to obtain curriculum assignment difficulty data; A teaching personalized resource recommendation module, which is used to recommend teaching personalized resources according to the curriculum assignment difficulty data and the paper mutilation degree data to obtain teaching personalized resource data; A teaching platform creation module, which is used to update the curriculum knowledge graph data according to the teaching personalized resource data to obtain updated curriculum knowledge graph data, and create a teaching platform according to the updated curriculum knowledge graph data to obtain teaching platform data.

[0085] The creation system for the multi-functional AI teaching platform of the present invention can implement any creation method for the multi-functional AI teaching platform of the present invention, and is a medium for coordinating operations and signal transmissions between various modules to complete the creation method for the multi-functional AI teaching platform. The internal modules of the system cooperate with each other, thereby enhancing the intelligent and personalized service capabilities of the teaching platform.

[0086] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Thus, all changes falling within the meaning and scope of the equivalent elements of the application document are intended to be embraced by the present invention.

[0087] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather will conform to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for creating a multifunctional AI teaching platform, characterized in that: The following steps are involved: Step S1: Acquire teaching course data, and construct a course knowledge graph based on the teaching course data, thereby obtaining course knowledge graph data; Step S2: extracting course assignment features according to the teaching course data, thereby obtaining course assignment data; performing paper defect detection according to the course assignment data, thereby obtaining paper defect data; Step S3: extracting course assignment interaction features and course assignment error features based on the course assignment data, thereby obtaining course assignment interaction data and course assignment error data; Conducting course assignment difficulty analysis based on course assignment interaction data and course assignment error data, thereby obtaining course assignment difficulty data; Step S4: Recommend personalized teaching resources based on the course homework difficulty data and paper incompleteness data, thereby obtaining personalized teaching resource data; Step S5: Update the course knowledge graph data according to the personalized teaching resource data to obtain the course knowledge graph update data, and create a teaching platform according to the course knowledge graph update data to obtain the teaching platform data.

2. The method for creating a multifunctional AI teaching platform according to claim 1, characterized in that: Step S1 is specifically as follows: Step S11: obtaining teaching course data, and marking knowledge points according to the teaching course data, thereby obtaining knowledge point data; Step S12: performing entity recognition according to the knowledge point data, thereby obtaining knowledge point entity data; Step S13: extracting relationships from knowledge point entity data to obtain knowledge point entity relationship data; Step S14: Construct a course knowledge graph based on the knowledge point entity data and the knowledge point entity relationship data to obtain the course knowledge graph data.

3. The method for creating a multifunctional AI teaching platform according to claim 2, characterized in that: Step S13 is specifically as follows: Step S131: formatting the knowledge point entity data to obtain knowledge point entity formatted data; Step S132: extracting grammatical features according to the knowledge point entity relationship data, thereby obtaining knowledge point entity relationship grammatical feature data; Step S133: constructing a knowledge point entity relationship recognition model according to the knowledge point entity relationship grammatical feature data, thereby obtaining a knowledge point entity relationship recognition model; Step S134: extracting entity pairs from the knowledge point entity formatted data according to the knowledge point entity relationship recognition model, thereby obtaining entity pair data; Step S135: Perform relationship judgment on the entity pair data according to the knowledge point entity relationship grammatical feature data, so as to obtain the knowledge point entity relationship data.

4. The method for creating a multifunctional AI teaching platform according to claim 1, characterized in that: Step S2 is specifically as follows: Step S21: extracting course assignment features according to the teaching course data, thereby obtaining course assignment data; Step S22: collecting paper images of the course assignment data, thereby obtaining a course assignment paper image; Step S23: performing grayscale conversion on the homework paper image to obtain a grayscale image of the homework paper; Step S24: performing paper contour edge detection according to the course homework paper grayscale image, thereby obtaining paper contour edge data; Step S25: performing defective area recognition on the paper outline edge data, thereby obtaining paper defective area data; Step S26: Calculate the degree of damage of the grayscale image of the homework paper according to the paper damage area data, so as to obtain paper damage degree data.

5. The method for creating a multifunctional AI teaching platform according to claim 4, characterized in that: Step S25 is specifically as follows: Step S251: Calculating the distance between adjacent contour points on the paper contour edge data, thereby obtaining the adjacent contour point distance data; Step S252: obtaining a safety threshold of the distance between adjacent contour points; Step S253: performing paper break detection on the adjacent contour point distance data according to the adjacent contour point distance safety threshold, thereby obtaining paper break data; Step S254: classifying the paper breakage data into different types, thereby obtaining paper tearing data and paper corner chipping data; Step S255: locating the paper tear area of ​​the course homework paper grayscale image according to the paper tear data, thereby obtaining paper tear area data; Step S256: locating the paper missing corner area of ​​the course homework paper grayscale image according to the paper missing corner data, thereby obtaining the paper missing corner area data; Step S257: merging the paper defective area data according to the paper tearing area data and the paper corner chipping area data, thereby obtaining the paper defective area data.

6. The method for creating a multifunctional AI teaching platform according to claim 1, characterized in that: Step S3 is specifically as follows: Step S31: extracting course homework interaction features and course homework error features according to the course homework data, thereby obtaining course homework interaction data and course homework error data; Step S32: Performing subject depth analysis on the course assignment interaction data to obtain subject depth data; Step S33: performing error type complexity analysis on the course assignment error data, thereby obtaining error type complexity data; Step S34: Evaluate the difficulty of the course assignment based on the error type complexity data and the subject depth data, thereby obtaining the course assignment difficulty data.

7. The method for creating a multifunctional AI teaching platform according to claim 6, characterized in that: Step S32 is specifically as follows: Step S321: performing interactive frequency statistics on the course homework interactive data, thereby obtaining high-frequency course homework interactive data; Step S322: extracting high-level interactive features based on the high-frequency course homework interactive data, thereby obtaining high-level interactive data; Step S323: performing subject reasoning depth analysis on the high-level interaction data, thereby obtaining subject reasoning depth data; Step S324: Obtaining student course homework answer data; Step S325: Performing a course homework answer logic analysis based on the student course homework answer data, thereby obtaining the course homework answer logic data; Step S326: Divide the subject reasoning depth data into subject depths according to the course assignment answer logic data, thereby obtaining subject depth data.

8. The method for creating a multifunctional AI teaching platform according to claim 7, characterized in that: Step S335 is specifically as follows: Marking the answer process according to the student's course homework answer data, thereby obtaining the answer process data; Performing answer skipping detection on the answering process data, thereby obtaining answer skipping data; Obtain the reasoning path data of students’ answers to course assignments; According to the reasoning path data of the students' course homework answers, abnormal reasoning is identified on the answer process data, so as to obtain abnormal reasoning data; The logic of answering the course assignment is integrated according to the answer skipping data and the abnormal reasoning data, so as to obtain the logic data of answering the course assignment.

9. The method for creating a multifunctional AI teaching platform according to claim 6, characterized in that: Step S33 is specifically as follows: Step S331: classifying the course assignment error data into error types, thereby obtaining course assignment basic error data and course assignment application error data; Step S332: Count component usage errors based on basic course assignment error data, thereby obtaining component usage error data; Step S333: performing circuit debugging error statistics according to the course homework application error data, thereby obtaining circuit debugging error data; Step S334: performing error complexity calculation according to component usage error data and circuit debugging error data, thereby obtaining error type complexity data.

10. A system for creating a multifunctional AI teaching platform, characterized in that: Used to execute the method for creating a multifunctional AI teaching platform as claimed in claim 1, the creation system for the multifunctional AI teaching platform comprises: The course knowledge graph construction module is used to obtain teaching course data and construct a course knowledge graph based on the teaching course data, thereby obtaining course knowledge graph data; The paper defect detection module is used to extract course assignment features based on teaching course data, thereby obtaining course assignment data; and to perform paper defect detection based on course assignment data, thereby obtaining paper defect data; A course assignment difficulty analysis module is used to extract course assignment interaction features and course assignment error features based on course assignment data, thereby obtaining course assignment interaction data and course assignment error data; and to perform course assignment difficulty analysis based on the course assignment interaction data and course assignment error data, thereby obtaining course assignment difficulty data; The personalized teaching resource recommendation module is used to recommend personalized teaching resources based on the course homework difficulty data and paper incompleteness data, thereby obtaining personalized teaching resource data; The teaching platform creation module is used to update the course knowledge graph data according to the personalized teaching resource data, thereby obtaining the course knowledge graph update data, and to create the teaching platform according to the course knowledge graph update data, thereby obtaining the teaching platform data.

Images