Big model-driven smart education resource recommendation method and system

Abstract

The present invention discloses a large-scale model-driven intelligent educational resource recommendation method and system, which relates to the field of data processing technology. The method includes: forming an educational resource set, and using a first intelligent agent to identify the points of action of the resources to establish an educational resource pool; a second intelligent agent combines learning records and interest preferences to analyze and generate a first resource demand set; a third intelligent agent combines learning goals and learning records to generate a second resource demand set; compares the differences between the first resource demand set and the second resource demand set, generates a resource demand list, and provides the list as a personalized recommendation list to students. The present invention solves the technical problems that existing educational resource recommendation methods cannot meet the personalized needs of students and the recommendation results lack dynamic adaptability and accuracy. It achieves the technical effect of realizing personalized and dynamically adaptable educational resource recommendations and improving recommendation accuracy through multi-agent collaborative analysis and dynamic feedback mechanism.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a large model-driven smart education resource recommendation method and system. Background Art

[0002] With the increasing popularity of online education, educational resource recommendation systems have become a crucial tool for improving learning experience and efficiency. However, existing systems lack the ability to provide personalized recommendations. Firstly, recommendation systems struggle to fully integrate diverse data sources, such as students' learning goals, learning history, and interests, resulting in inaccurate recommendations. Secondly, recommendation results lack dynamic adaptability and cannot be adjusted in real time based on students' learning progress and evolving needs. Summary of the Invention

[0003] This application provides a large-model-driven intelligent educational resource recommendation method and system to solve the technical problems that existing educational resource recommendation methods cannot meet students' personalized needs and the recommendation results lack dynamic adaptability and accuracy.

[0004] The first aspect of the present application provides a large-model-driven intelligent educational resource recommendation method, which includes: forming an educational resource set, and using a first intelligent agent among the multiple intelligent agents to identify the action point of the first resource in the educational resource set to obtain a first action label set; establishing an educational resource pool based on the correspondence between the first resource and the first action label set; obtaining a multi-source database of student users, wherein the multi-source database includes the learning goals, learning records and interest preferences of the student users; using a second intelligent agent among the multiple intelligent agents, in combination with the educational resource pool, to perform a collaborative analysis of the learning records and the interest preferences to obtain a first resource requirement set; using a third intelligent agent among the multiple intelligent agents, in combination with the educational resource pool, to perform a collaborative analysis of the learning goals and the learning records to obtain a second resource requirement set; introducing an intelligent agent debate strategy to compare and analyze the first resource requirement set with the second resource requirement set, generate a resource requirement list, and use the resource requirement list as a target recommendation list for the student user.

[0005] The second aspect of the present application provides a large model-driven intelligent education resource recommendation system, the system comprising: an action point identification module, the action point identification module is used to form an education resource set, and perform action point identification on the first resource in the education resource set through the first intelligent agent among the multiple intelligent agents to obtain a first action label set; an education resource pool establishment module, the education resource pool establishment module is used to establish an education resource pool according to the correspondence between the first resource and the first action label set; a multi-source data acquisition module, the multi-source data acquisition module is used to obtain a multi-source database of student users, wherein the multi-source database includes the learning goals, learning records and interest preferences of the student users; a first resource demand analysis module Block, the first resource demand analysis module is used to use the second intelligent agent among the multiple intelligent agents in combination with the educational resource pool to collaboratively analyze the learning records and the interest preferences to obtain a first resource demand set; the second resource demand analysis module, the second resource demand analysis module is used to use the third intelligent agent among the multiple intelligent agents in combination with the educational resource pool to collaboratively analyze the learning goals and the learning records to obtain a second resource demand set; the comparative analysis module, the comparative analysis module is used to introduce an intelligent agent debate strategy to compare and analyze the first resource demand set and the second resource demand set, generate a resource demand list, and use the resource demand list as a target recommendation list for the student user.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The large-scale model-driven intelligent educational resource recommendation method and system provided in this application relate to the field of data processing technology. Through the collaboration of multiple intelligent agents, an educational resource set is constructed and an educational resource pool is generated. Combined with the students' learning goals, records and interest preferences, the second and third intelligent agents collaborate to analyze and generate a resource demand set. An intelligent agent debate strategy is introduced to compare the two types of demand sets, and finally a personalized recommendation list is generated to accurately recommend educational resources that meet the students' needs. This solves the technical problems that existing educational resource recommendation methods cannot meet the personalized needs of students and the recommendation results lack dynamic adaptability and accuracy. It realizes the technical effect of realizing personalized and dynamically adaptable educational resource recommendations and improving recommendation accuracy through multi-agent collaborative analysis and dynamic feedback mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A flowchart of a large model-driven smart education resource recommendation method provided in an embodiment of the present application;

[0010] Figure 2 A schematic diagram of the structure of the large model-driven smart education resource recommendation system provided in the embodiment of this application.

[0011] Explanation of the accompanying symbols: action point identification module 11, education resource pool establishment module 12, multi-source data acquisition module 13, first resource demand analysis module 14, second resource demand analysis module 15, comparative analysis module 16. DETAILED DESCRIPTION

[0012] This application provides a large-model-driven intelligent educational resource recommendation method and system to solve the technical problems that existing educational resource recommendation methods cannot meet students' personalized needs and the recommendation results lack dynamic adaptability and accuracy.

[0013] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0014] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0015] Example 1, as Figure 1 As shown, this application provides a large model-driven smart education resource recommendation method, which includes:

[0016] P10: Build an educational resource set, and use a first agent among the multiple agents to identify an action point on a first resource in the educational resource set to obtain a first action label set.

[0017] Furthermore, step P10 in the embodiment of the present application further includes:

[0018] P11: Extract the knowledge point set and skill point set embedded in the first intelligent agent; P12: Match and identify the first resource based on the knowledge point set and the skill point set in turn, and obtain the first knowledge point label set and the first skill point label set respectively; P13: Form the first action label set based on the first knowledge point label set and the first skill point label set.

[0019] It should be understood that a comprehensive and diverse educational resource set must first be assembled. This set encompasses a wide range of educational resources, including but not limited to textbooks, courseware, lesson plans, test questions, video courses, and more. These resources form the fundamental data source for the recommendation system. To enable the system to deeply understand the characteristics and value of each educational resource, the first agent among multiple agents must identify the points of action for each resource in the set, thereby generating a first set of action labels. This process is one of the key steps in achieving personalized educational resource recommendations.

[0020] Specifically, the first step in action point identification is to extract the embedded knowledge point set and skill point set from the first agent. A knowledge point set refers to the specific knowledge content related to educational resources. For example, in mathematics, knowledge points may include linear algebra and probability theory. A skill point set focuses on learner skills that educational resources can cultivate or enhance, such as data analysis and logical thinking. These knowledge point and skill point sets are obtained through the pre-trained knowledge base of the large model and the annotation and organization of domain experts. The large model has already learned a large amount of text data during the pre-training phase and is able to identify and extract knowledge points and skill points related to educational resources.

[0021] Next, educational resources are matched and identified based on the extracted knowledge point sets and skill point sets. First, knowledge point matching and identification are performed, comparing the content of the educational resource with the knowledge point set to identify the knowledge points covered in the educational resource. For example, for a mathematics textbook, text analysis can identify the knowledge points contained therein, such as linear algebra and calculus. This process utilizes the semantic understanding capabilities of the large model, which can accurately match the text content in the educational resource with the predefined knowledge point set to generate the first knowledge point label set. Then, skill point matching and identification are performed. Similarly, the content of the educational resource is compared with the skill point set to generate the first skill point label set.

[0022] Finally, based on the first knowledge point label set and the first skill point label set, the first role label set is formed. The role label set refers to a collection of specific roles and application scenarios that educational resources can play in the educational process. For example, the role label set of a mathematics textbook may include basic knowledge explanation, example problem analysis, etc. By combining knowledge point labels and skill point labels, the role and application scenarios of educational resources can be more comprehensively described. For example, a mathematics textbook that includes linear algebra knowledge points and data analysis skill points can have a role label set that includes basic linear algebra knowledge explanation, data analysis ability training, etc. This role label set can not only reflect the content of the educational resource, but also reflect its actual application value in the educational process.

[0023] Through the above steps, a detailed and accurate role label set is generated for each educational resource, providing a solid foundation for subsequent educational resource recommendations.

[0024] P20: Establish an education resource pool based on the correspondence between the first resource and the first role tag set.

[0025] Specifically, an educational resource pool is established based on the correspondence between the first resource and the first role tag set. This process is to systematically integrate the detailed labeled and classified educational resources so that personalized educational resources can be efficiently retrieved and matched later.

[0026] For example, first, each educational resource and its corresponding first set of role tags are associated and stored. Each educational resource (such as textbooks, courseware, lesson plans, etc.) will establish a direct connection with the previously identified role tags (such as basic knowledge explanation, programming practice, etc.). This connection can be implemented in the form of a key-value pair in the database, where the educational resource serves as the key and its corresponding role tag set serves as the value. For example, a mathematics textbook may be marked with role tags such as linear algebra basic knowledge explanation and calculus example analysis. These tags will be stored as attributes of the textbook in the educational resource pool.

[0027] Then, all labeled educational resources are classified and organized according to their role labels to create a structured educational resource pool, so that the system can quickly retrieve the corresponding educational resources based on different role labels. For example, when the system needs to recommend a textbook on the basics of linear algebra for students, it can directly search for resources with this role label in the educational resource pool without having to screen them one by one from a large number of unclassified resources, thereby improving retrieval efficiency and achieving refined management and accurate recommendation of educational resources. For example, the following table is shown:

[0028] Table 1 Resource pool list

[0029] Resource Name Resource Type Main knowledge point tags Main Skill Point Tags Calculus textbook Teaching Materials Limits, differential laws, integrals Mathematical problem solving Algebra Problem Solving Videos video Algebraic operations, equation solving Problem-solving skills, logical reasoning Mathematical thinking exercises Test Questions Mathematical thinking, problem analysis Problem-solving skills, mathematical reasoning Mathematics Lesson Plan Lesson Plan Sequences, Limit Theory Academic writing, instructional design

[0030] Furthermore, the establishment of an educational resource pool also requires consideration of resource updates and maintenance. As educational resources are continuously updated and expanded, the data within the pool also needs to be updated accordingly. This includes adding new educational resources and their role tags, as well as adjusting and supplementing the role tags of existing resources. Furthermore, the data within the pool needs to be regularly reviewed and cleaned to ensure its accuracy and validity. This not only helps improve resource utilization but also provides students with more precise and personalized learning support.

[0031] P30: Acquire a multi-source database of student users, wherein the multi-source database includes the learning goals, learning records, and interest preferences of the student users.

[0032] Optionally, to achieve accurate personalized recommendations, a comprehensive understanding of student learning situations and needs is necessary. Therefore, acquiring a multi-source database of student users is a crucial step. This multi-source database includes student learning goals, learning history, and interests, providing rich information support for subsequent recommendations.

[0033] First, collect student learning goals. Learning goals are specific outcomes or competence levels that students hope to achieve through learning. These goals can be short-term, such as achieving a high score on an upcoming exam, or long-term, such as mastering expertise in a specific field to prepare for future career development. They can be set by students themselves, or teachers or learning advisors can provide recommendations based on students' learning progress and competence levels. Data sources can include student self-reports, questionnaires, and study plans. Alternatively, teachers can set or adjust learning goals for students based on their learning performance and potential, and record this information in the learning management system.

[0034] Next, collect student users' learning records. Learning records record in detail the various behaviors and performances of students during the learning process, including course viewing time, homework completion, test scores, participation in discussions, etc., which can reflect students' learning habits, learning progress, and knowledge mastery. Data sources can include online learning platforms, where students' activities such as watching course videos, submitting homework, and taking tests will generate a large number of learning records. These platforms are usually equipped with data tracking and analysis functions that can automatically collect and store this data. And through teacher feedback, teachers observe and evaluate students' learning performance in class and manually enter feedback information into the learning management system as part of the learning record.

[0035] Finally, collect student user interest preferences. Understanding student interest preferences helps recommend educational resources that match their interests, thereby increasing student learning motivation and engagement. Data sources can include student self-reports, where students can clearly express their interests in different subjects and topics through questionnaires or interest tests; and learning behavior analysis, which can infer student interest preferences by analyzing student behavior on the learning platform, such as the types of videos watched, the topics discussed, and the courses selected.

[0036] After obtaining a student's learning goals, learning history, and interests, this data needs to be integrated into a unified multi-source database. This database should feature structured storage, storing different types of data (such as text, numbers, and timestamps) in structured tables to facilitate query and analysis, and ensure that data from different sources can be linked. For example, a student's learning goals can be linked to their corresponding learning history and interests for comprehensive analysis. Furthermore, as students progress and their interests change, the multi-source database needs to be regularly updated to reflect their latest learning progress.

[0037] By establishing such a comprehensive and dynamically updated multi-source database, the recommendation system can gain an in-depth understanding of each student's learning needs and preferences, thereby providing students with more accurate and personalized educational resource recommendations.

[0038] P40: Through the second intelligent agent among the multiple intelligent agents, the learning records and the interest preferences are collaboratively analyzed in combination with the educational resource pool to obtain a first resource requirement set.

[0039] Furthermore, step P40 in this embodiment of the present application further includes:

[0040] P41: respectively obtain the knowledge point advancement list of the knowledge point set and the skill point advancement list of the skill point set; P42: based on the learned role set of the student user obtained by analyzing the learning record, filter the knowledge point advancement list and the skill point advancement list to obtain a first candidate resource pool; P43: based on the preferred role set of the student user obtained by analyzing the interest preference, filter the first candidate resource pool to obtain a second candidate resource pool; P44: the second intelligent agent records the resources in the second candidate resource pool as the first resource demand set.

[0041] It should be understood that in order to accurately meet the learning needs of student users, it is necessary to use the second intelligent agent among multiple intelligent agents to conduct a collaborative analysis of the student users' learning records and interest preferences in combination with the educational resource pool to obtain the first resource demand set.

[0042] First, the second agent obtains the advancement lists of the knowledge point set and the skill point set respectively. The knowledge point advancement list and the skill point advancement list are a guide to the student's learning progress, showing the learning order of each knowledge point and skill point that the student needs to master. For example, for a mathematics course, the knowledge point advancement list may start from basic algebra and geometry knowledge, and gradually advance to advanced content such as calculus and linear algebra; the skill point advancement list may include everything from basic mathematical problem-solving skills to advanced mathematical modeling capabilities. The second agent will use these advancement lists to determine the student's current learning stage, identify the knowledge points and skill points that they have not yet mastered, and filter out corresponding educational resources for these knowledge points and skill points.

[0043] Next, the second agent will filter the knowledge point advancement list and skill point advancement list based on the student's learned role set to obtain the first candidate resource pool. The learned role set refers to the knowledge and skills that the student has mastered, which reflects the level that the student has reached in the learning process. By analyzing the learned role set, the second agent can identify the gaps in the student's knowledge or skill points and select relevant resources from the advancement list. These resources will be used to fill the gaps in the student's learning and ensure that the recommended resources meet the student's current learning needs. For example, if a student has mastered basic algebra but has not yet learned calculus, the second agent will filter out calculus-related textbooks, videos, or exercises from the resource pool as part of the first candidate resource pool.

[0044] Then, the second agent will further screen according to the student's interest preferences to obtain the second candidate resource pool. Interest preferences reflect the student's personal interests and tendencies in the learning process. Students may be more interested in certain subjects, and this interest will affect their acceptance and participation in learning resources. For example, if a student has a strong interest in programming, but there is less mathematics content involved in his learning record, the second agent will give priority to recommending programming-related educational resources rather than resources related to mathematics content. Through this interest-based screening, the second agent can ensure that the recommended resources not only meet the students' learning needs, but also stimulate students' interest in learning and improve their learning enthusiasm. For example, as shown in the following table:

[0045] Table 2 Statistics of students' interest preferences

[0046] Resource Name Student browsing frequency Student interest preferences Whether the predetermined frequency limit is reached Calculus textbook 15 High (Mathematics) yes Programming video courses 5 High (Programming) no Mathematical thinking exercises 10 Medium (Logical Reasoning) yes Physics textbooks 3 Low (Physics) no

[0047] Finally, the second agent organizes the resources in the second candidate resource pool and records them as the first resource demand set. This first resource demand set is a selection of educational resources that best meet the student's current learning needs and interests. These resources not only match the student's current learning progress and knowledge base, but also meet their interests and preferences. They serve as input to the subsequent recommendation system, ensuring that subsequent recommendations accurately match the student's needs and maximize learning outcomes and student experience.

[0048] Furthermore, the embodiment of the present application further includes step P44a, which further includes:

[0049] P44-1a: Extract the implicit behavioral strategy embedded in the second intelligent agent; P44-2a: Based on the implicit behavioral strategy, combine the educational resource pool to perform implicit analysis on the learning behavior in the learning record to obtain an implicit resource requirement set; P44-3a: Calibrate the first resource requirement set based on the implicit resource requirement set.

[0050] In a possible embodiment of the present application, in order to further improve the accuracy and personalization of recommendations, the first resource requirement set can be calibrated by deeply analyzing the implicit meaning of student users' learning behavior to ensure that the recommended resources not only meet explicit needs but also meet potential learning needs.

[0051] First, embedded behavioral implicit strategies are extracted from the second agent. These strategies analyze students' learning behavior data to uncover the underlying learning needs and motivations behind explicit behaviors. For example, if a student repeatedly watches instructional videos and submits homework assignments on a particular topic but still receives unsatisfactory grades, these behaviors may indicate that the student is struggling with the topic or needs more practice. Leveraging the deep learning capabilities and natural language processing techniques of large models, these behavioral implicit strategies can identify and interpret these implicit needs.

[0052] Next, the second agent performs an implicit analysis of the learning behaviors in the students' learning records based on the extracted behavioral implicit strategies and the data in the educational resource pool. The students' learning behavior data is extracted from the learning records, such as the course viewing time, the number of homework submissions, test scores, discussion participation, etc. The behavioral implicit strategies are then used to analyze the potential needs behind these learning behaviors. For example, if a student repeatedly pauses and rewatches the video of a certain chapter and scores low in the related homework, this may indicate that the student has difficulty understanding the knowledge points in that chapter. Through the semantic understanding and reasoning capabilities of the large model, it is possible to further analyze that students may need more exercises, detailed example analysis, or related tutoring videos, and then generate an implicit resource demand set. This set contains the educational resources that students potentially need. These resources may not be directly reflected in explicit learning records or interest preferences, but are crucial to students' learning progress.

[0053] Finally, the first resource requirement set is calibrated based on the implicit resource requirement set. The implicit resource requirement set is compared and analyzed with the first resource requirement set to identify the differences and complementary relationships between the two. For example, if the implicit resource requirement set contains detailed example analysis resources required by a student, but the first resource requirement set does not contain such resources, then the first resource requirement set needs to be supplemented by adding the necessary resources identified in the implicit resource requirement set to the first resource requirement set to ensure that the recommended resources can fully cover the student's learning needs, including both explicit and potential needs. The calibrated first resource requirement set will serve as the final resource requirement set for subsequent educational resource recommendations.

[0054] This calibration process ensures a high degree of personalization in the recommendation system and improves the accuracy of recommendations. This analysis method based on behavioral implicit strategies enables the system to continuously optimize resource recommendations during the dynamic learning process, helping students maintain high learning efficiency and motivation despite their ever-changing learning needs.

[0055] Furthermore, step P44-2a of the embodiment of the present application further includes:

[0056] P44-21a: Extract browsing behavior records from the learning behavior; P44-22a: Obtain any browsing educational resources in the browsing behavior records, and obtain the student user's arbitrary browsing frequency of the arbitrary browsing educational resources; P44-23a: If the arbitrary browsing frequency reaches a predetermined frequency limit, then according to the behavior implicit strategy, add the arbitrary browsing educational resources to the implicit resource demand set.

[0057] Specifically, the process of generating the implicit resource requirement set can be further refined and refined. When conducting implicit analysis, browsing behavior records are first extracted from students' learning records. Browsing behavior records provide a detailed record of students' browsing activities on the learning platform, including pages viewed, duration of stay, and click paths. These records can reflect students' attention and interests in different educational resources. For example, a student may repeatedly browse a specific course chapter or video tutorial, which may indicate the appeal or importance of this resource to the student.

[0058] Next, we collect the frequency of students browsing any educational resource from their browsing behavior records. For example, we can count how many times students access a course video, check a particular exercise, or jump to a specific chapter in the textbook within a certain time period. Through this frequency analysis, we can determine the level of attention paid to each resource and understand students' potential learning needs.

[0059] Next, determine whether the student's browsing frequency of any educational resource has reached a predetermined frequency limit. If the frequency reaches the predetermined frequency limit, then, based on the behavioral implicit strategy, the educational resource is added to the implicit resource demand set. The predetermined frequency limit is a pre-set threshold used to determine whether a student's interest in a resource is strong enough to be included in the implicit resource demand set. The setting of this threshold can be dynamically adjusted based on the system's empirical data to ensure that the student's true needs are captured. If the browsing frequency exceeds the predetermined limit, the resource is considered a potential demand resource for the student.

[0060] This browsing behavior-based analysis method enables the system to automatically discover students' hidden needs and make intelligent recommendations for resources through behavioral implicit strategies.

[0061] P50: Through the third intelligent agent among the multiple intelligent agents, the learning objectives and the learning records are collaboratively analyzed in combination with the educational resource pool to obtain a second resource requirement set.

[0062] Optionally, to more comprehensively meet the learning needs of students, a third agent among the multiple agents can be used to collaboratively analyze the student's learning goals and learning history in conjunction with the educational resource pool to derive a second set of resource requirements. This process aims to accurately identify the educational resources a student may need currently and in the future by comprehensively considering their learning goals and actual learning progress.

[0063] First, the third agent obtains the student's learning goals and learning history. Learning goals reflect the specific outcomes or competence levels that students hope to achieve through learning, while learning history details the student's various behaviors and performance during the learning process. By analyzing the degree of alignment between learning goals and learning history, the third agent can identify potential gaps and needs in students' progress toward achieving their learning goals.

[0064] For example, if a student's learning goal is to master the knowledge of "Data Structures and Algorithms," but their learning records show that they have low test scores and few homework submissions on the knowledge point "Linked Lists," this indicates that the student may need more learning resources to strengthen their understanding and application of this knowledge point. The third agent will combine data from the educational resource pool to find educational resources related to "Linked Lists," such as detailed explanation videos, exercises, case studies, etc., and include these resources in the second resource requirement set.

[0065] The third agent also considers the student's learning progress and the coherence of their knowledge system. For example, if a student has mastered arrays and stacks but has yet to learn queues, the third agent might recommend queue-related learning resources to help the student build a complete knowledge system. Furthermore, if the student's learning goals are long-term, the third agent might recommend advanced algorithms and data structure courses based on the student's long-term learning plan.

[0066] During the collaborative analysis process, the third agent leverages the semantic understanding and reasoning capabilities of the large model to deeply explore the potential relationships between learning goals and learning records. For example, by analyzing students' study time and performance changes on different knowledge points, it can predict the difficulties they may encounter in subsequent learning and recommend relevant learning resources in advance.

[0067] Ultimately, the third agent generates a second resource requirement set, which includes the educational resources that students may need based on their learning goals and learning records. These resources can not only help students fill their current learning gaps but also support them in achieving their long-term learning goals, thereby providing them with more comprehensive and personalized learning support.

[0068] P60: Introduce an intelligent agent debate strategy to compare and analyze the first resource requirement set and the second resource requirement set, generate a resource requirement list, and use the resource requirement list as a target recommendation list for the student user.

[0069] Specifically, an agent debate strategy was introduced to compare and analyze the first and second resource requirement sets, ultimately generating a list of resource requirements. This process simulates the human debate mechanism, allowing different agents to evaluate and discuss resource requirements from different perspectives, thereby ensuring the comprehensiveness and rationality of the recommendation results.

[0070] For example, first, the debate coordinator among multiple agents compares and analyzes a first resource requirement set and a second resource requirement set. The first resource requirement set is generated based on the student user's learning history and interest preferences, reflecting the student's current learning needs and interests. The second resource requirement set is generated based on the student's learning goals and learning history, reflecting the additional resources the student needs to achieve their learning goals.

[0071] The debate coordinator compares the resources in the two resource request sets one by one, analyzing each resource's frequency of occurrence, importance, and alignment with students' learning goals and interests. For example, if a resource appears in both the first and second resource request sets and is highly aligned with students' learning goals and interests, it will be prioritized for inclusion in the final resource request list.

[0072] Next, we introduce an agent debate strategy. Multiple agents (including but not limited to the second and third agents) will engage in a debate about the necessity, importance, and applicability of each resource. Each agent, based on its own analysis and strategy, will offer an evaluation of the resource. For example, the second agent might emphasize how well a resource meets a student's interests, while the third agent might emphasize its importance in helping students achieve their learning goals. This debate mechanism allows for a full exchange and collision of viewpoints and opinions among different agents, leading to a more comprehensive assessment of the value of each resource.

[0073] During the debate, the debate coordinator is responsible for summarizing and organizing the opinions of each agent and conducting a comprehensive evaluation of each resource. If a resource is recognized by the majority of agents during the debate, it will be included in the final resource demand list. The debate coordinator also ranks the resource demands based on the debate results, ensuring that the resources in the recommended list are arranged according to importance and priority.

[0074] Finally, the generated resource requirement list serves as a target recommendation list for student users. This recommendation list not only includes resources that meet students' current learning needs and interests, but also considers the additional resources they will need to achieve their learning goals. This provides students with a comprehensive, personalized, and forward-looking learning resource recommendation solution. By introducing an agent-based debate strategy, the large-scale model-driven intelligent education resource recommendation method can more accurately identify and meet students' learning needs, improving the quality and applicability of recommendation results.

[0075] Furthermore, before using the resource requirement list as a target recommendation list for the student user, the embodiment of the present application further includes step P60a, which further includes:

[0076] P61a: Constructing a learning profile of the student user based on the multi-source database; P62a: Traversing the learning profile in the online user portraits to obtain a matching portrait set; P63a: Using the matching user set corresponding to the matching portrait set as the peers of the student user, and using the matching portrait set as the peer information of the peers; P64a: Adjusting the resource demand list according to the peer information.

[0077] It should be understood that in order to further optimize the recommendation results and ensure that the recommended resources not only meet the student users' own learning needs, but also refer to the learning paths and resource usage of peers of the same age or similar learning background, the recommendation system can understand students' needs more accurately by constructing student learning profiles and comparing them with other student profiles.

[0078] First, a learning profile of the student user is constructed based on a multi-source database. This database contains multi-dimensional data such as the student's learning goals, learning history, and interests. By comprehensively analyzing this data, a learning profile can be generated that comprehensively reflects the student's learning characteristics. The learning profile includes not only basic student information such as grade and major, but also detailed information such as the student's learning progress, knowledge mastery, learning style, and interests. For example, a student's learning profile may show a high interest and good grades in mathematics, but difficulty with English writing.

[0079] Next, the constructed learning profile is traversed through the online user profile database to find other users with similar learning characteristics as the student user, thereby obtaining a matching profile set. The online user profile database stores a large number of learning profiles of other student users. Through comparison and analysis, it is possible to identify student groups with similar learning goals, learning progress, and interests as the target student user. For example, if the target student user is a sophomore interested in computer science and currently studying data structures, the matching profile set may include other sophomores majoring in computer science with similar backgrounds and interests.

[0080] The user groups corresponding to the matching profiles are then considered the target student's peers, and the matching profiles serve as the peer information for the peers. Peer information reflects the learning paths, resource usage, and learning outcomes of these similar students. By analyzing peers' learning behaviors and resource preferences, valuable information can be obtained to optimize resource recommendations for the target student. For example, if peers commonly use a specific online course when learning data structures and achieve good results, then this course is likely to be helpful to the target student as well.

[0081] Finally, the resource needs list is adjusted based on peer information. For example, resources frequently recommended by peers and relevant to the target student's learning goals can be prioritized in the resource needs list, or the resources in the list can be sorted and optimized to ensure that the recommended resources are more closely aligned with the target student's learning needs and learning path. Furthermore, peer learning paths can be referenced to recommend resources that will contribute to the student's long-term learning development, rather than simply satisfying their current learning needs.

[0082] Through the above steps, the embodiment of the present application not only considers the student user's own learning needs, but also incorporates the learning experience and resource usage of others, thereby providing students with more comprehensive, personalized, and valuable educational resource recommendations. This method helps students learn from the successful experiences of others, avoid common learning problems, and improve learning efficiency and effectiveness.

[0083] Furthermore, step P62a of the embodiment of the present application further includes:

[0084] P62-1a: randomly extracting a first portrait from the online user portraits; P62-2a: obtaining a first similarity between the learning portrait and the first portrait, wherein the calculation formula for the first similarity is expressed as follows:

[0085] ; ; ;in, Characterize the Euclidean similarity between the learning portrait and the first portrait, Characterize the cosine similarity between the learning portrait and the first portrait, Characterizing the first similarity, and a learning vector representing the learning portrait and a first vector representing the first portrait respectively, Characterize the learning vector The first dimension of the feature dimensional features, represents the feature dimension of the learning vector, is a positive integer, Characterizing the first vector The first dimension of the feature dimensional features, represents the dot product of the learning vector and the first vector, and Respectively represent the module lengths of the learning vector and the first vector; P62-3a: when the first similarity reaches a predetermined similarity limit, the first portrait is added to the matching portrait set.

[0086] Optionally, the comparison process between a student's learning profile and other student profiles can be further refined and optimized. By randomly extracting user profiles and calculating similarities, the peer profiles that best match the current student profile can be selected to form a matching profile set.

[0087] First, we randomly extract one of the online user profiles as the first profile. The online user profile database contains a large number of learning profiles of other student users. Each profile details the student's learning characteristics, such as learning goals, learning progress, and interests. Random sampling ensures sample diversity and representativeness, avoiding inaccurate matching due to selection bias.

[0088] Next, the current student's learning profile is compared with the extracted first profile, and the first similarity between the two is calculated. The similarity calculation can be based on multiple dimensions such as learning goals, learning behaviors, interest preferences, etc., and various similarity calculation methods (such as cosine similarity, Pearson correlation coefficient, etc.) are used to measure the similarity between the two profiles. Exemplarily, the above calculation formula is used to calculate the Euclidean similarity and cosine similarity between the learning profile and the first profile, respectively, and the Euclidean similarity and cosine similarity are averaged to obtain the first similarity. Euclidean similarity is used to measure the distance between two vectors in the feature space. The smaller the distance, the higher the similarity; while cosine similarity is used to measure the directional consistency of two vectors. The larger the value, the stronger the similarity between the two vectors. The combination of these two similarities can comprehensively evaluate the similarity between student profiles.

[0089] During the calculation process, the learning vector of the learning portrait and the first vector of the first portrait respectively represent the positions of the two students in the multidimensional feature space, where each dimension represents a different aspect of learning behavior or learning goals. For example, one dimension may represent a student's area of ​​interest, while another dimension may represent a student's learning progress in a certain subject. By calculating the dot product and modulus of these two vectors, the system can evaluate the similarity between student portraits. The dot product reflects the correlation between the two vectors in this space, while the modulus represents the size of each vector, further reflecting the learning intensity and amount of learning. Through these calculations, a first similarity value is ultimately obtained, which indicates the degree of similarity between the learning portrait and the first portrait. The higher the similarity, the more similar the two students are in their learning characteristics. Through this calculation, the similarity between the current student and other students in terms of learning needs, learning progress, interests, etc. is evaluated.

[0090] Then, the calculated first similarity is compared with the preset predetermined similarity limit. When the first similarity reaches the predetermined similarity limit, the first portrait is added to the matching portrait set. The predetermined similarity limit is a pre-set threshold value used to determine whether two portraits are similar enough. It can be flexibly adjusted according to the needs of the system to ensure the accuracy of the match. For example, if the predetermined similarity limit is set to 0.8, and the calculated first similarity between the learning portrait and the first portrait is 0.85, then the first portrait will be added to the matching portrait set. This process will be repeated until a sufficient number of matching portraits are extracted from the online user portraits to form a complete matching portrait set.

[0091] Through these steps, the system can efficiently filter out other user profiles with similar learning characteristics as the target student from a large number of online user profiles, thereby constructing a high-quality matching profile set. This matching profile set will serve as the basis for subsequent peer information analysis, providing students with more accurate and personalized educational resource recommendations.

[0092] Furthermore, after using the resource requirement list as a target recommendation list for the student user, the embodiment of the present application further includes step P60b, which further includes:

[0093] P61b: Activate a fourth agent among the multiple agents; P62b: Collect, through the fourth agent, multi-source recommendation effect records for the target recommendation list; P63b: Perform a weighted analysis on the multi-source recommendation effect records to obtain a recommendation effect fitness; P64b: If the recommendation effect fitness does not reach a predetermined fitness limit, optimize the recommendation for the first peer user of the student user. The multi-source recommendation effect records include at least teacher evaluation results, student evaluation results, and student written test results.

[0094] Specifically, in order to further optimize the recommendation effect of the personalized recommendation system, multiple intelligent agents can work together to perform weighted analysis of the recommendation effect based on multi-source recommendation effect records to ensure the quality and adaptability of the recommendation results.

[0095] First, activate the fourth agent. A key component of the system, the fourth agent is responsible for collecting and analyzing feedback from the recommendation system. By collaborating with other agents, this agent efficiently processes and integrates recommendation effectiveness data from various sources. Its mission is to continuously optimize recommendation algorithms and strategies based on student learning progress and feedback on resource recommendations, thereby improving the adaptability and accuracy of the recommendation system.

[0096] Next, the fourth agent begins collecting multi-source recommendation performance records for the target recommendation list. These records include teacher evaluation results, student evaluation results, and student written assessment results. Teacher evaluation results reflect teachers' comprehensive assessment of students' learning performance and class participation after using the recommended resources; student evaluation results directly come from students' feedback on satisfaction, ease of use, and helpfulness of the recommended resources; and student written assessment results assess students' knowledge mastery after using the recommended resources through written assessments such as test scores and homework completion. This multi-source data provides a comprehensive basis for subsequent analysis.

[0097] Then, a weighted analysis is performed on the collected multi-source recommendation effect records to obtain the recommendation effect fitness. Weighted analysis is to assign different weights based on the importance and reliability of recommendation effect records from different sources, thereby calculating a comprehensive recommendation effect fitness. For example, if the teacher evaluation effect is considered more important than the student evaluation effect, a higher weight can be given. Recommendation effect fitness is a quantitative indicator used to evaluate whether the overall effect of the recommendation resource has achieved the expected goal. In this way, the system can objectively evaluate the effectiveness and adaptability of the recommendation resources. For example, as shown in the following table:

[0098] Table 3 Comprehensive evaluation table of recommended resources

[0099] Recommended resource name Teacher evaluation results Student evaluation results Student written test results Weighted analysis score Calculus textbook 4.5 4.2 80% 4.4 Programming video courses 3.8 4.1 70% 3.9 Mathematical thinking exercises 4 4.3 85% 4.1 Physics textbooks 3.5 3.9 60% 3.7

[0100] Finally, if the recommendation effect fitness does not reach the predetermined fitness limit, an optimized recommendation is made to the student user's first peer user. The predetermined fitness limit is a pre-set threshold used to determine whether the recommendation effect is sufficient. If the recommendation effect fitness is lower than this threshold, it means that the current recommended resources may not fully meet the student user's learning needs. At this time, the student user's peer information, especially the first peer user's learning path and resource usage, will be referenced to optimize the student user's recommended resources. For example, if the first peer user has used different resources for similar learning goals and achieved better results, then these resources may be recommended to the student user to improve the recommendation effect fitness.

[0101] Through the above steps, the embodiment of this application not only provides personalized educational resource recommendations, but also continuously optimizes the recommendation results by collecting and analyzing multi-source recommendation effect records, ensuring that the recommended resources can truly help students achieve their learning goals and improve learning outcomes. This method not only improves the accuracy of recommendations, but also ensures the long-term effectiveness of recommended resources through dynamic adjustments.

[0102] In summary, the embodiments of the present application have at least the following technical effects:

[0103] This application uses multiple intelligent agents to work together. First, an educational resource set is formed, and the first intelligent agent identifies the points of action of the resources to generate a first set of action labels. Based on the correspondence between resources and labels, an educational resource pool is established. Next, multi-source data such as students' learning goals, learning records, and interest preferences are obtained. The second intelligent agent analyzes the learning records and interest preferences to generate a first resource requirement set; the third intelligent agent combines the learning goals and learning records to generate a second resource requirement set. By introducing an intelligent agent debate strategy, the differences between the first resource requirement set and the second resource requirement set are compared to generate a final resource requirement list, and the list is provided to students as a personalized recommendation list.

[0104] The technical effect of realizing personalized and dynamically adaptable educational resource recommendations and improving the accuracy of recommendations has been achieved through multi-agent collaborative analysis and dynamic feedback mechanism.

[0105] Example 2, based on the same inventive concept as the large model driven smart education resource recommendation method in the above embodiment, Figure 2 As shown, this application provides a large model-driven smart education resource recommendation system. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0106] An action point identification module 11 is used to form an educational resource set, and to perform action point identification on a first resource in the educational resource set by a first agent among the multiple agents to obtain a first action label set; an educational resource pool establishment module 12 is used to establish an educational resource pool based on the correspondence between the first resource and the first action label set; a multi-source data acquisition module 13 is used to obtain a multi-source database of student users, wherein the multi-source database includes the learning goals, learning records and interest preferences of the student users; a first resource demand analysis module 14 is used to analyze the first resource demand. It is used to perform a collaborative analysis of the learning records and the interest preferences through the second intelligent agent among the multiple intelligent agents in combination with the educational resource pool to obtain a first resource requirement set; the second resource requirement analysis module 15, the second resource requirement analysis module 15 is used to perform a collaborative analysis of the learning objectives and the learning records through the third intelligent agent among the multiple intelligent agents in combination with the educational resource pool to obtain a second resource requirement set; the comparative analysis module 16, the comparative analysis module 16 is used to introduce an intelligent agent debate strategy to perform a comparative analysis on the first resource requirement set and the second resource requirement set, generate a resource requirement list, and use the resource requirement list as a target recommendation list for the student user.

[0107] Furthermore, the action point identification module 11 is further configured to perform the following steps:

[0108] Extract the knowledge point set and skill point set embedded in the first intelligent agent; match and identify the first resource based on the knowledge point set and the skill point set in turn to obtain a first knowledge point label set and a first skill point label set respectively; and form the first action label set based on the first knowledge point label set and the first skill point label set.

[0109] Furthermore, the first resource demand analysis module 14 is further configured to perform the following steps:

[0110] A knowledge point advancement list of the knowledge point set and a skill point advancement list of the skill point set are respectively obtained; according to the learned role set of the student user obtained by analyzing the learning record, the knowledge point advancement list and the skill point advancement list are screened to obtain a first candidate resource pool; according to the preferred role set of the student user obtained by analyzing the interest preference, the first candidate resource pool is screened to obtain a second candidate resource pool; the second intelligent agent records the resources in the second candidate resource pool as the first resource demand set.

[0111] Furthermore, the first resource demand analysis module 14 is further configured to perform the following steps:

[0112] Extract the behavioral implicit strategy embedded in the second intelligent agent; perform implicit analysis on the learning behavior in the learning record based on the behavioral implicit strategy and in combination with the educational resource pool to obtain an implicit resource requirement set; calibrate the first resource requirement set based on the implicit resource requirement set.

[0113] Furthermore, the first resource demand analysis module 14 is further configured to perform the following steps:

[0114] Extracting browsing behavior records from the learning behavior; obtaining any browsing educational resources in the browsing behavior records, and obtaining any browsing frequency of the student user on the any browsing educational resources; if the any browsing frequency reaches a predetermined frequency limit, then according to the behavior implicit strategy, adding the any browsing educational resources to the implicit resource demand set.

[0115] Furthermore, the comparison and analysis module 16 is further configured to perform the following steps:

[0116] A learning profile of the student user is constructed based on the multi-source database; the learning profile is traversed in the online user profile to obtain a matching profile set; a matching user set corresponding to the matching profile set is used as the peers of the student user, and the matching profile set is used as the peer information of the peers; and the resource demand list is adjusted according to the peer information.

[0117] Furthermore, the comparison and analysis module 16 is further configured to perform the following steps:

[0118] Randomly extract a first portrait from the online user portraits; obtain a first similarity between the learning portrait and the first portrait, wherein the calculation formula of the first similarity is expressed as follows:

[0119] ; ; ;in, Characterize the Euclidean similarity between the learning portrait and the first portrait, Characterize the cosine similarity between the learning portrait and the first portrait, Characterizing the first similarity, and a learning vector representing the learning portrait and a first vector representing the first portrait respectively, Characterize the learning vector The first dimension of the feature dimensional features, Characterizing the first vector The first dimension of the feature dimensional features, represents the feature dimension of the learning vector, is a positive integer, represents the dot product of the learning vector and the first vector, and The module lengths of the learning vector and the first vector are respectively characterized; when the first similarity reaches a predetermined similarity limit, the first portrait is added to the matching portrait set.

[0120] Furthermore, the comparison and analysis module 16 is further configured to perform the following steps:

[0121] Activate a fourth agent among the multiple agents; collect, through the fourth agent, multi-source recommendation effect records for the target recommendation list; perform weighted analysis on the multi-source recommendation effect records to obtain a recommendation effect fitness; and if the recommendation effect fitness does not reach a predetermined fitness limit, perform an optimized recommendation for the first peer user of the student user. The multi-source recommendation effect records include at least teacher evaluation effects, student evaluation effects, and student written test effects.

[0122] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0124] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A large-scale model-driven intelligent educational resource recommendation method, characterized by: The large model includes multiple intelligent agents, and the method for recommending smart educational resources driven by the large model includes: assembling an educational resource set, and performing action point identification on a first resource in the educational resource set by a first agent among the plurality of agents, including extracting a knowledge point set and a skill point set embedded in the first agent to obtain a first action label set; Establishing an educational resource pool based on the correspondence between the first resource and the first role tag set; Acquire a multi-source database of student users, wherein the multi-source database includes the learning goals, learning records, and interest preferences of the student users; performing a collaborative analysis of the learning record and the interest preference by a second agent among the multiple agents in combination with the educational resource pool to obtain a first resource requirement set; performing, by a third agent among the plurality of agents, a collaborative analysis of the learning objectives and the learning records in combination with the educational resource pool to obtain a second resource requirement set; Introducing an intelligent agent debate strategy to compare and analyze the first resource requirement set and the second resource requirement set, generating a resource requirement list, and using the resource requirement list as a target recommendation list for the student user; The agent debate strategy is introduced to compare and analyze the first resource requirement set and the second resource requirement set, including: The debate coordinator among multiple agents compares the resources in the first resource demand set and the second resource demand set one by one, and analyzes the frequency and importance of each resource in the two demand sets, as well as its match with students' learning goals and interests; Based on the frequency, importance, and matching degree with students' learning goals and interests, multiple other agents debate the necessity, importance, and applicability of each resource, generating their own evaluation opinions. The debate coordinator synthesizes the evaluation opinions of each intelligent agent to generate the resource demand list.

2. The large model-driven smart education resource recommendation method according to claim 1, characterized in that: An educational resource set is formed, and an action point identification is performed on a first resource in the educational resource set by a first agent among the multiple agents to obtain a first action label set, including: Extracting a set of knowledge points and a set of skill points embedded in the first agent; Matching and identifying the first resource based on the knowledge point set and the skill point set in sequence to obtain a first knowledge point label set and a first skill point label set respectively; The first role tag set is formed according to the first knowledge point tag set and the first skill point tag set.

3. The large model-driven smart education resource recommendation method according to claim 2, characterized in that: A second agent among the multiple agents performs collaborative analysis on the learning record and the interest preference in combination with the educational resource pool to obtain a first resource requirement set, including: Respectively obtaining a knowledge point advancement list of the knowledge point set and a skill point advancement list of the skill point set; Filtering the knowledge point advancement list and the skill point advancement list according to the learned role set of the student user obtained by analyzing the learning record to obtain a first candidate resource pool; Filtering the first candidate resource pool according to the preference role set of the student user obtained by analyzing the interest preferences to obtain a second candidate resource pool; The second agent records the resources in the second candidate resource pool as the first resource requirement set.

4. The large model-driven smart education resource recommendation method according to claim 3, characterized in that: After the second agent records the resources in the second candidate resource pool as the first resource requirement set, the method further includes: extracting the behavioral implicit strategy embedded in the second intelligent agent; According to the behavior implicit strategy, in combination with the educational resource pool, the learning behavior in the learning record is implicitly analyzed to obtain an implicit resource demand set; The first resource requirement set is calibrated based on the implicit resource requirement set.

5. The large model-driven smart education resource recommendation method according to claim 4, characterized in that: According to the behavior implicit strategy, in combination with the educational resource pool, the learning behavior in the learning record is implicitly analyzed to obtain an implicit resource requirement set, including: Extracting browsing behavior records from the learning behavior; Obtain any browsed educational resources in the browsing behavior record, and obtain any browsing frequency of the student user on the browsed educational resources; If the arbitrary browsing frequency reaches a predetermined frequency limit, the arbitrary browsing educational resource is added to the implicit resource demand set according to the behavior implicit strategy.

6. The large model-driven smart education resource recommendation method according to claim 1, characterized in that: Before introducing the intelligent agent debate strategy to compare and analyze the first resource requirement set and the second resource requirement set to generate a resource requirement list, and using the resource requirement list as a target recommendation list for the student user, the method further includes: Constructing a learning profile of the student user based on the multi-source database; Traversing the learning portrait in the online user portraits to obtain a matching portrait set; Using the matching user set corresponding to the matching portrait set as the peers of the student user, and using the matching portrait set as the peer information of the peers; The resource requirement list is adjusted according to the peer information.

7. The large model-driven smart education resource recommendation method according to claim 6, characterized in that: Traverse the learning profile in the online user profile to obtain a matching profile set, including: Randomly extracting a first portrait from the online user portraits; Obtain a first similarity between the learning portrait and the first portrait, wherein the calculation formula of the first similarity is expressed as follows: ； ； ； Characterize the Euclidean similarity between the learning portrait and the first portrait, Characterize the cosine similarity between the learning portrait and the first portrait, Characterizing the first similarity, and a learning vector representing the learning portrait and a first vector representing the first portrait respectively, Characterize the learning vector The first dimension of the feature dimensional features, represents the feature dimension of the learning vector, is a positive integer, Characterizing the first vector The first dimension of the feature dimensional features, represents the dot product of the learning vector and the first vector, and respectively representing the modulus of the learning vector and the first vector; When the first similarity reaches a predetermined similarity limit, the first portrait is added to the matching portrait set.

8. The large model-driven smart education resource recommendation method according to claim 1, characterized in that: After introducing the agent debate strategy to compare and analyze the first resource requirement set and the second resource requirement set to generate a resource requirement list, and using the resource requirement list as a target recommendation list for the student user, the method further includes: activating a fourth agent among the plurality of agents; Collecting multi-source recommendation effect records of the target recommendation list through the fourth agent; Performing weighted analysis on the multi-source recommendation effect records to obtain the recommendation effect fitness; If the fitness of the recommendation effect does not reach the predetermined fitness limit, an optimization recommendation is performed on the first peer user of the student user.

9. The large model-driven smart education resource recommendation method according to claim 8, characterized in that: The multi-source recommendation effect record includes at least teacher evaluation effect, student evaluation effect and student written test effect.

10. A large-scale model-driven intelligent educational resource recommendation system, characterized by: The system is applied to the large model-driven smart education resource recommendation method according to any one of claims 1 to 9, and the system includes: an action point identification module, the action point identification module being used to form an educational resource set and perform action point identification on a first resource in the educational resource set by a first agent among the multiple agents, including extracting a knowledge point set and a skill point set embedded in the first agent to obtain a first action label set; an education resource pool establishment module, the education resource pool establishment module being configured to establish an education resource pool according to a correspondence between the first resource and the first action tag set; A multi-source data acquisition module, wherein the multi-source data acquisition module is used to acquire a multi-source database of student users, wherein the multi-source database includes the learning goals, learning records and interest preferences of the student users; a first resource demand analysis module, configured to perform a collaborative analysis of the learning records and the interest preferences using a second agent among the multiple agents in combination with the educational resource pool to obtain a first resource demand set; a second resource requirement analysis module, configured to perform a collaborative analysis of the learning objectives and the learning records using a third agent among the multiple agents in combination with the educational resource pool to obtain a second resource requirement set; a comparative analysis module, the comparative analysis module being configured to introduce an agent debate strategy to perform comparative analysis on the first resource requirement set and the second resource requirement set, generate a resource requirement list, and use the resource requirement list as a target recommendation list for the student user; Wherein, the comparative analysis module is further used for: The debate coordinator among multiple agents compares the resources in the first resource demand set and the second resource demand set one by one, and analyzes the frequency and importance of each resource in the two demand sets, as well as its match with students' learning goals and interests; Based on the frequency, importance, and matching degree with students' learning goals and interests, multiple other agents debate the necessity, importance, and applicability of each resource, generating their own evaluation opinions. The debate coordinator synthesizes the evaluation opinions of each intelligent agent to generate the resource demand list.

Images