A thought politics lesson knowledge point dynamic pushing method and system

Abstract

The application discloses a kind of ideological and political course knowledge point dynamic push method and system, including the following steps: the multidimensional learning data of student is collected, multidimensional learning data includes answer text data, discussion speech data and behavior selection data;Call pre-trained position detection model, the multi-source semantic analysis of student's answer text, discussion speech and behavior option is carried out, and the value orientation tendency of student is implicitly inferred.According to the ideological and political course knowledge point dynamic push method and system of the application embodiment, two-dimensional adaptation is suitable for the attribute of ideological and political education: the adaptation system of double dimension of cognitive hierarchy and value orientation is constructed, on the basis that corresponding difficulty learning content is matched according to the cognitive ability of student, corresponding depth guiding content is matched simultaneously combined with value identification state, the double personalized adaptation of ideological and political course knowledge learning and value leading is realized, and the pertinence of teaching guidance is improved.

Description

Technical Field

[0001] This application relates to the technical field of ideological and political education, and in particular to a method and system for dynamically pushing knowledge points in ideological and political courses. Background Technology

[0002] Currently, most of the personalized delivery solutions in online ideological and political education courses in colleges and universities only focus on students' knowledge mastery to classify them into different levels. They can only deliver learning content of different difficulty to different students, without taking into account that students' levels of identification with ideological and political content are different. They cannot provide tiered guidance based on ideological identification and have not fully played the role of ideological and political education courses in both teaching knowledge and guiding values.

[0003] While some programs aim to analyze students' values, they primarily rely on directly distributing questionnaires and conducting assessments to collect information. Firstly, students are prone to resistance; secondly, many students deliberately choose "standard answers," meaning the results don't reflect their true thoughts. Furthermore, this ideologically related scoring data is mostly stored and used in plain text with explicit scores, posing privacy risks of data breaches and easily labeling students with fixed ideological views, which contradicts the gradual and progressive guidance principle of ideological and political education. Summary of the Invention

[0004] This application aims to at least partially address one of the technical problems in the related art.

[0005] Therefore, one objective of this application is to provide a method and system for dynamically pushing knowledge points of ideological and political courses, which adapts to the attributes of ideological and political education in two dimensions: constructing an adaptation system with cognitive level and value orientation in two dimensions. On the basis of matching learning content of corresponding difficulty according to students' cognitive ability, it simultaneously combines value identification status to match guidance content of corresponding depth, realizing dual personalized adaptation of ideological and political course knowledge learning and value guidance, and improving the pertinence of teaching guidance.

[0006] To achieve the above objectives, the first aspect of this application proposes a method for dynamically pushing knowledge points in ideological and political education courses, comprising the following steps: Collect students’ multidimensional learning data, which includes answer text data, discussion speech data, and behavior selection data; By calling a pre-trained stance detection model, multi-source semantic analysis is performed on students' answer texts, discussion speeches, and behavioral options to implicitly infer students' value orientation tendencies and obtain a comprehensive value orientation score. The comprehensive value orientation score is mapped to several preset fuzzy value orientation levels and then output. Value orientation data is stored using homomorphic encryption technology. When participating in cognitive level calibration calculations, numerical calculations are completed within the ciphertext space. During decryption, only the adaptation level identifier is output, and the original scoring data is not output. A student cognitive model is constructed based on multidimensional learning data that includes value orientation dimensions, and the student's cognitive level is identified. Based on the defined cognitive levels and value orientation ambiguity levels, the adaptation of ideological and political education learning resources and the guidance for answering questions are driven.

[0007] According to an embodiment of this application, a method and system for dynamically pushing knowledge points in ideological and political education courses are proposed. The system adapts to the attributes of ideological and political education in two dimensions: it constructs an adaptation system with cognitive level and value orientation. On the basis of matching learning content of corresponding difficulty according to students' cognitive ability, it simultaneously combines the value recognition status to match guidance content of corresponding depth, thereby realizing the dual personalized adaptation of ideological and political education knowledge learning and value guidance, and improving the pertinence of teaching guidance.

[0008] In addition, the method and system for dynamically pushing knowledge points of ideological and political courses proposed above in this application may also have the following additional technical features: In one embodiment of this application, the multi-source semantic analysis includes three data streams: response text semantic analysis, discussion speech semantic analysis, and behavioral option preference analysis. After value orientation inference is performed independently on each of the three data streams, a weighted fusion is performed to obtain a comprehensive inference result. The weighted fusion formula is as follows:

[0009] in, for Value orientation rating at any moment To score points based on the inference of the answer text. Inferring scores based on discussion and presentation. Infer scores for behavioral options. , , The weights of the three data streams are respectively and satisfy the following conditions: , The updated weights for this observation, This is the overall score from the previous moment.

[0010] In one embodiment of this application, speaking weight is discussed. The settings are dynamically adjusted based on speech length and content quality factors. , Scale synchronously according to the initial weight ratio, keeping the sum of the weights equal to 1. The adjusted weights satisfy the following: , , for Adjust the weights; The weight of discussion speech The adjusted formula is as follows:

[0011] in, To establish the base values ​​for discussing the weight of each statement, The normalized speech length. For content quality factors, and This is the preset adjustment coefficient; The content quality factor

[0012] in, This refers to the number of terms in the speech text that match the pre-defined dictionary of core ideological and political values. The total number of words in the speech text; The preset ideological and political core value keyword dictionary is the same dictionary used by the stance detection model.

[0013] In one embodiment of this application, the homomorphic encryption technology employs the Paillier additive homomorphic encryption scheme, which converts value orientation data into an n-dimensional vector. This means that each dimension corresponds to a core ideological and political value dimension, and each component is encrypted and stored separately; value-oriented dimension label vectors are pre-configured for each topic in the ideological and political topic database. The value orientation dimension label vector corresponds one-to-one with the dimension of the value orientation vector; The encrypted value orientation vector and the value orientation dimension label vector of the question are matched in the ciphertext space. The matching degree is sorted and suitable questions are selected to generate a personalized answer sequence. Among them, the dimension with high recognition is selected first, and the critical thinking and analysis type questions are selected first, and the dimension with the need to improve recognition is selected first, and the case guidance type questions are selected first. The formula for calculating the matching degree of the ciphertext space is as follows:

[0014]

[0015] in, For students' value orientation vector, For the first The value orientation dimension label vector of the question. For Paillier encryption functions, The degree of matching in ciphertext form; After ranking the matching degree, critical thinking and analysis questions are selected first for dimensions with high recognition, and case-guided questions are selected first for dimensions with lower recognition.

[0016] In one embodiment of this application, the fuzzy levels include at least three levels: high level of agreement, medium level of agreement, and level of agreement needing improvement. The level division adopts a hysteresis threshold mechanism, and the current level is set as follows: The current rating is The threshold was raised to The threshold was lowered to The second upward adjustment threshold between high and low approval ratings. Second down-adjustment threshold ,and The current value orientation comprehensive score is ; when The level of recognition needs to be improved and At that time, the rating will be upgraded to "Medium" (meaning "medium approval rating"). when For moderate recognition and At that time, the rating will be upgraded to "high approval rating"; when For high recognition and At that time, the rating will be downgraded to "medium" in terms of approval rating. when For moderate recognition and At that time, the rating will be downgraded to "recognition needs improvement".

[0017] In one embodiment of this application, based on the fuzzy level of value orientation determined by the hysteresis threshold mechanism, corresponding types of learning content are selected and a guiding sequence is generated. High recognition corresponds to critical thinking and analytical questions and open-ended topics. The content includes case studies and explanations of principles related to the level of recognition. There is room for improvement in the availability of inspiring stories and role model examples.

[0018] In one embodiment of this application, a label reset step is also included, which automatically performs a reset operation on cognitive level labels, including value orientation dimensions, according to an adaptive reset cycle, restoring all cognitive labels to their initial unlabeled state. The adaptive reset cycle formula ,in, Based on the reset cycle, To reflect recent learning activity, For the rate of cognitive change, and The preset adjustment coefficient; The rate of cognitive change Based on comprehensive value orientation score The standard deviation within the preset time window is calculated. Recent learning activity The value is the normalized number of student learning behaviors within a preset time window, and its range is [value range missing]. ; The rate of cognitive change Based on comprehensive value orientation score The standard deviation is calculated within a preset time window, specifically by first calculating the sample standard deviation of the comprehensive value orientation score sequence within the time window, and then linearly normalizing the sample standard deviation to a preset maximum standard deviation threshold. The interval is used to obtain the rate of cognitive change. ; When the sample standard deviation is greater than or equal to the preset maximum standard deviation threshold The value is 1.

[0019] In one embodiment of this application, based on the student's cognitive level and the fuzzy level of value orientation determined by the hysteresis threshold mechanism, a two-dimensional joint search is performed in the ideological and political education resource database to calculate the comprehensive matching score of the resources. The formula for the comprehensive matching score is as follows:

[0020]

[0021] in, The score is based on the cognitive difficulty matching. Matching scores based on value orientation. For professional matching score, For interest and preference matching score, , , , For the corresponding weight and ; Personalized learning resource sequences are generated based on the overall matching score.

[0022] The second aspect of this application proposes a method and system for dynamically pushing knowledge points of ideological and political courses, including a data acquisition module for collecting students' multidimensional learning data, which includes answer text data, discussion speech data and behavior selection data. The value orientation analysis module is used to call the pre-trained stance detection model to perform multi-source semantic analysis on the multi-dimensional learning data, implicitly infer the students' value orientation tendencies, and obtain a comprehensive value orientation score. The rating mapping module is used to map the comprehensive value orientation score to a number of preset fuzzy value orientation ratings and output them. The encrypted storage module is used to store value orientation data using homomorphic encryption technology. When participating in cognitive level calibration calculations, it completes numerical calculations in the ciphertext space, and only outputs the level identifier when decrypting. The cognitive level labeling module is used to construct a student cognitive model based on multidimensional learning data that includes value orientation dimensions, and to label the student's cognitive level. The adaptation and guidance module is used to drive the adaptation of ideological and political education learning resources and answer guidance based on the calibrated cognitive level and value orientation ambiguity level.

[0023] The third aspect of this application proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any step of the dynamic push method for ideological and political education knowledge points proposed in the first aspect.

[0024] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0025] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating a method and system for dynamically pushing knowledge points in ideological and political education courses according to an embodiment of this application. Figure 2 This is a system block diagram of a method and system for dynamically pushing knowledge points in ideological and political education courses according to another embodiment of this application. Detailed Implementation

[0026] Embodiments of this application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. Rather, embodiments of this application include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0027] The following is in conjunction with the appendix Figure 1 This application describes a method for dynamically pushing knowledge points in ideological and political education, comprising the following steps: Collect students’ multidimensional learning data, which includes answer text data, discussion speech data, and behavior selection data; By calling a pre-trained stance detection model, multi-source semantic analysis is performed on students' answer texts, discussion speeches, and behavioral options to implicitly infer students' value orientation tendencies and obtain a comprehensive value orientation score. The comprehensive value orientation score is mapped to several preset fuzzy value orientation levels and then output. Value orientation data is stored using homomorphic encryption technology. When participating in cognitive level calibration calculations, numerical calculations are completed within the ciphertext space. During decryption, only the adaptation level identifier is output, and the original scoring data is not output. A student cognitive model is constructed based on multidimensional learning data that includes value orientation dimensions, and the student's cognitive level is identified. Based on the defined cognitive levels and value orientation ambiguity levels, the adaptation of ideological and political education learning resources and the guidance for answering questions are driven.

[0028] The implementation process of this method will be explained in detail below, taking into account specific implementation scenarios.

[0029] This embodiment of the method for dynamically pushing knowledge points in ideological and political education courses is deployed on an online teaching platform for ideological and political education courses in universities and is applicable to undergraduate courses in ideological and political theory. The method relies on students' daily learning behavior data and does not require students to participate in specific assessments. The specific implementation steps are as follows: Step 1: Collect multidimensional learning data from students.

[0030] The platform's data collection components collect three types of data in real time during the student learning process, serving as data sources for subsequent analysis: first, written answers, including student responses to subjective questions, material analysis questions, and essay questions; second, discussion and speaking data, including posts on the course forum, comments in group discussions, and classroom interaction comments; and third, behavioral choice data, including choices made in answering objective questions, access to extended resources, and learning path selections. Data collection follows the principle of minimum necessity, limiting the scope to data required for teaching, and obtaining informed consent from students before collection.

[0031] Step 2: Multi-source semantic analysis to implicitly infer students' value orientation tendencies.

[0032] A pre-trained ideological and political education stance detection model is invoked to perform semantic analysis on three types of data to infer students' value orientation tendencies. In this embodiment, the stance detection model is based on a pre-trained Chinese language model and fine-tuned using labeled student response data. It can output tendency scores for multiple value dimensions such as collectivism, patriotism, social responsibility, and rule of law awareness. The single-dimensional score ranges from -1 to 1, with positive values ​​corresponding to positive tendencies and negative values ​​corresponding to negative tendencies. For the two types of text data, namely response texts and discussion speeches, the semantic inference scores for each dimension are directly output by the model. For behavioral choice data, the tendency scores for the corresponding dimension are converted through preset option-value dimension mapping rules. After weighted fusion of the three types of data, a comprehensive inference result of students' value orientation is obtained. The inference process is completed automatically in the background without the students' awareness.

[0033] Step 3: Map the inference results to several preset fuzzy levels and output them.

[0034] The comprehensive inference result of value orientation is mapped to a preset fuzzy level and output. This embodiment sets three fuzzy levels: recognition level needs improvement, recognition level is medium, and recognition level is high. The level determination adopts a hysteresis threshold mechanism: adjacent levels are respectively set with upward and downward thresholds, and the upward threshold is higher than the downward threshold; when the comprehensive score continuously reaches the corresponding threshold, the student's value orientation level is changed. The system only outputs the level identifier and does not output the precise score.

[0035] Step 4: Homomorphic encryption storage and ciphertext operation of value orientation data.

[0036] Value orientation data is stored using homomorphic encryption. When participating in cognitive level calibration calculations, numerical calculations are performed within the ciphertext space, and decryption only outputs the adaptation level identifier. This embodiment uses the Paillier additive homomorphic encryption scheme, and the specific implementation is as follows: Encrypted storage: After the students' multi-dimensional value orientation score vectors are integerized, they are encrypted component by component using a public key and stored in an independent encrypted database, which is physically isolated from students' identity information and business learning data; the decryption private key is stored and managed by multiple parties, and no single party can complete the decryption independently.

[0037] Ciphertext computation: When value orientation data is used for cognitive level labeling, resource matching and other calculations, the calculation is completed directly in the ciphertext space without decrypting the original data; relying on the additive homomorphic property of the Paillier algorithm, operations such as dot product and weighted summation can be performed.

[0038] Output results: After the calculation is completed, only the final adaptation level result is decrypted and output; the original value orientation score data is not returned.

[0039] Step 5: Construct a student cognitive model that includes a value orientation dimension and label the cognitive levels.

[0040] Based on collected multidimensional learning data, a student cognitive model including a value orientation dimension is constructed, and cognitive levels are labeled. This embodiment's cognitive model includes four dimensions: cognitive foundation, interest preference, value orientation, and learning resilience, each with preset weight coefficients. Specifically, the cognitive foundation dimension is calculated based on knowledge point mastery and answer accuracy; the interest preference dimension is obtained based on statistics of student resource access behavior; the value orientation dimension uses encrypted level identifiers in the calculation, without incorporating raw scores; and the learning resilience dimension is calculated based on the time spent answering difficult questions and the completion rate of revisiting incorrect questions. Based on the comprehensive score of the cognitive model, students' cognitive levels are labeled into three levels: basic, developmental, and advanced, corresponding to different levels of knowledge mastery and thinking ability, serving as the basis for subsequent resource delivery and answer guidance.

[0041] Step 6: Adapt learning resources for ideological and political education courses and guide students in answering questions.

[0042] Based on the identified cognitive levels and value orientation ambiguity levels, corresponding ideological and political education learning resources are matched to generate personalized answer paths. Specifically, this includes: Learning resource matching: Knowledge point explanations and case studies of corresponding difficulty are matched according to cognitive level, and guiding resources of corresponding types are matched according to value orientation level. For example, students at the basic level who need to improve their understanding will be given priority access to basic concept explanations and typical case studies; students at the advanced level who have high understanding will be given priority access to current affairs topics and extended learning resources.

[0043] Answer Path Guidance: The initial difficulty of the questions is determined based on the student's cognitive level. Basic level students start with memorization and comprehension questions, gradually increasing the difficulty; advanced level students begin with application and analysis questions, supplemented by extension questions. The question type emphasis is also adjusted based on the student's value orientation level: dimensions with high recognition are primarily addressed with critical thinking and analytical questions, while dimensions requiring improvement are primarily addressed with case-guided questions. During the question-and-answer process, the system can trigger supplementary learning of prior knowledge or provide tiered prompts based on the student's responses.

[0044] In one embodiment of this application, multi-source semantic analysis includes three data streams: semantic analysis of response text, semantic analysis of discussion speech, and behavioral option preference analysis. After value orientation inference is performed independently on each of the three data streams, a weighted fusion is performed to obtain a comprehensive inference result. The weighted fusion formula is as follows:

[0045] in, for Value orientation rating at any moment To score points based on the inference of the answer text. Inferring scores based on discussion and presentation. Infer scores for behavioral options. , , The weights of the three data streams are respectively and satisfy the following conditions: , The updated weights for this observation, This is the overall score from the previous moment.

[0046] This embodiment further discusses the speaking weight. The settings are dynamically adjusted based on speech length and content quality factors. , Scale synchronously according to the initial weight ratio, keeping the sum of the weights equal to 1. The adjusted weights satisfy the following: , , for Adjust the weights; Discussion of speaking weight The adjusted formula is as follows:

[0047] in, To establish the base values ​​for discussing the weight of each statement, The normalized speech length. For content quality factors, and This is the preset adjustment coefficient; Content quality factor

[0048] in, This refers to the number of terms in the speech text that match the pre-defined dictionary of core ideological and political values. The total number of words in the speech text; The keyword dictionary for core ideological and political values ​​is the same as the keyword dictionary used by the stance detection model.

[0049] Specifically, the triggering scenarios for weighted fusion calculation based on overall value orientation.

[0050] When students exhibit the following three types of effective learning behaviors on the platform, the system will trigger value orientation inference from the corresponding data source and execute the aforementioned weighted fusion formula to update the student's comprehensive value orientation score: When submitting text-based answers: After students complete and submit assignments or quizzes with text-based answers, such as subjective questions, material analysis questions, and essay questions, the system performs semantic inference on the text and generates a score for that answer. Substitute the values ​​into the formula to update the overall score.

[0051] When a student posts a valid discussion message: After a student posts a public message in the course forum, group discussion area, or classroom interaction section, and the content is verified as valid, the system performs semantic inference on the message and generates a score for that discussion message. Simultaneously, the speaking weight is dynamically adjusted, and the overall score is updated by substituting it into the formula.

[0052] When a student makes a behavioral selection action: After completing actions such as answering objective questions, selecting designated extended learning resources, or choosing a self-selected answer path, the system generates a behavioral option score based on the selection result. Substitute the values ​​into the formula to update the overall score.

[0053] Substituting the corresponding score obtained from the current observation into the exponentially weighted moving average fusion formula, the comprehensive score at time t is calculated. The formula is:

[0054] To control system overhead, consecutive similar operations within the same learning unit are merged into a single observation cycle for unified calculation and update; single operation data that is meaningless or has no valid content is not included in the scoring calculation and does not trigger a scoring update.

[0055] It should be noted that the weight of discussion speeches... Dynamically adjusted trigger conditions: This will only be executed if the newly added data in this observation is discussion or speech data. The dynamic adjustment calculation is as follows: If only response text and action option data are generated this time, The current base value remains unchanged, and the adjustment logic is not triggered. The adjusted weight is subject to a value range constraint, as described in this embodiment. The adjusted range is limited to To avoid giving an individual message too much or too little weight, and to ensure the overall balance of the three weighting methods, weight adjustments are only effective for the current observation. In the next update of non-message data, [the weighting will be adjusted accordingly]. Regression to the baseline value.

[0056] This embodiment does not rely on a single homework answer to judge students' ideological and political identification level. Instead, it uses a variety of behavioral data collected by the platform from students' learning process, analyzing several different forms such as written answers, public speeches, and self-selected answers together, and finally integrates them to calculate a comprehensive value orientation score.

[0057] The reason for using so many dimensions is to avoid inaccurate results from relying on only one type of data. For example, students might write formulaic answers to subjective questions, which may not reflect their true thoughts; focusing solely on discussion comments is too casual and inconsistent in quality. Cross-validating data from multiple different formats ensures that the results are closer to students' actual understanding and are more reliable.

[0058] In one embodiment of this application, the homomorphic encryption technology employs the Paillier additive homomorphic encryption scheme, which converts value orientation data into an n-dimensional vector. This means that each dimension corresponds to a core ideological and political value dimension, and each component is encrypted and stored separately; value-oriented dimension label vectors are pre-configured for each topic in the ideological and political topic database. The value orientation dimension label vector and the value orientation vector have a one-to-one correspondence in dimensions; The encrypted value orientation vector and the value orientation dimension label vector of the question are matched in the ciphertext space. The matching degree is sorted and suitable questions are selected to generate a personalized answer sequence. Among them, the dimension with high recognition is selected first, and the critical thinking and analysis type questions are selected first, and the dimension with the need to improve recognition is selected first, and the case guidance type questions are selected first. The formula for calculating the matching degree of the ciphertext space is as follows:

[0059]

[0060] in, For students' value orientation vector, For the first The value orientation dimension label vector of the question. For Paillier encryption functions, The degree of matching in ciphertext form; After ranking the matching degree, critical thinking and analysis questions are selected first for dimensions with high recognition, and case-guided questions are selected first for dimensions with lower recognition.

[0061] The core of ideological and political education is to cultivate people. The worst thing to do is to label students with fixed tags such as "low ideological awareness". If the value orientation score is written down and can be seen by the teachers and administrators, it is easy to form stereotypes and even affect students' evaluation for honors, party membership and comprehensive quality evaluation, which goes against the original intention of ideological and political education.

[0062] This encrypted computing mechanism essentially locks access to the original data from the technical level, outputting only the appropriate questions and level labels throughout the process. No one can obtain the precise original score, thus fundamentally avoiding the ethical issues of "using data to score and label students' thoughts," which is something that ordinary access control cannot achieve.

[0063] The specific implementation process is as follows: 1. Value-oriented data is stored in encrypted form.

[0064] Based on the calculation results of the scores for each dimension of value orientation, the students' value orientation data is organized into an n-dimensional vector form to complete the data structuring. The vector expression is as follows: in The total number of core values ​​of ideological and political education These are the plaintext scores for the corresponding dimensions.

[0065] The Paillier additive homomorphic encryption scheme is used to encrypt each component of the vector component by component using the system's public key, resulting in the fully ciphertext value orientation vector. ,in This uses the Paillier encryption function. After encryption, the ciphertext vector is stored in a separate encrypted database, physically isolated from student identity information and regular learning business data. The decryption private key is managed using a multi-party storage mechanism, preventing a single entity from independently completing decryption. The original plaintext score generated is not retained; only the ciphertext data is preserved.

[0066] 2. Pre-configuration of value-oriented tags for the question bank (preliminary preparation step).

[0067] When adding ideological and political education questions to the question bank, a value orientation dimension label vector is simultaneously assigned to each question, corresponding one-to-one with the student's value orientation vector dimension. The tag vector expression for this question is: in These represent the guidance emphasis coefficients for the corresponding value dimension of the question. The question tag vector is a publicly configured attribute, stored in plaintext in the question bank, and does not require encryption.

[0068] 3. Calculation of ciphertext spatial matching degree.

[0069] When it is necessary to generate personalized answer sequences for students, the encrypted value orientation vector of the student and the plaintext label vector of the candidate questions are retrieved. Relying on the additive homomorphic property of the Paillier algorithm, the dot product matching degree of the two is directly calculated in the encrypted space. The matching degree calculation formula is as follows: In the formula, The matching results are in encrypted form. The entire calculation process does not decrypt the student's original value score; all numerical calculations are completed in encrypted form, and there is no path for the original score to be leaked during the calculation process.

[0070] 4. Adapt to question filtering and result output.

[0071] Candidate questions are relatively ranked based on the encrypted matching results, and questions with higher matching scores are selected. Simultaneously, the question type configuration is adjusted based on the student's fuzzy level across various dimensions: critical thinking and analytical questions are prioritized for dimensions with high acceptance, while case-guided questions are prioritized for dimensions where acceptance needs improvement. Ultimately, only personalized answer sequences and fuzzy level labels are output to the teaching front-end; no original precise scores with any value orientation are returned, ensuring that the student's specific scoring will not be revealed during application.

[0072] In one embodiment of this application, the fuzzy level includes at least three levels: high level of agreement, medium level of agreement, and level of agreement needing improvement. The level division adopts a hysteresis threshold mechanism, and the current level is set as follows: The current rating is The threshold was raised to The threshold was lowered to The second upward adjustment threshold between high and low approval ratings. Second down-adjustment threshold ,and The current value orientation comprehensive score is ; when The level of recognition needs to be improved and At that time, the rating will be upgraded to "Medium" (meaning "medium approval rating"). when For moderate recognition and At that time, the rating will be upgraded to "high approval rating"; when For high recognition and At that time, the rating will be downgraded to "medium" in terms of approval rating. when For moderate recognition and At that time, the rating will be downgraded to "recognition needs improvement".

[0073] This embodiment uses a value orientation fuzzy level mapping step to desensitize and weaken the sense of labeling: the precise numerical score is converted into three fuzzy levels: "high recognition, medium recognition, and recognition needs improvement". Subsequently, only the fuzzy level is used to match the corresponding question type. The precise score is not exposed to the outside world throughout the process, which further weakens the problem of "giving students precise scores and attaching fixed labels", which is in line with the educational attribute of ideological and political education.

[0074] To avoid frequent fluctuations in grades: A hysteresis threshold mechanism is adopted, which sets different score lines for grade adjustments. This prevents grades from jumping back and forth repeatedly when the overall score fluctuates slightly around the critical value, ensuring stable grade results that conform to the gradual changes in students' value perception.

[0075] 1. Level threshold settings.

[0076] Four score lines are set, corresponding to the upward and downward adjustment standards for two grade transitions. The four score lines, from highest to lowest, are: upward adjustment line for students moving from intermediate to advanced, downward adjustment line for students moving from advanced to intermediate, upward adjustment line for students needing to be upgraded to intermediate, and downward adjustment line for students moving from intermediate to needing to be upgraded. Within the same grade transition, the upward adjustment score line is always higher than the downward adjustment score line, forming a buffer zone between the two lines: when a student's overall score falls within this buffer zone, the original grade remains unchanged, without any adjustment.

[0077] 2. Rules for determining level updates.

[0078] The grade update not only considers the current overall score, but also takes into account the student's previous grade. The specific determination rules are as follows: If the original level is "Approval rating needs improvement": the level will be upgraded to "Approval rating is medium" only when the current overall score reaches the upper limit of the "Need to improve" level; otherwise, the original level will remain unchanged.

[0079] If the original level is "Medium approval rating": it will only be upgraded to "High approval rating" when the current comprehensive score reaches the upward adjustment line of "Medium to Upgrade"; it will only be downgraded to "Approval rating needs improvement" when the current comprehensive score falls to the downward adjustment line of "Medium to Downgrade". Otherwise, the original level will remain unchanged.

[0080] If the original level is "High level of recognition": it will only be downgraded to "Medium level of recognition" if the current overall score reaches the lower limit of the "High to Medium" downgrade; otherwise, the original level will remain unchanged.

[0081] To make it easier to understand, let's take a more intuitive numerical example: Suppose we set four score lines as follows: 65 points are needed to improve to the middle level, 60 points are needed to improve from the middle level; 85 points are needed to improve from the middle level to the high level, and 80 points are needed to improve from the high level to the middle level.

[0082] The student was originally "recognition needs improvement". The score increased from 61 to 64, but it did not reach the standard of 65 for promotion. The grade remained unchanged and the student was not promoted just because they passed the passing grade. After a student is upgraded to "medium recognition", their score drops from 66 to 62, but does not fall below the 60-point downgrade standard. The grade remains unchanged and will not fluctuate repeatedly due to small score fluctuations. The 60-65 and 80-85 score ranges serve as buffer zones, completely preventing scores from fluctuating around the critical values ​​and causing frequent changes in grades.

[0083] The final output fuzzy level is the core basis for generating personalized answer sequences—dimensions with high consensus are prioritized for critical thinking and analytical questions, while dimensions with lower consensus are prioritized for case-guided questions. Meanwhile, the precise comprehensive score is stored and processed only in an encrypted backend environment, with the frontend only displaying the fuzzy level, further strengthening data privacy and educational ethics protection at the output level.

[0084] As one possible implementation, based on the student's cognitive level and the fuzzy level of value orientation determined by the hysteresis threshold mechanism, a two-dimensional joint search is performed in the ideological and political education resource database to calculate the comprehensive matching score. The formula for the comprehensive matching score is...

[0085]

[0086] in, The score is based on the cognitive difficulty matching. Matching scores based on value orientation. For professional matching score, For interest and preference matching score, , , , For the corresponding weight and ; Personalized learning resource sequences are generated based on the overall matching score.

[0087] The same learning resources for ideological and political education courses cannot be recommended to everyone. This step involves taking into account students' learning abilities, ideological identification, professional background, and personal learning preferences to calculate the suitability of each resource for each student. The most suitable resources are then recommended to students to ensure that they can understand and are willing to learn, while also achieving the corresponding ideological and political guidance.

[0088] Formula for calculating the total score of resource suitability.

[0089] This formula calculates a "suitability score" for each learning resource. The higher the score, the more suitable it is for the student. Let's break down the meaning of each item in layman's terms: Cognitive difficulty matching score): corresponding The weighting refers to "whether the difficulty of the resource matches the student's current cognitive level." If the student's cognitive level is high, then resources for advanced critical thinking and in-depth expansion will be given high marks; if the student's foundation is weak, then resources for basic explanation and concept popularization will be given high marks, avoiding resources that are too difficult to understand or too simple to learn anything.

[0090] (Value Orientation Matching Score): Corresponding to The weighting refers to "whether the ideological and political guidance direction of this resource matches the student's current level of value identification." For example, if a student's identification level in a certain dimension needs improvement, resources with case explanations and story-based guidance will be given high scores; if a student's identification level is already high, resources with dialectical analysis and in-depth exploration will be given high scores, thus achieving tiered guidance.

[0091] (Professional fit score): Corresponding The weighting is based on "how relevant the resource is to the student's major." For example, providing ideological and political education content that combines engineering cases to engineering students, and providing materials related to the benevolence of doctors to medical students. The closer the resource is to the major, the higher the score, and the more students will feel engaged with it.

[0092] (Interest and Preference Matching Score): Corresponding Weighting refers to "whether the format and theme of this resource match the students' learning preferences." For example, if students usually like watching videos, micro-lessons and short video resources will be given high scores; if students like reading text, articles and documents will be given high scores, mainly to enhance their learning motivation.

[0093] Weight setting logic.

[0094] The weights of the four dimensions satisfy The priorities are arranged from highest to lowest: First, "Can the difficulty level be kept up?" is the most important, as it is a course and must match the students' learning abilities; second, "Is the value guidance correct?" is the core educational goal of ideological and political education; next, "Is it related to the student's major?" is a supplementary factor to enhance the learning experience; and finally, "Does the student like the format?" is a supplementary factor to optimize the learning experience. This arrangement is to avoid putting the cart before the horse, ensuring that resource recommendations first meet the core needs of teaching and education, and then consider the learning experience.

[0095] In actual operation, the system first uses the student's cognitive level and value orientation fuzzy level to filter the resource library, removing resources that are obviously unsuitable in difficulty or orientation, thus narrowing the calculation scope. The remaining resources are then used to calculate the comprehensive suitability score using the formula above, and finally sorted from high to low scores to generate a personalized learning resource sequence to be pushed to the student.

[0096] As one possible implementation of this embodiment, based on the fuzzy level of value orientation determined by the hysteresis threshold mechanism, corresponding types of learning content are selected and a guiding sequence is generated. High recognition corresponds to critical thinking and analytical questions and open-ended topics. The content includes case studies and explanations of principles related to the level of recognition. There is room for improvement in the availability of inspiring stories and role model examples.

[0097] Specifically, the hierarchical adaptation rules are as follows: Highly recognized: Matches critical thinking and analytical questions with open-ended topics, guiding students to engage in in-depth dialectical thinking and further deepen their understanding.

[0098] Medium level of recognition: Matching case studies and principle explanations helps students consolidate their understanding and solidify their value perception.

[0099] Awareness needs improvement: Match inspiring stories and role model examples with empathetic guidance to gradually build value consensus.

[0100] Meanwhile, because the level determination adopts a delayed threshold mechanism, the level status is stable, and the corresponding content type will not switch frequently, thus maintaining a smooth learning guidance pace.

[0101] In one embodiment of this application, a label reset step is also included, which automatically performs a reset operation on cognitive level labels, including value orientation dimensions, according to an adaptive reset cycle, restoring all cognitive labels to their initial unlabeled state. Adaptive reset cycle formula ,in, Based on the reset cycle, To reflect recent learning activity, For the rate of cognitive change, and The preset adjustment coefficient; Cognitive change rate Based on comprehensive value orientation score The standard deviation within the preset time window is calculated. Recent learning activity The value is the normalized number of student learning behaviors within a preset time window, and its range is [value range missing]. ; Cognitive change rate Based on comprehensive value orientation score The standard deviation is calculated within a preset time window, specifically by first calculating the sample standard deviation of the comprehensive value orientation score sequence within the time window, and then linearly normalizing the sample standard deviation to a preset maximum standard deviation threshold. The interval is used to obtain the rate of cognitive change. ;

[0102] When the sample standard deviation is greater than or equal to the preset maximum standard deviation threshold The value is 1.

[0103] Specifically, students' cognitive abilities and value identification status are constantly evolving. Regularly resetting and relabeling all labels can ensure that the labels always match the students' current real status, avoiding the long-term lag of old labels and the resulting mismatch between learning resources and questions.

[0104] Abandoning the rigid model of fixed cycles, the reset interval is dynamically adjusted according to the students' learning activity level and the speed of cognitive change—students who learn frequently and whose cognition fluctuates greatly have a shorter reset cycle to ensure accuracy; students who learn infrequently and whose cognition is stable have a longer cycle to reduce unnecessary system computing overhead.

[0105] The formula for calculating the reset period is:

[0106] The base reset cycle, also known as the system's default reset interval (e.g., 30 days), is the baseline value for cycle calculation.

[0107] Recent learning activity level, with a fixed range of values. The total number of effective learning behaviors of students within a preset time window is counted and then linearly normalized to obtain this value; the more frequently students learn and the more effective behaviors they exhibit, the closer R approaches 1.

[0108] Cognitive change rate, with a fixed range of values. This is used to quantify the fluctuation range of students' value orientation cognition.

[0109] , : Preset adjustment coefficients to control the impact of learning activity and cognitive change rate on the reset cycle. The values ​​can be adjusted according to the needs of the actual teaching scenario.

[0110] Generally speaking, the higher the student's learning activity and the faster their cognitive changes, the larger the value of the denominator in the formula, and the longer the final reset cycle. The shorter the interval, the more frequently the labels are reset; conversely, the longer the interval, the better the system's efficiency while ensuring label accuracy.

[0111] When the system runtime reaches the currently calculated adaptive reset cycle, a reset operation is automatically triggered: all previously labeled cognitive labels, such as student cognitive level tags and value orientation dimension levels, are restored to their initial unlabeled state. After the reset is complete, the system will re-perform cognitive level labeling, value orientation score calculation, and level determination based on a new round of learning behavior data, completing the iterative update of labels.

[0112] This application discloses a dynamic push system for ideological and political education knowledge points, including a data acquisition module for collecting students' multidimensional learning data, which includes answer text data, discussion and speech data, and behavior selection data. The value orientation analysis module is used to call the pre-trained stance detection model to perform multi-source semantic analysis on multi-dimensional learning data, implicitly infer students' value orientation tendencies, and obtain a comprehensive value orientation score. The rating mapping module is used to map the comprehensive value orientation score to several preset fuzzy value orientation ratings and output them. The encrypted storage module is used to store value orientation data using homomorphic encryption technology. When participating in cognitive level calibration calculations, it completes numerical calculations in the ciphertext space, and only outputs the level identifier when decrypting. The cognitive level labeling module is used to construct a student cognitive model based on multidimensional learning data that includes value orientation dimensions, and to label the student's cognitive level. The adaptation and guidance module is used to drive the adaptation of ideological and political education learning resources and answer guidance based on the calibrated cognitive level and value orientation ambiguity level.

[0113] This application is a personalized learning guidance system for ideological and political courses. Its core logic is: instead of relying on direct ideological surveys, it implicitly analyzes students' value identification status from their daily learning behavior, while marking students' cognitive ability levels. Based on this, it dynamically pushes appropriate knowledge points and learning resources and plans personalized answer paths. Sensitive data is protected through encryption mechanisms throughout the process to avoid the ethical issues of "giving students precise scores and fixed labels for their thoughts".

[0114] Functional descriptions of each module.

[0115] The data acquisition module is the system's data entry point, responsible for collecting three types of behavioral data naturally generated by students during the course, which form the foundational material for the entire analysis. These three types of data correspond to three different scenarios: written responses, public speaking, and self-selection. The aim is to improve the authenticity of subsequent analysis results by cross-validating multi-dimensional data, preventing the deliberate manipulation of single data points.

[0116] The value orientation analysis module is the core analysis component of the system. It calls a pre-trained stance detection algorithm model to perform semantic analysis and interpretation on the collected multi-source learning data. Without directly asking students for their opinions, it can implicitly infer students' value identification tendencies and finally calculate a comprehensive score of students' value orientation.

[0117] The rating mapping module is responsible for converting precise numerical scores into vague rating labels such as "high recognition," "medium recognition," and "recognition needs improvement," and employs a hysteresis threshold mechanism to prevent frequent rating fluctuations. On the one hand, this weakens the labeling effect of precise scores, aligning with the attributes of ideological and political education; on the other hand, vague ratings are more suitable for direct use in subsequent content adaptation, without needing to expose specific scores.

[0118] The encrypted storage module is responsible for the secure storage and computation of value-sensitive data. It adopts homomorphic encryption technology: the original scoring data is stored after encryption, and when it is used in cognitive level calculations, the entire calculation is completed in encrypted state without decrypting the original data. When the data is finally decrypted and output, only a fuzzy level identifier is output. No role can obtain the student's precise original score, thus protecting data privacy and educational ethics from the technical level.

[0119] The cognitive level labeling module is responsible for building a complete student cognitive model. It not only considers the value orientation dimension but also combines multi-dimensional learning data to comprehensively label the student's cognitive ability level. Together with the fuzzy value orientation level, it constitutes the two core bases for subsequent content adaptation, achieving a two-dimensional matching of "ability difficulty adaptation + value guidance adaptation".

[0120] The adaptation and guidance module is the final output execution stage of the system. Based on the previously defined cognitive level and value orientation ambiguity level, it selects and matches corresponding learning resources and question types from the ideological and political education resource library, and generates personalized knowledge point push sequences and answer guidance paths, so that students with different foundations and different levels of identification can receive learning content that is adapted to their own situation.

[0121] This application discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any step of a method for dynamically pushing knowledge points in ideological and political education courses.

[0122] Overall, this personalized learning system for ideological and political education courses operates automatically based on students' daily learning data, requiring no additional assessments. The overall process is simplified as follows: Data collection: Automatically collects three types of behavioral data from students: their answers to subjective questions, their participation in discussions, and their learning choices. Meaningless content is filtered out, and similar consecutive operations are merged.

[0123] Analysis and scoring: A dedicated model is used to implicitly analyze three types of data to calculate a comprehensive value orientation score; the score is updated in a way that "new data accounts for a small part and historical scores account for a large part" to avoid sudden changes in scores due to a single action, and the weight of discussion comments will also be dynamically adjusted according to the quality of the content.

[0124] Level Conversion: Convert precise scores into three levels of fuzzy labels: "Agreement Needs Improvement," "Medium," and "High." Use a buffer threshold to prevent the level from jumping back and forth when the score fluctuates slightly. Only the level is output throughout the process, and the specific score is not displayed.

[0125] Encryption protection: The original value orientation data is stored using homomorphic encryption, and all subsequent calculations are completed in an encrypted state. In the end, only the output level result is decrypted, and no one can obtain the original precise score.

[0126] Cognitive levels are defined by combining four dimensions: learning foundation, interest, value orientation, and learning resilience. Students' cognitive abilities are defined as three levels: basic level, developmental level, and improvement level.

[0127] Personalized adaptation: Based on "cognitive level + value level", it matches learning resources and questions of corresponding difficulty and type, generates exclusive learning paths and answer content, and provides guidance of different depths for different levels.

[0128] Dynamic recalibration: The label recalibration cycle is automatically adjusted based on students' learning activity and the speed of cognitive change; after the expiration, the old labels are cleared and recalibrated with new learning data to ensure that the results always reflect the students' true state.

[0129] In summary, the method and system for dynamically pushing knowledge points in ideological and political education according to the embodiments of this application are adapted to the attributes of ideological and political education in two dimensions: constructing an adaptation system with cognitive level and value orientation in two dimensions, matching learning content of corresponding difficulty according to students' cognitive ability, and simultaneously matching guidance content of corresponding depth according to value identification status, realizing dual personalized adaptation of ideological and political education knowledge learning and value guidance, and improving the pertinence of teaching guidance.

[0130] In the description of this specification, the use of terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refers to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0131] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for dynamically pushing knowledge points in ideological and political education courses, characterized in that, Includes the following steps: Collect students’ multidimensional learning data, which includes answer text data, discussion speech data, and behavior selection data; By calling a pre-trained stance detection model, multi-source semantic analysis is performed on students' answer texts, discussion speeches, and behavioral options to implicitly infer students' value orientation tendencies and obtain a comprehensive value orientation score. The comprehensive value orientation score is mapped to several preset fuzzy value orientation levels and then output. Value orientation data is stored using homomorphic encryption technology. When participating in cognitive level calibration calculations, numerical calculations are completed within the ciphertext space. During decryption, only the adaptation level identifier is output, and the original scoring data is not output. A student cognitive model is constructed based on multidimensional learning data that includes value orientation dimensions, and the student's cognitive level is identified. Based on the defined cognitive levels and value orientation ambiguity levels, the adaptation of ideological and political education learning resources and the guidance for answering questions are driven.

2. The method according to claim 1, characterized in that, The multi-source semantic analysis includes three data sources: semantic analysis of response texts, semantic analysis of discussion statements, and behavioral option preference analysis. After value orientation inference is performed independently on each of the three data sources, a weighted fusion is performed to obtain a comprehensive inference result. The weighted fusion formula is as follows: in, for Value orientation rating at any moment To score points based on the inference of the answer text. Inferring scores based on discussion and presentation. Infer scores for behavioral options. , , The weights of the three data streams are respectively and satisfy the following conditions: , The updated weights for this observation, This is the overall score from the previous moment.

3. The method according to claim 2, characterized in that, Discussion of speaking weight The settings are dynamically adjusted based on speech length and content quality factors. , Scale synchronously according to the initial weight ratio, keeping the sum of the weights equal to 1. The adjusted weights satisfy the following: , , for Adjust the weights; The weight of discussion speech The adjusted formula is as follows: in, To establish the base values ​​for discussing the weight of each statement, The normalized speech length. For content quality factors, and This is the preset adjustment coefficient; The content quality factor in, This refers to the number of terms in the speech text that match the pre-defined dictionary of core ideological and political values. The total number of words in the speech text; The preset ideological and political core value keyword dictionary is the same dictionary used by the stance detection model.

4. The method according to claim 1, characterized in that, The homomorphic encryption technology employs the Paillier additive homomorphic encryption scheme, which transforms value orientation data into an n-dimensional vector. This means that each dimension corresponds to a core ideological and political value dimension, and each component is encrypted and stored separately; value-oriented dimension label vectors are pre-configured for each topic in the ideological and political topic database. The value orientation dimension label vector corresponds one-to-one with the dimension of the value orientation vector; The encrypted value orientation vector and the value orientation dimension label vector of the question are matched in the ciphertext space. The matching degree is sorted and suitable questions are selected to generate a personalized answer sequence. Among them, the dimension with high recognition is selected first, and the critical thinking and analysis type questions are selected first, and the dimension with the need to improve recognition is selected first, and the case guidance type questions are selected first. The formula for calculating the matching degree of the ciphertext space is as follows: in, For students' value orientation vector, For the first The value orientation dimension label vector of the question. For Paillier encryption functions, The degree of matching in ciphertext form; After ranking the matching degree, critical thinking and analysis questions are selected first for dimensions with high recognition, and case-guided questions are selected first for dimensions with lower recognition.

5. The method according to claim 1, characterized in that, The fuzzy levels include at least three levels: high level of agreement, medium level of agreement, and level of agreement needing improvement. The level division adopts a hysteresis threshold mechanism, assuming the current level is [value missing]. The current rating is The threshold was raised to The threshold was lowered to The second upward adjustment threshold between high and low approval ratings. Second down-adjustment threshold ,and The current value orientation comprehensive score is ; when The level of recognition needs to be improved and At that time, the rating will be upgraded to "Medium" (meaning "medium approval rating"). when For moderate recognition and At that time, the rating will be upgraded to "high approval rating"; when For high recognition and At that time, the rating will be downgraded to "medium" in terms of approval rating. when For moderate recognition and At that time, the rating will be downgraded to "recognition needs improvement".

6. The method according to claim 5, characterized in that, Based on the fuzziness level of value orientation determined by the hysteresis threshold mechanism, corresponding types of learning content are selected and a guiding sequence is generated. High recognition corresponds to critical thinking and analytical questions and open-ended topics. The content includes case studies and explanations of principles related to the level of recognition. There is room for improvement in the availability of inspiring stories and role model examples.

7. The method according to claim 1, characterized in that, It also includes a label reset step, which automatically performs the reset operation of cognitive level labels, including value orientation dimensions, according to the adaptive reset cycle, restoring all cognitive labels to the initial unlabeled state; The adaptive reset cycle formula ,in, Based on the reset cycle, To reflect recent learning activity, For the rate of cognitive change, and The preset adjustment coefficient; The rate of cognitive change Based on comprehensive value orientation score The standard deviation within the preset time window is calculated. Recent learning activity The value is the normalized number of student learning behaviors within a preset time window, and its range is [value range missing]. ; The rate of cognitive change Based on comprehensive value orientation score The standard deviation is calculated within a preset time window, specifically by first calculating the sample standard deviation of the comprehensive value orientation score sequence within the time window, and then linearly normalizing the sample standard deviation to a preset maximum standard deviation threshold. The interval is used to obtain the rate of cognitive change. ; When the sample standard deviation is greater than or equal to the preset maximum standard deviation threshold The value is 1.

8. The method according to claim 5, characterized in that, Based on students' cognitive levels and the fuzzy level of value orientation determined by the lag threshold mechanism, a two-dimensional joint search is conducted in the ideological and political education resource database to calculate the comprehensive matching score. The formula for the comprehensive matching score is as follows: in, The score is based on the cognitive difficulty matching. Matching scores based on value orientation. For professional matching score, For interest and preference matching score, , , , For the corresponding weight and ; Personalized learning resource sequences are generated based on the overall matching score.

9. A dynamic push system for ideological and political education knowledge points, characterized in that, It includes a data acquisition module for collecting students' multidimensional learning data, which includes answer text data, discussion speech data, and behavior selection data; The value orientation analysis module is used to call the pre-trained stance detection model to perform multi-source semantic analysis on the multi-dimensional learning data, implicitly infer the students' value orientation tendencies, and obtain a comprehensive value orientation score. The rating mapping module is used to map the comprehensive value orientation score to a number of preset fuzzy value orientation ratings and output them. The encrypted storage module is used to store value orientation data using homomorphic encryption technology. When participating in cognitive level calibration calculations, it completes numerical calculations in the ciphertext space, and only outputs the level identifier when decrypting. The cognitive level labeling module is used to construct a student cognitive model based on multidimensional learning data that includes value orientation dimensions, and to label the student's cognitive level. The adaptation and guidance module is used to drive the adaptation of ideological and political education learning resources and answer guidance based on the calibrated cognitive level and value orientation ambiguity level.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 8.

Images