Learning situation assessment method, electronic device, readable storage medium and system

Abstract

This application provides a learning assessment method, electronic device, readable storage medium, and system. The method includes: acquiring a first type of learning data and a second type of learning data, wherein the first type of learning data consists of response result data collected in an uncontrolled environment, and the second type of learning data consists of response result data collected in a controlled environment; extracting verification items from the second type of learning data, wherein the verification items have a preset correlation with the corresponding response result data in the first type of learning data; comparing the verification items with the corresponding response result data in the second type of learning data, and adjusting the confidence weight of the second type of learning data according to the comparison result to obtain calibrated learning data. This application can use controlled acquisition data to calibrate uncontrolled acquisition data, effectively filter data noise, and improve the accuracy of learning assessment.

Description

Technical Field

[0001] This application relates to the field of educational informatization technology, specifically to a learning assessment method, electronic device, readable storage medium, and system. Background Technology

[0002] In educational practice, student learning assessment is a crucial basis for guiding teaching decisions and achieving personalized instruction. Related technologies primarily rely on homework and exam data for student learning assessment.

[0003] Assessments based on exam data, such as analyzing scores from midterms and finals, are generally reliable because the data collection environment is controlled and can accurately reflect students' knowledge acquisition levels. However, these exams are infrequent, the data update cycle is long, and there is a significant lag, making it difficult to reflect students' daily learning progress in a timely manner and providing real-time guidance for daily teaching.

[0004] Assignment-based assessment tracks learning progress by recording the accuracy of students' daily assignments. This type of data is collected frequently and provides rich information about the learning process. However, the data collection environment is often uncontrolled; students may complete assignments by checking answers with each other or using question-searching software, leading to noise in the data and making it difficult to guarantee its authenticity and reliability. Summary of the Invention

[0005] This application provides a learning assessment method, electronic device, readable storage medium, and system that can obtain real and dynamic learning information by fusing different types of data to meet data frequency and data reliability requirements.

[0006] Firstly, embodiments of this application provide a method for assessing learning progress, including: Acquire a first type of learning data and a second type of learning data. The first type of learning data is the answer result data collected in an uncontrolled environment, and the second type of learning data is the answer result data collected in a controlled environment. At least one verification item is extracted from the second type of learning data, and the verification item has a preset correlation with the corresponding answer result data in the first type of learning data; The verification item is compared with the corresponding answer result data in the first type of learning data, and the confidence weight of the first type of learning data is adjusted according to the comparison result to obtain the calibrated learning data.

[0007] Secondly, embodiments of this application also provide a learning assessment device, including: The data acquisition module is used to acquire a first type of learning data and a second type of learning data. The first type of learning data is the answer result data collected in an uncontrolled environment, and the second type of learning data is the answer result data collected in a controlled environment. The calibration module is used to extract at least one verification item from the second type of learning data, wherein the verification item has a preset correlation with the corresponding answer result data in the first type of learning data; and to compare the verification item with the corresponding answer result data in the first type of learning data, and adjust the confidence weight of the first type of learning data according to the comparison result to obtain calibrated learning data.

[0008] Thirdly, embodiments of this application also provide an electronic device, including a processor and a memory, wherein the memory stores multiple instructions; the processor loads instructions from the memory to execute the steps of any of the learning assessment methods provided in embodiments of this application.

[0009] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute the steps of any of the learning assessment methods provided in embodiments of this application.

[0010] Fifthly, embodiments of this application also provide a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps in any of the learning assessment methods provided in embodiments of this application.

[0011] Fifthly, embodiments of this application also provide a learning assessment system, including: Any electronic device provided in the embodiments of this application; The terminal acquisition layer, which is communicatively connected to the electronic device, includes: An uncontrolled acquisition terminal is used to acquire the first type of learning data, which is the answer result data collected in an uncontrolled environment; The controlled acquisition terminal is used to acquire the second type of learning data, which is the answer result data collected in a controlled environment.

[0012] This application's embodiments acquire daily practice data collected in an uncontrolled environment and answer result data collected in a controlled environment. The confidence level of the practice data is calibrated using validation items in the answer result data, effectively filtering noise from daily practice and solving the problems of low reliability of homework data and low frequency of exam data in traditional assessments. Based on this, the forgetting curve is used to predict the knowledge decay coefficient, dynamically generating personalized test sets containing new knowledge points, historical errors, and critical forgetting points, achieving full lifecycle memory management. Simultaneously, by calculating learning rate and ability value to construct a two-dimensional learning profile, the learning potential of different students with the same score can be distinguished, providing a more nuanced basis for differentiated instruction, thereby significantly improving the accuracy of learning assessment and the scientific nature of teaching decisions. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 These are schematic diagrams illustrating application scenarios of the learning assessment methods provided in some embodiments of this application; Figure 2 This is a flowchart illustrating the learning assessment method provided in some embodiments of this application; Figure 3 This is another flowchart illustrating the learning assessment method provided in some embodiments of this application; Figure 4 This is another flowchart illustrating a learning assessment method provided in some embodiments of this application; Figure 5 This is a schematic diagram of the learning assessment device provided in some embodiments of this application; Figure 6 This is another structural schematic diagram of the learning assessment device provided in some embodiments of this application; Figure 7 These are schematic diagrams of the structure of electronic devices provided in some embodiments of this application; Figure 8 This is a schematic diagram of the learning assessment system provided in some embodiments of this application. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0017] "A and / or B" includes the following three combinations: A only, B only, and a combination of A and B.

[0018] The use of "applies to" or "configured to" in this application implies open and inclusive language, which does not exclude the applicability to or configuration to devices performing additional tasks or steps. Additionally, the use of "based on" implies openness and inclusivity, because processes, steps, calculations, or other actions "based on" one or more of the stated conditions or values ​​may in practice be based on additional conditions or values ​​beyond those stated.

[0019] In this application, the term "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0020] This application provides a learning assessment method, apparatus, electronic device, and computer-readable storage medium. Specifically, this embodiment will be described from the perspective of the learning assessment apparatus, which can be integrated into an electronic device. That is, the learning assessment method of this application embodiment can be executed by an electronic device. Optionally, the electronic device may include a terminal device or a server. The terminal device may be a mobile phone, tablet computer, laptop computer, personal computer (PC), interactive teaching tablet, or smart classroom terminal, etc.

[0021] The learning assessment method provided in this application can be applied to systems such as smart teaching systems. This smart teaching system may include terminal devices and a server. The terminal devices may be devices with data acquisition and interaction functions, that is, devices capable of collecting student answer data and interacting with users. The terminal devices and the server can communicate bidirectionally via a network.

[0022] Optionally, the server can be a standalone server, or a server network or server cluster, including but not limited to computers, network hosts, single network servers, multiple network server sets, or cloud servers composed of multiple servers. Cloud servers consist of a large number of computers or network servers based on cloud computing.

[0023] For example, such as Figure 1 As shown, terminal device 10 collects a first type of learning data (students' homework data) and a second type of learning data (weekly / monthly / quarterly assessment data), and sends the collected data to server 20. Server 20 obtains the first type of learning data and the second type of learning data. The first type of learning data is the answer result data collected in an uncontrolled environment, such as the answer result data generated by students in their daily learning process. The second type of learning data is the answer result data collected in a controlled environment, such as the exam result data of students in weekly / monthly / quarterly tests. At least one verification item is extracted from the second type of learning data. This verification item has a preset correlation with the corresponding answer result data in the first type of learning data. The verification item is compared with the corresponding answer result data in the first type of learning data. The confidence weight of the first type of learning data is adjusted according to the comparison result to obtain calibrated learning data. The calibrated learning data is then returned to terminal device 10.

[0024] In the embodiments of this application, to facilitate understanding of the technical solution, the following technical terms are first explained: Answer result data refers to the data set generated after a student completes an answer to a question, and it should include at least the following information: Question Identifier: A unique code or index that identifies the question. Answer content: The answer information submitted by students can be in the form of text, options, images, etc. Correctness indicator: The answer result is judged by automatic grading technology, including two states: correct or incorrect. Answering time: The time a student spends from starting to answering until submitting their answer.

[0025] Answer results data are the basic units for constructing both types of learning data. For example, in a math homework exercise, if a student completes 10 questions, the system will generate 10 answer results data, each corresponding to a complete answer to one question. These answer results data are stored in time series format, reflecting the student's learning process and knowledge mastery within a specific time period.

[0026] A knowledge point is a basic unit within a subject's knowledge system, used to represent the specific knowledge content that students need to master. Each knowledge point has the following attributes: Knowledge point identifier: A unique code or index that identifies the knowledge point, for example, "K_0123" represents "finding the root of a quadratic equation in one variable"; Difficulty parameter: A quantitative indicator used to characterize the difficulty of mastering this knowledge point. The value range is [0, 1], and the larger the value, the higher the difficulty. Knowledge graph position: The node position of this knowledge point in the subject knowledge graph, including its relationship with preceding and subsequent knowledge points.

[0027] Knowledge points are the basic units for organizing learning content and evaluating learning effectiveness. For example, in mathematics, the Pythagorean theorem and quadratic equations are independent knowledge points. In the embodiments of this application, the correlation between verification items and answer result data, memory strength parameters, and learning rate calculations are all organized and managed at the knowledge point level.

[0028] The answer results data represent students' specific performance on particular knowledge points. Each answer result is associated with a knowledge point identifier, indicating the knowledge content tested in that answer. By aggregating multiple answer results for the same knowledge point, the system forms a learning trajectory for that knowledge point, enabling learning analysis and forgetting prediction. For example, a student might have multiple answer results for the Pythagorean theorem, including results from daily exercises and weekly tests. By analyzing the temporal changes in these answer results, the system can determine the student's mastery of the knowledge point and its forgetting trend.

[0029] A controlled environment refers to an environment where students' answering process is effectively monitored, accurately reflecting their knowledge mastery. Controlled environments include, but are not limited to: school-organized weekly and monthly tests, midterm and final exams, in-class quizzes (proctored by teachers), experimental assessments, and in-person interviews. Second-type learning data collected under such environments has high reliability, and the initial confidence weight is set to a high value.

[0030] An uncontrolled environment refers to an environment where students' answering process is not effectively supervised, potentially leading to behaviors such as peer review, use of question-searching software, or perfunctory completion of tasks. Primary learning data collected in uncontrolled environments includes, but is not limited to, homework exercises, self-practice, and online learning platform exercises. Data collected in such environments is frequent, but the initial confidence weights are set to low values.

[0031] The following detailed description is provided in conjunction with the accompanying drawings. In this embodiment, the execution subject is a terminal device as an example. It should be noted that the order of description in the following embodiments is not intended to limit the preferred order of the embodiments. Although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be performed in a different order than that shown in the accompanying drawings.

[0032] Please refer to Figure 2 The specific process of this learning assessment method can be summarized in steps S201 to S203, where: Step S201: Obtain the first type of learning data and the second type of learning data. The first type of learning data is the answer result data collected in an uncontrolled environment, and the second type of learning data is the answer result data collected in a controlled environment.

[0033] Optionally, the first type of learning data consists of students' answer results generated during their daily learning process. This answer result data may include information such as the question label, answer content, correctness indicator, and answer time when students complete daily exercises. For example, in mathematics teaching, the first type of learning data is obtained by collecting students' answers to their daily homework problems, including whether each question is correct or not, the solution steps, and the completion time. The first type of learning data is collected frequently and can reflect the students' learning process. However, due to uncontrollable environments, such as students comparing answers with each other or using question-searching software, the data noise is high. Therefore, the initial confidence level of the first type of learning data is set to a low value (e.g., 0.5).

[0034] Optionally, the second type of learning data consists of student test responses generated in a controlled environment. This response data can include student answer records in pre-designed testing scenarios such as weekly, monthly, quarterly, and semester tests. For example, the second type of learning data can be obtained by collecting students' answers to weekly math tests, including the correctness indicators for each question. The second type of learning data is collected in a controlled environment, such as an examination room or under teacher supervision, and can accurately reflect students' levels. Therefore, the initial reliability of the second type of learning data is set to a high value (e.g., 1.0). However, this type of data has a long collection period, low frequency, and strong lag characteristics, making it insufficient for guiding daily teaching.

[0035] Step S202: Extract at least one verification item from the second type of learning data. The verification item has a preset correlation with the corresponding answer result data in the first type of learning data.

[0036] The pre-defined correlation means that the knowledge content examined by the verification item is comparable to the knowledge content involved in a certain answer result data in the first type of learning data, thereby ensuring the accuracy of the calibration results.

[0037] As an optional embodiment, the preset association relationship includes the knowledge point identifier corresponding to the verification item being the same as the knowledge point identifier corresponding to the answer result data.

[0038] Specifically, each question in the test question database is pre-marked with a knowledge point identifier, which is used to uniquely identify the knowledge content tested in the question. The knowledge point identifier system is constructed according to the subject knowledge structure. For example, in mathematics, the knowledge point identifier corresponding to "finding the roots of a quadratic equation in one variable" is "K_0123", and the knowledge point identifier corresponding to "Pythagorean theorem" is "K_0456".

[0039] When establishing the association, the knowledge point identifiers of the questions corresponding to the answer results are first obtained. Then, questions with the same knowledge point identifiers are selected from the test question database as verification items. For example, if a student completes a question about "finding the roots of a quadratic equation" in an assignment, the knowledge point identifier for that question is "K_0123". In subsequent weekly tests, a question also labeled "K_0123" is inserted as a verification item, thus ensuring that the verification item and the answer results test the same knowledge content.

[0040] As another optional embodiment, the preset association relationship includes the cognitive classification level corresponding to the verification item being the same as the cognitive classification level corresponding to the answer result data.

[0041] Specifically, each question in the test question database is pre-labeled with a cognitive classification level, which is based on Bloom's Taxonomy of Cognition and includes six levels: memory, comprehension, application, analysis, evaluation, and creation. Questions at different levels test different cognitive abilities of students.

[0042] When establishing associations, the system first obtains the cognitive classification level of the questions corresponding to the answer results data, and then selects questions with the same cognitive classification level from the test question database as verification items. For example, if a homework question requires students to "recite the content of the Pythagorean theorem," this question belongs to the "memory" level; the system will then embed a question that also tests the "memory" level in the weekly test as a verification item, such as "please write the expression of the Pythagorean theorem." If a homework question requires students to "use the Pythagorean theorem to calculate the side length of a right triangle," this question belongs to the "application" level; the system will then select a question that also belongs to the "application" level as a verification item.

[0043] As another optional embodiment, the preset association relationship includes the difference between the difficulty parameter corresponding to the verification item and the difficulty parameter corresponding to the answer result data being less than a preset threshold.

[0044] Specifically, each question in the test question database is pre-labeled with a difficulty parameter. This parameter is calculated based on the item response theory model and typically ranges from [-3, 3] to [0, 1]. A higher value indicates a more difficult question. The difficulty parameter reflects the question's ability to differentiate between students' ability levels.

[0045] When establishing the association, the system first obtains the difficulty parameter of the question corresponding to the answer result data. Then, it selects questions from the question database whose absolute difference between the difficulty parameter and the difficulty parameter of the answer result data is less than a preset threshold as verification items. For example, if the preset threshold is set to 0.2 and the difficulty parameter of a certain assignment question is 0.5, the system selects questions with a difficulty parameter in the range of 0.3 to 0.7 as verification items to ensure that the verification items and the answer result data have a similar difficulty level.

[0046] As another optional embodiment, the preset association relationship includes the fact that the question format corresponding to the verification item is the same as the question format corresponding to the answer result data.

[0047] Specifically, each question in the test question database has a defined question format, including but not limited to multiple choice, fill-in-the-blank, true / false, short answer, calculation, and proof questions. Different question formats have different requirements for students' answers and different assessment focuses.

[0048] When establishing a connection, the question format corresponding to the answer results data is first obtained, and then questions with the same question format are selected from the question database as verification items. For example, if the homework question is a multiple-choice question, then multiple-choice questions are selected as verification items; if the homework question is a fill-in-the-blank question, then fill-in-the-blank questions are selected as verification items.

[0049] The four types of relationships mentioned above can be used independently or in combination. For example, a verification item can be required to simultaneously meet two conditions: the knowledge point identifier is the same and the difference in difficulty parameter is less than a preset threshold, in order to further improve the matching accuracy of the verification item.

[0050] Step S203: Compare the verification item with the corresponding answer result data in the first type of learning data, adjust the confidence weight of the first type of learning data according to the comparison result, and obtain the calibrated learning data.

[0051] By fusing and calibrating the first type of learning data with the second type of learning data through the above steps, we can obtain learning data with higher reliability.

[0052] The learning assessment method provided in this application addresses the problems of insufficient reliability of uncontrolled first-type learning data and excessively low frequency of controlled second-type learning data in related technologies by introducing response data collected in a controlled environment as a verification benchmark to calibrate response data generated during daily learning. By utilizing verification items in the second-type learning data that are homogeneous with the first-type learning data, and comparing the correctness consistency between the verification items and the first-type learning data, noise in the first-type learning data can be identified and the corresponding confidence weights adjusted. This effectively filters data noise while preserving the high-frequency characteristics of the first-type learning data, achieving an organic fusion of high reliability and high-frequency data, thereby obtaining a realistic and dynamic learning profile.

[0053] In some embodiments, at least one validation item is extracted from the second type of learning data, including: Obtain the correct answer data from the first type of learning data as the seed set; Based on the preset association relationship, candidate verification items that match the answer results data in the seed set are retrieved from the preset question database to form a candidate verification item set; wherein, the questions in the question database are pre-constructed with feature vectors, and the feature vectors contain at least one feature value used to represent the association relationship; Based on the student's current ability estimate, candidate verification items with information content exceeding a preset information content threshold are selected from the candidate verification item set as verification items.

[0054] Specifically, by scanning the answer data generated by students during their daily learning process—that is, the first type of learning data—the system filters out the answer data marked as "correct." This data represents the knowledge points that students believe they have mastered, and their authenticity needs to be verified through subsequent validation items. "Correct" answer data is chosen as the seed because if a student submits a correct answer but has not actually mastered the knowledge point (e.g., plagiarism, or searching for answers online), then verification is required to identify this.

[0055] For example, if a student completes 10 questions in a math assignment, answering 8 correctly and 2 incorrectly, the data corresponding to the 8 correctly answered questions will be used as seeds to form a seed set. Each seed answer data includes a question identifier, a knowledge point identifier, a difficulty parameter, a cognitive classification level, and information about the question format.

[0056] Each question in the question database has a pre-constructed feature vector, which contains at least one feature value representing the association relationship. Based on the association relationships described in the foregoing embodiments, the homogeneous dimensions corresponding to these feature values ​​include, but are not limited to: Knowledge point identification dimension: The knowledge points tested in the question are represented by one-hot encoding or embedded vectors, and the corresponding feature values ​​are one-hot encoding or vector values; Cognitive classification hierarchy dimension: Numerical codes are used to represent the Bloom cognitive hierarchy in which the question is located (e.g., memory=1, comprehension=2, application=3, analysis=4, evaluation=5, creation=6), and the corresponding feature values ​​are numerical codes; Difficulty parameter dimension: The difficulty parameter value is calculated using the project response theory model, and the corresponding feature value is the difficulty parameter value; Question Format Dimension: One-hot encoding is used to represent the question type (such as multiple choice, fill-in-the-blank, problem-solving, etc.), and the corresponding feature value is the one-hot encoding.

[0057] Each answer result in the seed set serves as a query vector for vectorized retrieval within the question data. For example, the retrieval criteria are questions that meet preset threshold conditions across four homogeneous dimensions. For instance, identical knowledge point identifiers, identical cognitive classification levels, difficulty parameter differences less than 0.2, and identical question formats are used as retrieval criteria to recall questions from the question database that meet all conditions as candidate validation items.

[0058] Perform the above retrieval operation on each answer result in the seed set, merge all retrieval results, remove duplicates, and form a candidate verification item set. For example, if the seed set contains 3 answer results, and each result yields 5 candidate verification items, the deduplicated candidate verification item set will contain 12 candidate questions.

[0059] In some embodiments, based on the student's current ability estimate, candidate verification items with information content greater than a preset information content threshold are selected from the candidate verification item set as verification items, including: Obtain an estimate of the student's current ability; For each candidate validation item in the candidate validation item set, calculate the information value of the candidate validation item at the capability estimate; Candidate verification items with information content values ​​greater than the information content threshold are selected as verification items.

[0060] Student's current ability estimate θ This is calculated based on students' past answer data and reflects their current level of knowledge mastery. For each candidate validation item in the candidate validation item set, the ability estimate for that candidate validation item is calculated. θ The information content value at each point. The magnitude of the information content value reflects the estimated ability of the question to distinguish between different individuals. θ The effectiveness of the test among the nearby student population. The higher the information content value, the more accurately the test can measure students' ability level.

[0061] For example, the candidate verification item with the highest information content value is selected from the candidate verification item set as the final verification item. For instance, if there are three questions in the candidate verification item set with information content values ​​of 0.8, 1.2, and 0.5 respectively, the question with an information content value of 1.2 is selected as the verification item and embedded into the assessment test paper.

[0062] If multiple questions in the candidate verification item set have the same or similar information content values, the system can further combine other factors for selection, such as the question's discrimination parameter and the frequency of the question's use in the question bank.

[0063] The verification item extraction method provided in this embodiment uses a three-stage mechanism of seed extraction, vectorized retrieval, and information content filtering to enable verification items to be dynamically adjusted according to the student's actual ability level, thereby improving the accuracy of calibration results and the relevance of testing.

[0064] In some embodiments, the verification item is compared with the corresponding answer result data in the first type of learning data, and the confidence weight of the first type of learning data is adjusted according to the comparison result to obtain calibrated learning data, including: Obtain the first accuracy rate of the verification item and the second accuracy rate of the corresponding answer result data in the first type of learning data; wherein, the first accuracy rate is calculated based on the answer correctness identifier of the verification item, and the second accuracy rate is calculated based on the answer correctness identifier of the corresponding answer result data in the first type of learning data; Based on the difference between the first and second accuracy rates, the confidence weight of the first type of learning data is determined, and the calibrated learning data is obtained.

[0065] As mentioned earlier, the correctness indicator includes two states: "correct" or "incorrect." As an optional implementation, the correctness indicator for each question can be obtained through automatic grading technology. For example, for objective questions (such as multiple-choice and fill-in-the-blank questions), the student's answer is compared with the corresponding preset answer information. If they match, it is marked as correct; otherwise, it is marked as incorrect. For subjective questions (such as problem-solving questions), semantic analysis technology can be used to determine the degree of matching between the answer and the preset answer information. If the matching degree exceeds a preset threshold, it is marked as correct; otherwise, it is marked as incorrect.

[0066] Based on this, the first accuracy rate represents the proportion of questions answered correctly by students in controlled tests out of the total number of questions corresponding to the validation items. The second accuracy rate represents the proportion of questions answered correctly by students in daily practice out of the total number of questions corresponding to the homogeneous questions in the first type of learning data.

[0067] When a validation item contains a single question, the first accuracy rate is 0 or 1 (0 indicates an incorrect answer, and 1 indicates a correct answer). When a validation item contains multiple questions, the first accuracy rate is the ratio of the number of correct answers to the total number of questions. Similarly, the second accuracy rate of the corresponding answer results data in the first learning data is calculated in the same way.

[0068] For example, if a weekly test includes 3 verification questions and all students answer them correctly, the first accuracy rate is 1.0; in the corresponding homework, if there are 3 questions of the same type as the verification questions and students answer 2 correctly, the second accuracy rate is 0.667.

[0069] Specifically, the difference between the first accuracy rate of the validation item and the second accuracy rate of the corresponding answer results data in the first learning data is calculated. The smaller the difference, the higher the consistency between the two; the larger the difference, the more obvious the contradiction between them. Based on the magnitude of the difference and the specific accuracy rates of the validation item and the first learning data, the credibility of the first learning data is comprehensively determined, and its confidence weight is adjusted accordingly.

[0070] When the difference value is small, meaning the verification item is highly consistent with the corresponding answer results in the first learning data, the first learning data is considered credible evidence. In this case, the confidence weight of the first learning data is increased, giving it a higher weight in subsequent learning assessments, and it is used together with the second learning data to update the student's ability model.

[0071] When the difference is large, and the first accuracy rate of the validation item is less than the second accuracy rate of the corresponding answer result data in the first learning data, it is determined that the first learning data may contain noise (such as plagiarism or searching for answers). In this case, the confidence weight of the first learning data is reduced, and a rollback mechanism is executed, that is, the student's ability model is updated only based on the second learning data, ignoring the first learning data in this instance.

[0072] When the discrepancy is large, and the second accuracy rate of the corresponding answer result data in the first learning data is greater than the first accuracy rate of the verification item, it is determined that the student's actual ability meets the standard, but there may be problems with their daily learning attitude (such as being perfunctory). In this case, the confidence weight of the first learning data remains unchanged, and an attitude warning is marked for teachers and parents to pay attention to.

[0073] Through the above-mentioned differentiated processing, while preserving the high-frequency characteristics of the first learning data, data noise can be effectively filtered out and attitude problems can be identified. This ensures that the calibrated learning data can not only reflect the students' true ability level, but also provide early warning information for teaching management.

[0074] In some embodiments, determining the confidence weight of the first type of learning data based on the difference between the first correctness and the second correctness includes: Calculate the consistency coefficient between the first and second accuracy rates.

[0075] If the consistency coefficient is greater than the preset consistency threshold, the confidence weight of the first type of learning data is increased.

[0076] If the consistency coefficient is less than the preset consistency threshold and the first accuracy is less than the second accuracy, then the confidence weight of the first type of learning data is reduced.

[0077] If the consistency coefficient is greater than the preset consistency threshold and the first accuracy rate is greater than the second accuracy rate, then the confidence weight of the first type of learning data remains unchanged, and warning label information is generated.

[0078] Specifically, obtain the first accuracy of the validation item. Acc Anchor The second accuracy rate of the corresponding answer results data in the first learning data Acc Homework Calculate the consistency coefficient C =| Acc Anchor - Acc Homework This coefficient is used to quantify the degree of consistency between the verification item and the corresponding answer result data in the first learning data. The value range is [0, 1]. The smaller the value, the more consistent the two are, and the larger the value, the greater the difference between the two.

[0079] For example, in a certain assignment, the student got all 5 questions correct, that is... Acc Homework =1.0; This corresponds to all three verification items in the weekly test being answered correctly, i.e. Acc Anchor = A consistency coefficient of 1.0 indicates a high degree of consistency. C =0. If all questions corresponding to the verification item are incorrect, that is... Acc Anchor = If the consistency coefficient is 0, then the consistency coefficient is 0. C =1.0.

[0080] Based on the consistency coefficient C The value of is dynamically adjusted based on the specific accuracy of the responses to the validation items and the corresponding first learning data. The confidence weight for the previous period is set to . W t-1 (Initial value set to 0.5), consistency threshold C th Set the learning rate to 0.8. α Set to 0.2, sensitivity index k Set the weight to 2. The weight update formula is as follows: W t = W t-1 ×(1+ α · sgn ( C - C th )·| C - C th | k ) When the consistency coefficient C Less than the preset consistency threshold C th When the confidence level is reached, it indicates that the validation item and the corresponding response data in the first learning data are highly consistent, and the validation item verifies the authenticity of the first learning data. At this point, the first learning data is determined to be credible evidence, and its confidence weight is increased to approach the upper limit (e.g., 1.0). After the weight update, the first learning data is incorporated into the student ability estimate. θ Update calculation.

[0081] When the consistency coefficient C Less than the preset consistency threshold C th And the second accuracy rate of the corresponding answer results data in the first learning data. Acc HomeworkSignificantly higher than the first correct answer of the validation items Acc Anchor If the first learning data is found to be noisy, such as students copying homework or using question-searching software, then the confidence weight of this first learning data is reduced, causing it to decay exponentially to the lower limit (e.g., 0.1).

[0082] For example, in a certain assignment, a student answered all 10 questions correctly, that is... Acc Homework = Version 1.0, but all three questions corresponding to the verification item in the weekly test were incorrect, i.e. Acc Anchor =0, then C =1.0, which is greater than the consistency threshold of 0.2. Therefore, the first learning data is determined to contain noise, and the confidence weight of the first learning data is reduced from 0.5 to 0.1.

[0083] Simultaneously, a rollback mechanism is implemented: positive updates to student ability estimates based on the first learning data are revoked, retaining only the data as a negative sample or ignoring it entirely, to prevent false high scores from contaminating the student's learning profile. For example, if a student's "quadratic equation" ability score was previously increased by 0.2 based on the first learning data, the rollback will restore the ability score to its pre-increase level.

[0084] When the consistency coefficient C Less than the preset consistency threshold C th And the first correctness of the validation items Acc Anchor Significantly higher than the first accuracy rate of the corresponding answer results data in the first learning data. Acc Homework This indicates that the student's actual ability meets the standard, but there may be attitude problems in daily learning (such as perfunctory work or careless reading of questions). At this time, although... C If the value is large, but it belongs to the special case of "ability exceeding performance", the confidence weight of the first learning data is not reduced. Instead, the ability value is locked and not lowered, and a warning mark information is generated to activate the hierarchical response mechanism.

[0085] For example, in a certain assignment, a student got all 10 questions wrong, that is... Acc Homework =0, but all 3 questions corresponding to the verification item in the weekly test were answered correctly, i.e. Acc Anchor =1.0, then C =1.0, which is greater than the consistency threshold of 0.2. At this point, it is determined to be an attitude problem, and the confidence weight of the first learning data is not reduced.

[0086] In some embodiments, after keeping the confidence weights of the first type of learning data unchanged, the method further includes: Generate early warning marker information; Send a corresponding prompt message to the teacher's end, which indicates that there is a difference in the accuracy rate between the first type of learning data and the second type of learning data; Add status tags to student learning portfolios. These status tags are used to identify whether the response data corresponding to the first type of learning data is of the attitude fluctuation type; or A comparison report is sent to parents, which shows the accuracy rate of the first type of learning data compared with that of the second type of learning data.

[0087] Specifically, when the accuracy rate of the first type of learning data is lower than that of the second type of learning data (i.e., the homework performance is worse than the assessment performance), the system generates an early warning flag, which is used to trigger a tiered response mechanism. This flag is used to distinguish between two different types of learning situations: insufficient ability and attitude problems, and guides the system to adopt differentiated information delivery strategies.

[0088] When the accuracy rate of the first type of learning data is significantly lower than the accuracy rate of the second type of learning data, that is, when the student's homework performance is significantly worse than their assessment performance, the following tiered response mechanism will be implemented: At the first level, the student is marked as a "key student" in the teacher's classroom learning dashboard, and a notification message is sent to the teacher's end in a highlighted format. This notification message indicates a difference in the accuracy rates between the first and second types of learning data. The information displayed on the teacher's end includes the student's name, a comparison of homework and assessment performance, and historical records of attitude fluctuations, facilitating timely understanding of the situation and communication with the student.

[0089] The second level involves recording a "attitude fluctuation" tag in the student's learning portfolio and linking it to comparative data from the assignment and assessment. This tag is used to accumulate the student's attitude change history, forming a long-term tracking record for subsequent learning analysis and teaching decisions. This record does not directly affect the student's ability assessment; it only serves as supplementary information to the learning profile.

[0090] The third level involves sending weekly or monthly reports to parents, which include a chart comparing homework performance with actual abilities. This chart visually illustrates the difference between the student's performance in homework and their actual abilities as reflected in assessments. The chart is not pushed out in real-time to avoid potential conflicts between school and home caused by real-time alerts. Instead, it uses regular reports to keep parents informed, encouraging them to focus on the process of completing assignments rather than just the results.

[0091] Simultaneously with the aforementioned tiered response, based on the weight adjustment results, the response data is associated with and stored along with the corresponding confidence weights to obtain calibrated learning data. In this data, each response data point is accompanied by a confidence weight coefficient, used to characterize the reliability of the response data.

[0092] The tiered response mechanism provided in this embodiment achieves differentiated handling of attitude issues through a three-tiered response system consisting of teacher prompts, student learning record entries, and parent reports. This not only protects the accuracy of ability estimation but also provides effective information support for teaching management.

[0093] Please see Figure 3 This is a flowchart illustrating another learning assessment method provided in an embodiment of this application. It should be noted that... Figure 3 Steps S301 to S303 in the process are Figure 2 The steps S201 to S203 correspond to each other, and their specific implementation methods can be found in [reference needed]. Figure 2 The relevant descriptions will not be repeated here. Based on the calibration steps described above, this embodiment further introduces a forgetting model to predict the critical forgetting point and generate personalized test sets.

[0094] Specifically, the method also includes the following steps: Step S304: Based on the calibrated learning data, obtain the learning time interval corresponding to each knowledge point with historical answer result data. The historical answer result data includes the answer result data associated with the knowledge points in the first type of learning data and the second type of learning data.

[0095] Specifically, historical answer data for each knowledge point is retrieved from the calibrated learning data, including the timestamp of the most recent examination or review of that knowledge point. The difference between the current moment and that timestamp is calculated as the learning time interval Δt for that knowledge point. The unit of the learning time interval is hours or days, reflecting the length of time since the student last encountered that knowledge point.

[0096] For example, if a student learned the Pythagorean theorem 30 days ago and no related answer data has been detected by the system since then, the learning time interval for this knowledge point is Δt = 720 hours (30 days × 24 hours).

[0097] Step S305: Calculate the decay coefficient of each knowledge point according to the preset decay function. The decay function is used to characterize the mapping relationship between the learning time interval and the memory strength parameter.

[0098] The initial memory strength parameter is inversely proportional to the difficulty parameter of the knowledge point: the higher the difficulty of the knowledge point, the smaller the initial memory strength parameter, which means that the memory decays faster; the lower the difficulty of the knowledge point, the larger the initial memory strength parameter, which means that the memory is retained for a longer time, as shown in Table 1 below.

[0099]

[0100] Table 1 shows the mapping relationship between the knowledge point difficulty parameter and the initial memory strength parameter. The memory retention rate R(t) of each knowledge point is calculated based on the preset decay function:

[0101] Where Δt is the learning time interval. S 0 represents the initial memory strength parameter. The memory retention rate R(t) ranges from (0, 1], with smaller values ​​indicating a higher degree of forgetting.

[0102] In some embodiments, the attenuation coefficient of each knowledge point is calculated according to a preset attenuation function, including: Obtain the initial memory strength parameters for each knowledge point; The decay coefficient is calculated based on the ratio of the learning time interval to the initial memory strength parameter.

[0103] Specifically, the memory retention rate is converted into a decay coefficient as described above. The decay coefficient D = 1 - R(t), and the larger the value, the more severe the forgetting.

[0104] For example, regarding the knowledge point of "composite functions" with a difficulty level of "difficult", S 0 = 24 hours. If a student has studied this knowledge point for 48 hours (Δt = 48), then the memory retention rate is... The value is approximately 0.135, and the attenuation coefficient D = 1 - 0.135 = 0.865, indicating that this knowledge point has been severely forgotten.

[0105] Step S306: Based on the attenuation coefficient of each knowledge point, select test questions from the preset test question database to generate a test set containing first type test question data, second type test question data and third type test question data. The first type test question data corresponds to knowledge points for which no historical answer data has appeared. The second type test question data corresponds to test questions in the historical answer data that are marked as incorrect. The third type test question data corresponds to knowledge points for which the attenuation coefficient is lower than the preset threshold.

[0106] In an optional embodiment, the first type of test question data corresponds to knowledge points for which no historical answer data has appeared, i.e., new knowledge points. The filtering of the first type of test question data includes: The system selects all questions from a pre-set question database that test knowledge points for which no historical answer data has been found, such as all questions covering new knowledge points covered in this week's teaching syllabus, and forms a candidate question set.

[0107] Obtain estimates of students' current abilities in relevant cognitive dimensions. θ For each question in the candidate set, calculate the ability estimate for that question. θ Information value I at the location θ The information content value is used to measure how effectively a question can differentiate students at a specific ability level. Its calculation formula is as follows:

[0108] in, a The discrimination parameter of the question. b This is a difficulty parameter for the question. A higher information content value indicates a greater level of difficulty for the question. θ The higher the accuracy of student measurements.

[0109] Questions with the highest information content and whose difficulty parameters meet the matching range (e.g., |b-θ|<0.3) are selected as the first type of test data.

[0110] To ensure that the difficulty of the test questions is neither too discouraging for students nor too ineffective in assessing their level of understanding, the system does not simply aim to maximize the amount of information provided. Instead, it first filters and matches the test questions based on difficulty levels. Specifically, the system sets a difficulty matching range and filters difficulty parameters. b Satisfy | b - θ| The questions, among which Match the preset difficulty threshold (e.g.) This range ensures that the selected questions are appropriate for the students' current ability level, avoiding inaccurate measurements due to questions being too difficult or too easy. From the candidate question set that meets the difficulty matching range, the system uses an optimization algorithm to extract questions. Under the premise of satisfying preset constraints on question type distribution and total number of questions, the system selects the optimal combination of questions from the candidate question set as the first type of test data. In this way, the system can maximize the accuracy of measuring students' ability levels while ensuring a reasonable test structure.

[0111] In an optional embodiment, the second type of test question data corresponds to test questions marked as incorrect in the historical answer results data, i.e., historical wrong questions. High-frequency error points are extracted from the historical wrong question set to generate variant questions as the second type of test question data. Specifically, the historical error records of students on each knowledge point are counted, and knowledge points with error frequency higher than a preset threshold are screened out. Variant questions that test the same knowledge points as the original wrong questions but have different question stem parameters are selected from the question bank.

[0112] In an optional embodiment, the third type of test question data corresponds to knowledge points with a decay coefficient lower than a preset threshold, i.e., critical forgetting points. Based on the decay coefficient of each knowledge point, knowledge points with a decay coefficient lower than the preset threshold (e.g., a memory retention rate lower than 60%) are selected as critical forgetting points. For example, in the above example, the decay coefficient D of "composite function" is 0.865, which is lower than the preset threshold and is determined to be a critical forgetting point. Questions related to this knowledge point are selected from the question bank as the third type of test question data and inserted into the test set.

[0113] The three types of test data are combined to generate a test set. The order of the test questions in the test set can be arranged according to the teaching strategy: usually, the first type of test data (test items for new knowledge points) is placed at the beginning to check the mastery of new knowledge points; the third type of test data (test items for critical forgetting points) is interspersed in to achieve "recall-style" review; and the second type of test data (test items for historical wrong questions) is placed at the end for consolidation and reinforcement.

[0114] For example, a test set for a particular student includes: The first type of test data includes 3 questions on the new knowledge point of "quadratic equations in one variable"; The second type of test data includes two variant questions on the historical errors related to "factorization"; The third type of test data includes one question about the critical point of forgetting the Pythagorean theorem.

[0115] Optionally, the generated test set can also be pushed to the teacher's or student's end for subsequent assessment implementation.

[0116] The personalized test set generation method based on the forgetting model provided in this embodiment achieves a quantitative assessment of students' knowledge forgetting status by calculating the learning time interval and decay coefficient. For new knowledge points, the system matches the most suitable test questions according to the student's ability level; for historical incorrect questions, the system reinforces and consolidates them through variant questions; for critical forgetting points, the system actively triggers wake-up tests, thereby solving the problem of ignoring the laws of knowledge forgetting in related technologies and realizing a leap from "single test" to "full life cycle memory management". Through personalized test sets, students can review in time before forgetting occurs, effectively improving the knowledge retention rate; teachers can obtain students' true knowledge mastery status, providing a basis for subsequent teaching decisions.

[0117] In some embodiments, the method further includes: If the answer results data corresponding to the third type of test questions match the corresponding preset answer information, then the memory strength parameter of the knowledge point will be increased according to the preset growth coefficient. If the answer results data corresponding to the third type of test questions do not match the corresponding preset answer information, the memory strength parameter of the knowledge point will be reduced according to the preset attenuation coefficient.

[0118] Specifically, the system receives students' answers to the third type of test questions and matches these answers with corresponding preset answer information. For objective questions (such as multiple choice and fill-in-the-blank questions), the system directly compares the students' answers with the preset answer information; if they match, it is considered a match; otherwise, it is considered a mismatch. For subjective questions (such as problem-solving and proof questions), semantic analysis technology is used to determine the degree of matching between the answer and the standard answer; if the matching degree exceeds a preset threshold (such as 80%), it is considered a match; otherwise, it is considered a mismatch.

[0119] For example, a student completes a critical forgetting point test question about the Pythagorean theorem and answers "a² + b² = c²", which perfectly matches the preset answer information, and the system determines it as a match. Another student answers "a + b = c", which does not match the preset answer information, and the system determines it as a mismatch.

[0120] If the answer data corresponding to the third type of test questions matches the corresponding preset answer information, it indicates that the student's memory of the knowledge points is well maintained, and the "recall" is successfully completed. At this time, the memory strength parameter of the corresponding knowledge points is increased according to the preset growth coefficient.

[0121] Specifically, the memory strength parameter before the update is set as follows: S old The growth coefficient is μ Its default value is 1.5, which increases with the number of consecutive correct answers, and the upper limit of memory strength is [value missing]. S max (e.g., 180 days, corresponding to the long-term memory area). Updated memory strength parameters. S new The calculation formula is as follows: S new = min( S old ×(1+ μ ), S max ) This formula shows that the memory strength parameter increases non-linearly with the growth coefficient, but does not exceed a preset upper limit. S max This is to prevent unlimited growth from leading to excessively long review intervals.

[0122] For example, the current memory strength parameter of a certain knowledge point S old =24 hours, growth coefficient μ =1.5, then S new=min(24×(1+1.5), 180)=60 hours. If the growth coefficient increases to 2.0 for the second consecutive match, then... S new =min(24×(1+2.0), 180)=72 hours, until the upper limit of 180 hours is reached.

[0123] If the answer data corresponding to the third type of test questions does not match the corresponding preset answer information, it indicates that the student has forgotten the knowledge point, and the recall has failed. In this case, the memory strength parameter of the knowledge point is reduced according to the preset decay coefficient.

[0124] Specifically, the memory strength parameter before the update is set as follows: S old The attenuation coefficient is λ Its default value is 0.4, and the initial memory strength parameter is... S 0. Updated memory strength parameters S new The calculation formula is as follows: S new = max( S old × λ , S 0) This formula indicates that the memory strength parameter decreases with the decay coefficient, but does not fall below the initial memory strength parameter. S 0, to prevent excessive decay of parameters due to a single instance of forgetting. Once this "punishment rollback" mechanism is triggered, the next review interval needs to be significantly shortened to reinforce memory again.

[0125] For example, the current memory strength parameter of a certain knowledge point S old =72 hours, attenuation coefficient λ =0.4, initial strength S 0 = 12 hours, then S new = max(72×0.4, 12) = 28.8 hours. If there is a second consecutive mismatch, and the attenuation coefficient remains at 0.4, then S new = max(28.8×0.4,12)=12 hours.

[0126] Updated memory strength parameters S new The memory is stored in association with corresponding knowledge points, serving as the basis for the next round of forgetting prediction. The updated memory strength parameters will be used to calculate the subsequent memory retention rate, thereby affecting the selection of the next critical forgetting point and the determination of the recall timing.

[0127] For example, if the memory strength parameter for a certain knowledge point is updated to 36 hours, the system will recalculate the memory retention rate of that knowledge point after 36 hours. If it is lower than the preset threshold, it will be included in the third type of test data again.

[0128] The dynamic update mechanism for memory strength parameters provided in this embodiment achieves precise modeling of individual learning patterns through differentiated processing for matching and non-matching cases. Specifically, when a student successfully recalls a knowledge point, the memory strength parameter increases by a growth coefficient, extending the next review interval and avoiding ineffective repetitive training; when a student fails to recall a knowledge point, the memory strength parameter decreases by a decay coefficient, shortening the next review interval and promptly compensating for forgetting. This "growth for correct answers, rollback for incorrect answers" mechanism conforms to the Ebbinghaus forgetting curve, dynamically adjusting the review pace according to the individual's actual mastery, making the review schedule more scientific and efficient. Simultaneously, by setting upper and lower limits for memory strength, a reasonable range for the parameters is ensured, preventing abnormal fluctuations in extreme cases and ensuring the stability and robustness of the system.

[0129] Please see Figure 4 This is a flowchart illustrating another learning assessment method provided in an embodiment of this application. It should be noted that... Figure 4 Steps S401 to S403 in the process are Figure 2 The steps S201 to S203 correspond to each other, and their specific implementation methods can be found in [reference needed]. Figure 2 The relevant descriptions will not be repeated here. Based on the calibration steps described above, this embodiment further introduces a two-dimensional evaluation mechanism to achieve a comprehensive assessment of students' learning rate and ability.

[0130] Specifically, the method also includes the following steps: Step S404: Based on the calibrated learning data, obtain the student's answer result data sequence for the target knowledge point. The answer result data sequence includes the correctness indicators and answer duration of multiple answers.

[0131] A sequence of answer results refers to the collection of all student answers to a specific knowledge point within a given time period, arranged in chronological order. Each answer result must include at least a correctness indicator and the answer duration.

[0132] The correctness indicator includes two states: correct or incorrect, which are obtained by the system through automatic grading technology. The answering time is the time consumed by the student from starting to answering to submitting the answer, in seconds or minutes.

[0133] For example, the data sequence of students' answers to the question "quadratic equations in one variable" is as follows: First attempt: Correctness marked as correct, duration 45 seconds. Second attempt: Correctness flagged as correct, duration 38 seconds. Third time: Correctness marked as incorrect, duration 52 seconds. 4th time: Correctness marked as correct, duration 42 seconds. 5th time: Correctness marked as correct, duration 35 seconds. 6th time: Correctness marked as correct, duration 40 seconds. Step S405: Based on the sequence of answer results, determine the cumulative number of questions and effective learning time corresponding to the student's transition from the first state to the second state.

[0134] The cumulative number of questions is the number of answer result data included between the first state and the second state. The effective learning time is calculated based on the answer time in the answer result data included between the first state and the second state. The first state is when the proportion of correct answers marked as incorrect in a series of consecutive preset number of answers is higher than a first threshold. The second state is when the proportion of correct answers marked as correct in a series of consecutive preset number of answers is higher than a second threshold.

[0135] The first state is defined as "not mastered," and the specific quantitative standard is: in a predetermined number of consecutive attempts (e.g., 5 times), the percentage of correct answers marked as incorrect is higher than a first threshold (e.g., 60%). That is, if the number of incorrect answers exceeds 3 in 5 consecutive attempts, the student is judged to be in a state of not mastering the material.

[0136] The second state is defined as "stable mastery state," and the specific quantitative standard is: in a predetermined number of consecutive attempts (e.g., 3 times), the percentage of correct answers is higher than the second threshold (e.g., 80%). That is, if the number of correct answers reaches 3 out of 3 consecutive attempts (100% accuracy), the student is judged to have reached the stable mastery state.

[0137] The system identifies the cycle of transition from the first state to the second state from the sequence of answer results. Taking the example above, if the student answers correctly the first two times and incorrectly the third time, they have not yet reached a stable mastery state. Starting from the fourth time, if they answer correctly three consecutive times (the fourth, fifth, and sixth times), it is determined that the student has completed the transition from the unstable state to the stable mastery state during the period from the fourth to the sixth answer.

[0138] The total number of questions students actually answered for this knowledge point within the conversion period is counted as the cumulative number of questions. In the example above, students answered questions 3 times from the 4th to the 6th time, resulting in a cumulative number of 3 questions.

[0139] The formula for calculating effective learning time is:

[0140] in, t iThe theoretical standard time taken for each question (derived from full sample big data). η i This is the focus coefficient. The focus coefficient is determined as follows: if the answer interval is normal, meaning the time interval between adjacent answers does not exceed a preset threshold, then... η i =1; If there are obvious abnormal behaviors such as being idle (no operation for a long time) or responding instantly (responding time is much shorter than the standard time), then η i Take the smaller value (e.g., 0.5).

[0141] For example, if the standard time for the above three responses is 45 seconds, 40 seconds, and 45 seconds respectively, and there is no abnormal behavior in any of them, then the effective learning time is 45 + 40 + 45 = 130 seconds.

[0142] Step S406: Calculate the student's learning rate based on the cumulative number of questions, effective learning time, and the difficulty parameters of questions in the question database that are associated with the target knowledge point.

[0143] As an optional implementation, the learning rate LR The calculation can be performed using a linear formula: LR =

[0144] in, β 1 and β 2 represents the normalized weighting coefficient. ln It is the natural logarithm function, used to smooth extreme long-tailed data and avoid excessive fluctuations in calculation results due to extreme values. V The cumulative number of questions represents the total number of questions a student needs to complete to transition from the first to the second stage. This formula means that the learning rate is directly proportional to the difficulty of the knowledge point (the more difficult the knowledge point, the higher the rate), and inversely proportional to the weighted sum of the logarithm of the cumulative number of questions and the logarithm of the effective learning time (the more questions and the longer the time spent, the lower the rate). The introduction of the logarithmic function ensures that the impact of the number of questions and the time spent on the rate exhibits a diminishing marginal effect, avoiding evaluation distortion due to extreme data from a single instance. A higher learning rate value indicates a higher efficiency in mastering the knowledge point.

[0145] In some embodiments, the method further includes: Calculate students' ability scores based on the calibrated learning data; A learning profile is generated based on ability values ​​and learning rate.

[0146] Specifically, the system calculates the student's current ability estimate based on the student's previous answer data. θIn an optional embodiment, a three-parameter logistic model from item response theory is employed, which posits that the probability of a student answering a question correctly is determined by the student's ability value. and the characteristic parameters of the question (discrimination) Difficulty Speculation The decision is made jointly. Based on the student's previous responses, the current ability estimate of the student is iteratively calculated using maximum likelihood estimation or Bayesian estimation methods. θ The range of the ability estimate is usually [-3, 3] or normalized to [0, 1]. The larger the value, the stronger the student's ability.

[0147] For example, based on a student's multiple answers to questions on the topic of "quadratic equations in one variable," the system calculates the student's ability score. θ =0.72.

[0148] Specifically, in terms of ability values θ The horizontal axis represents the learning rate. LR A two-dimensional coordinate system is constructed with the vertical axis as the ordinate, mapping each student to a corresponding position within the coordinate system. Based on the distribution of students in the coordinate system, the system categorizes students into different types and generates differentiated teaching strategy suggestions.

[0149] For example, the following classification criteria can be set: High-potential, high-achieving type: Students with abilities exceeding a preset threshold (e.g., class median) and learning rates exceeding a preset threshold. These students have a solid foundation and high learning efficiency; it is recommended to provide them with extended learning resources and challenging tasks.

[0150] High-energy, low-efficiency type: Students with ability scores above the preset threshold but learning rates below it. These students meet the ability requirements but have low learning efficiency. It is recommended to optimize learning methods and reduce ineffective practice.

[0151] Low-ability, high-efficiency type: Students with abilities below the preset threshold but learning speed above it. These students have weak foundations but high learning efficiency. It is recommended to strengthen their understanding of basic concepts and increase targeted training.

[0152] Low-ability and inefficient type: Students with abilities and learning rates below the preset threshold have weak foundations and low learning efficiency. It is recommended to strengthen basic tutoring, adjust the learning pace, and gradually improve from simple knowledge points.

[0153] For example, a student's ability value θ =0.72, learning rate LR=0.0347. If the median ability value is set to 0.65 and the median learning rate is set to 0.0300, then the student is classified as "high-potential, high-efficiency". The generated teaching suggestion is: "This student has a solid foundation and high learning efficiency. It is recommended to increase extended learning resources, such as competition problems and interdisciplinary application problems, to further stimulate potential." Furthermore, the generated learning profiles are pushed to both the teacher and student ends. The teacher end displays a scatter plot of the class's overall ability-learning rate distribution, as well as classification tags and teaching suggestions for each student; the student end displays their position in the coordinate system and a comparison with the class average, helping students understand their own learning characteristics and areas for improvement.

[0154] The dual-dimensional learning assessment method provided in this embodiment solves the problem of a single evaluation dimension in related technologies by calculating two independent dimensions: learning rate and ability value. Specifically, learning rate reflects the efficiency with which students master knowledge and can distinguish students with different potentials at the same score: for example, students with the same score of 90 may be "talented" with a high learning rate and "diligent" with a low learning rate. This dual-dimensional assessment mechanism provides a more nuanced basis for differentiated instruction, enabling teachers to develop differentiated teaching strategies based on different student types and achieve true individualized instruction.

[0155] This embodiment also provides a learning assessment device, which can be specifically applied to a data processing device. This data processing device can be a standalone server or integrated into a server or cloud platform. For example, ... Figure 5 As shown, the learning assessment device 500 may include: The data acquisition module 501 is used to acquire a first type of learning data and a second type of learning data. The first type of learning data is the answer result data collected in an uncontrolled environment, and the second type of learning data is the answer result data collected in a controlled environment. Calibration module 502 is used for: Extract at least one validation item from the second type of learning data, wherein the validation item has a pre-defined correlation with the corresponding response data in the first type of learning data; and The verification items are compared with the corresponding answer results in the first type of learning data. The confidence weight of the first type of learning data is adjusted according to the comparison results to obtain the calibrated learning data.

[0156] Please see Figure 6 This is a schematic diagram of another structure of the learning assessment system provided in the embodiments of this application. The learning assessment device 500 further includes: Forgetting prediction module 503 is used for: Based on the calibrated learning data, the learning time intervals corresponding to each knowledge point with historical answer results data are obtained. The historical answer results data include the answer results data associated with the knowledge points in the first type of learning data and the second type of learning data. The attenuation coefficient of each knowledge point is calculated based on the preset attenuation function; Based on the attenuation coefficient of each knowledge point, test questions are selected from a pre-set test question database to generate a test set containing three types of test question data: the first type corresponds to knowledge points for which no historical answer data has appeared, the second type corresponds to test questions marked as incorrect in the historical answer data, and the third type corresponds to knowledge points for which the attenuation coefficient is lower than a pre-set threshold.

[0157] The two-dimensional evaluation module 504 is used for: Based on the calibrated learning data, obtain the sequence of students' answer results for the target knowledge points; The cumulative number of questions and effective learning time corresponding to the student's transition from the first state to the second state are determined based on the sequence of answer results data. Specifically, the first state is defined as the percentage of incorrect answers in a consecutive preset number of answers that is higher than a first threshold, and the second state is defined as the percentage of correct answers in a consecutive preset number of answers that is higher than a second threshold. The cumulative number of questions is the number of answer results data included between the first and second states, and the effective learning time is calculated based on the answer time in the answer results data included between the first and second states. The student's learning rate is calculated based on the cumulative number of questions, effective learning time, and the difficulty parameters of questions in the question bank that are related to the target knowledge points. Calculate student ability estimates based on calibrated learning data; and A learning profile is generated based on the estimated ability and learning rate.

[0158] In some embodiments, the calibration module 502 is specifically used to perform: Obtain the correct answer data from the first type of learning data as the seed set; Based on the feature values ​​of the preset association relationship, candidate verification items that match the answer result data in the seed set are retrieved from the preset test question database to form a candidate verification item set; Based on the student's current ability estimate, candidate verification items with information content greater than the preset information content threshold are selected from the candidate verification item set as verification items; In this database, the test questions are pre-constructed with feature vectors, each containing at least one feature value representing a relationship.

[0159] In some embodiments, the calibration module 502 is specifically used to perform: Obtain the first accuracy rate of the verification item and the second accuracy rate of the corresponding answer result data in the first learning data; Calculate the consistency coefficient between the first accuracy and the second accuracy; If the consistency coefficient is greater than the preset consistency threshold, the confidence weight of the first type of learning data is increased; If the consistency coefficient is less than the consistency threshold and the first accuracy is less than the second accuracy, then reduce the confidence weight of the first type of learning data. If the consistency coefficient is less than the consistency threshold and the first accuracy is greater than the second accuracy, then the confidence weight of the first type of learning data remains unchanged.

[0160] In some embodiments, the calibration module 502 is further configured to perform: After reducing the confidence weight of the first type of learning data, the update of the student's ability value based on the corresponding answer result data in the first type of learning data is withdrawn.

[0161] In some embodiments, the learning assessment device 500 further includes an early warning response module, which, after keeping the confidence weight of the first type of learning data unchanged, sends a prompt message to the teacher, adds a status label to the student's learning record, and sends a comparison report to the parent; wherein, the prompt message is used to indicate that there is a difference in the accuracy rate between the first type of learning data and the second type of learning data, the status label is used to identify that the answer result data corresponding to the first type of learning data is of the attitude fluctuation type, and the comparison report is used to display the accuracy rate comparison between the first type of learning data and the second type of learning data.

[0162] In some embodiments, the forgetting prediction module 503 is specifically configured to perform: Obtain the initial memory strength parameters for each knowledge point; The decay coefficient is calculated based on the ratio of the learning time interval to the initial memory strength parameter; If the answer results data corresponding to the third type of test questions match the corresponding preset answer information, then the memory strength parameter of the corresponding knowledge point will be increased according to the preset growth coefficient. If the answer results data corresponding to the third type of test questions do not match the corresponding preset answer information, the memory strength parameter of the corresponding knowledge point will be reduced according to the preset attenuation coefficient.

[0163] In some embodiments, the learning assessment device 500 further includes a spiral test paper generation module and a report generation module, wherein: The spiral test set module is used to receive the test set generated by the forgetting prediction module and push it to the teacher's or student's end. The report generation module receives the learning profile generated by the dual-dimensional evaluation module and pushes it to the teacher's or student's end.

[0164] In practice, each of the above modules can be implemented as an independent entity or can be combined arbitrarily to be implemented as the same or several entities. For the specific implementation methods and corresponding beneficial effects of each of the above modules, please refer to the previous method embodiments, which will not be repeated here.

[0165] Accordingly, this application also provides an electronic device, which can be a terminal, such as a smartphone, tablet computer, laptop computer, touch screen, game console, personal computer (PC), personal digital assistant (PDA), or other terminal device. Alternatively, the electronic device can be a server.

[0166] like Figure 7 As shown, Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device 700 includes a processor 701 with one or more processing cores, a memory 702 with one or more computer-readable storage media, and a computer program stored on the memory 702 and executable on the processor. The processor 701 and the memory 702 are electrically connected. Those skilled in the art will understand that the electronic device structure shown in the figure does not constitute a limitation on the electronic device, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0167] The processor 701 is the control center of the electronic device 700. It connects various parts of the electronic device 700 via various interfaces and lines. By running or loading software programs and / or units stored in the memory 702, and by calling data stored in the memory 702, it executes various functions and processes data of the electronic device 700, thereby providing overall monitoring of the electronic device 700. The processor 701 can be a central processing unit (CPU), a graphics processing unit (GPU), a network processor (NP), etc., and can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0168] In this embodiment, the processor 701 in the electronic device 700 loads the instructions corresponding to the processes of one or more applications into the memory 702 according to the following steps, and the processor 701 runs the applications stored in the memory 702 to realize various functions, such as: Acquire a first type of learning data and a second type of learning data. The first type of learning data is the answer result data collected in an uncontrolled environment, and the second type of learning data is the answer result data collected in a controlled environment. At least one verification item is extracted from the second type of learning data, and the verification item has a preset correlation with the corresponding answer result data in the first type of learning data; The verification item is compared with the corresponding answer result data in the first type of learning data, and the confidence weight of the first type of learning data is adjusted according to the comparison result to obtain the calibrated learning data.

[0169] The specific implementation of each of the above operations and their corresponding beneficial effects can be found in the previous embodiments, and will not be repeated here.

[0170] Optional, such as Figure 7 As shown, the electronic device 700 also includes: a touch display screen 703, a radio frequency circuit 704, an audio circuit 705, an input unit 706, and a power supply 707. The processor 701 is electrically connected to the touch display screen 703, the radio frequency circuit 704, the audio circuit 705, the input unit 706, and the power supply 707. Those skilled in the art will understand that... Figure 7 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0171] The touch display screen 703 can be used to display a graphical user interface (GUI) and receive operation commands generated by the user interacting with the GUI. The touch display screen 703 may include a display panel and a touch panel. The display panel can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces of the electronic device. These graphical user interfaces can be composed of graphics, text, icons, video, and any combination thereof. Optionally, the display panel can be configured using a liquid crystal display (LCD), an organic light-emitting diode (OLED), or other similar technology. The touch panel can be used to collect touch operations performed by the user on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near the touch panel), generate corresponding operation commands, and execute the corresponding program according to the operation commands. Optionally, the touch panel may include two parts: a touch detection device and a touch controller. The touch detection device detects the user's touch location and the signal generated by the touch operation, transmitting the signal to the touch controller. The touch controller receives touch information from the touch detection device, converts it into touch point coordinates, and sends it to the processor 701. It can also receive and execute commands from the processor 701. The touch panel can cover the display panel. When the touch panel detects a touch operation on or near it, it transmits the information to the processor 701 to determine the type of touch event. Subsequently, the processor 701 provides corresponding visual output on the display panel based on the type of touch event. In this embodiment, the touch panel and the display panel can be integrated into the touch display screen 703 to achieve input and output functions. However, in some embodiments, the touch panel and the touch display screen 703 can be implemented as two independent components to achieve input and output functions. That is, the touch display screen 703 can also be used as part of the input unit 706 to achieve input functions.

[0172] The radio frequency circuit 704 can be used to transmit and receive radio frequency signals to establish wireless communication with network devices or other electronic devices, and to transmit and receive signals with network devices or other electronic devices.

[0173] Audio circuitry 705 can be used to provide an audio interface between a user and an electronic device via a speaker and a microphone. Audio circuitry 705 converts received audio data into electrical signals, transmits them to the speaker, and the speaker converts them into sound signals for output. Conversely, the microphone converts collected sound signals into electrical signals, which are then received by audio circuitry 705, converted back into audio data, and then processed by processor 701 before being transmitted via radio frequency circuitry 704 to, for example, another electronic device, or output to memory 702 for further processing. Audio circuitry 705 may also include an earphone jack to facilitate communication between peripheral headphones and electronic devices.

[0174] The input unit 706 can be used to receive input numbers, characters, or user characteristic information (such as fingerprints, iris, facial information, etc.), and to generate keyboard, mouse, joystick, optical, or trackball signal inputs related to user settings and function control.

[0175] Power supply 707 is used to supply power to various components of electronic device 700. Optionally, power supply 707 can be logically connected to processor 701 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. Power supply 707 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0176] although Figure 7 As not shown in the diagram, the electronic device 700 may also include a camera, sensor, wireless fidelity module, Bluetooth module, etc., which will not be described in detail here.

[0177] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0178] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0179] Therefore, embodiments of this application provide a computer-readable storage medium storing multiple computer programs that can be loaded by a processor to execute any of the video recognition methods provided in this application. For example, the computer program can execute the steps of the following video recognition method: Acquire a first type of learning data and a second type of learning data. The first type of learning data is continuously collected answer result data, and the second type of learning data is answer result data collected in a controlled environment. At least one verification item is extracted from the second type of learning data, and the verification item has a preset correlation with the corresponding answer result data in the first type of learning data; The verification item is compared with the corresponding answer result data in the first type of learning data, and the confidence weight of the first type of learning data is adjusted according to the comparison result to obtain the calibrated learning data.

[0180] The specific implementation of each of the above operations and their corresponding beneficial effects can be found in the previous embodiments, and will not be repeated here.

[0181] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0182] Since the computer program stored in the computer-readable storage medium can execute any of the learning assessment methods provided in the embodiments of this application, it can achieve the beneficial effects that any of the learning assessment methods provided in the embodiments of this application can achieve, as detailed in the preceding embodiments, and will not be repeated here.

[0183] According to one aspect of this application, a computer program product or computer program is also provided, comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the methods provided in the various optional implementations of the above embodiments.

[0184] According to one aspect of this application, a learning assessment system 800 is also provided, such as... Figure 8 As shown, it includes any of the electronic devices 700 provided in the embodiments of this application; it also includes: The terminal acquisition layer 801, which is communicatively connected to the electronic device 700, includes: The uncontrolled acquisition terminal 8011 is used to acquire the first type of learning data, which is the answer result data collected in an uncontrolled environment; The controlled acquisition terminal 8012 is used to acquire the second type of learning data, which is the answer result data collected in a controlled environment.

[0185] Optionally, the uncontrolled data acquisition terminal 8011 can be deployed on student terminal devices (such as mobile phones, tablets, and teaching tablets) to collect data on students' answers to homework assignments and self-study exercises generated during their daily learning process. This type of data is collected frequently and can reflect the students' learning process, but due to the uncontrollable environment (such as students potentially checking answers with each other or using question-searching software), its initial confidence weight is set to a low value.

[0186] Optionally, the controlled data acquisition terminal 8012 can be deployed in examination room terminal equipment or teacher terminal equipment to collect students' answer results data in controlled environments such as weekly tests, monthly tests, quarterly tests, semester tests, in-class quizzes (proctored by teachers), experimental operation assessments, and offline interviews. This type of data acquisition environment is controlled and can truly reflect students' levels, with its initial confidence weight set to a relatively high value.

[0187] In some embodiments, the system 800 further includes a data processing layer 802, in which an electronic device 700 is integrated. The electronic device 700 is communicatively connected to a terminal acquisition layer 801 and is used to perform the learning assessment steps in the aforementioned method embodiments, including: Validation items are extracted from the second type of learning data. The validation items are compared with the corresponding answer results in the first type of learning data. The confidence weight of the first type of learning data is adjusted according to the comparison results to obtain the calibrated learning data. A test set containing three types of test questions is generated based on the calibrated learning data. Based on the calibrated learning data, students' learning rates and abilities are calculated, and learning profiles are generated.

[0188] In some embodiments, the system 800 further includes: Application service layer 803, which is communicatively connected to electronic device 700, includes: The spiral test set module 8031 ​​is used to receive the test set generated by the electronic device 700 and push it to the teacher's or student's end. The report generation module 8032 is used to receive the learning profile generated by the electronic device 700 and push it to the teacher's or student's end.

[0189] Optionally, the spiral test creation module 8031 ​​can push a personalized test set containing questions on new knowledge points, variations of historical incorrect questions, and questions on critical forgetting points to the teacher's end for classroom use, or directly to the student's end for self-study. The report generation module 8032 can push a learning profile containing ability values, learning rates, and two-dimensional classification labels to the teacher's end in the form of radar charts, scatter plots, or reports for the teacher's reference in teaching decisions; it can also push it to the student's end to help students understand their own learning characteristics and areas for improvement.

[0190] The learning assessment system provided in this embodiment achieves decoupling of data acquisition, data processing, and application services through a layered architecture. The terminal acquisition layer is responsible for acquiring multi-source data, the data processing layer is responsible for executing the core algorithm, and the application service layer is responsible for outputting the results. These layers collaborate through communication connections. This system can calibrate data from uncontrolled environments using high-reliability data acquired in a controlled environment, generate personalized test sets by combining forgetting curves, and generate learning profiles through two-dimensional evaluation, thereby effectively improving the accuracy of learning assessment and the scientific nature of teaching decisions.

[0191] In the above embodiments of the learning assessment device, equipment, computer-readable storage medium, computer program product, and learning assessment system, the descriptions of each embodiment have different focuses. Parts not detailed in a particular embodiment can be found in the relevant descriptions of other embodiments. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes and beneficial effects of the learning assessment system, computer-readable storage medium, computer program product, electronic device, and their corresponding units described above can be referred to the description of the video recognition method in the above embodiments, and will not be repeated here.

[0192] The foregoing has provided a detailed description of a learning assessment method, apparatus, device, computer-readable storage medium, computer program product, and learning assessment system provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for assessing learning progress, characterized in that, include: Acquire a first type of learning data and a second type of learning data. The first type of learning data is the answer result data collected in an uncontrolled environment, and the second type of learning data is the answer result data collected in a controlled environment. At least one verification item is extracted from the second type of learning data, and the verification item has a preset correlation with the corresponding answer result data in the first type of learning data; The verification item is compared with the corresponding answer result data in the first type of learning data, and the confidence weight of the first type of learning data is adjusted according to the comparison result to obtain the calibrated learning data.

2. The method according to claim 1, characterized in that, The step of extracting at least one verification item from the second type of learning data includes: Obtain the correct answer data from the first type of learning data as the seed set; Based on the feature values ​​of the preset association relationship, candidate verification items that match the answer result data in the seed set are retrieved from the preset test question database to form a candidate verification item set; Based on the student's current ability estimate, candidate verification items with information content greater than a preset information content threshold are selected from the candidate verification item set as the verification items; In this context, the test questions in the test question database are pre-constructed with feature vectors, and the feature vectors contain at least one feature value used to represent the association relationship.

3. The method according to claim 2, characterized in that, The step of selecting candidate verification items with information content greater than a preset information content threshold from the candidate verification item set based on the student's current ability estimate includes: Obtain an estimate of the student's current ability; For each candidate validation item in the candidate validation item set, calculate the information content value of the candidate validation item at the capability estimate; Candidate verification items whose information content value is greater than the information content threshold are selected as the verification items.

4. The method according to claim 1, characterized in that, The preset association relationship includes at least one of the following: The knowledge point identifier corresponding to the verification item is the same as the knowledge point identifier corresponding to the answer result data; The cognitive classification level corresponding to the verification item is the same as the cognitive classification level corresponding to the answer result data; The difference between the difficulty parameter corresponding to the verification item and the difficulty parameter corresponding to the answer result data is less than a preset threshold; The format of the test questions corresponding to the verification items is the same as the format of the test questions corresponding to the answer result data.

5. The method according to claim 1, characterized in that, The step of comparing the verification item with the corresponding answer result data in the first type of learning data, and adjusting the confidence weight of the first type of learning data according to the comparison result to obtain the calibrated learning data includes: Obtain the first accuracy rate of the verification item and the second accuracy rate of the corresponding answer result data in the first type of learning data; Based on the difference between the first accuracy rate and the second accuracy rate, the confidence weight of the first type of learning data is determined to obtain the calibrated learning data; The first accuracy rate is calculated based on the correctness identifier of the verification item, and the second accuracy rate is calculated based on the correctness identifier of the corresponding answer result data in the first type of learning data.

6. The method according to claim 5, characterized in that, The step of determining the confidence weight of the first type of learning data based on the difference between the first accuracy and the second accuracy includes: Calculate the consistency coefficient between the first accuracy rate and the second accuracy rate; If the consistency coefficient is greater than the preset consistency threshold, then the confidence weight of the first type of learning data is increased; If the consistency coefficient is less than the consistency threshold and the first accuracy is less than the second accuracy, then the confidence weight of the first type of learning data is reduced. If the consistency coefficient is less than the consistency threshold and the first accuracy is greater than the second accuracy, then the confidence weight of the first type of learning data remains unchanged.

7. The method according to claim 6, characterized in that, After reducing the confidence weight of the first type of learning data, the method further includes: Cancel the update of the student's ability estimate based on the corresponding answer result data in the first type of learning data.

8. The method according to claim 6, characterized in that, After keeping the confidence weights of the first type of learning data unchanged, the method further includes: Generate early warning marker information; Send a corresponding prompt message to the teacher's end, the prompt message being used to indicate that there is a difference in the accuracy rate between the first type of learning data and the second type of learning data; Add status tags to student learning portfolios; these status tags are used to identify whether the response data corresponding to the first type of learning data is of the attitude fluctuation type; or A comparison report is sent to the parents, which shows the accuracy comparison between the first type of learning data and the second type of learning data.

9. The method according to claim 1, characterized in that, Also includes: Based on the calibrated learning data, the learning time interval corresponding to each knowledge point with historical answer result data is obtained. The historical answer result data includes the answer result data associated with the knowledge point in the first type of learning data and the second type of learning data. The decay coefficient of each knowledge point is calculated according to a preset decay function, which is used to characterize the mapping relationship between the learning time interval and the memory strength parameter. Based on the attenuation coefficient of each knowledge point, test questions are selected from a preset test question database to generate a test set containing first type test question data, second type test question data, and third type test question data. The first type test question data corresponds to knowledge points for which no historical answer data has appeared, the second type test question data corresponds to test questions in the historical answer data that are marked as incorrect, and the third type test question data corresponds to knowledge points for which the attenuation coefficient is lower than a preset threshold.

10. The method according to claim 9, characterized in that, Also includes: If the answer result data corresponding to the third type of test question data matches the corresponding preset answer information, then the memory strength parameter of the corresponding knowledge point is increased according to the preset growth coefficient; If the answer result data corresponding to the third type of test question data does not match the corresponding preset answer information, the memory strength parameter of the corresponding knowledge point will be reduced according to the preset attenuation coefficient.

11. The method according to claim 1, characterized in that, Also includes: Based on the calibrated learning data, a sequence of students' answer results for the target knowledge points is obtained, including the correctness indicators and answer duration of multiple answers. Based on the sequence of answer results, determine the cumulative number of questions and effective learning time corresponding to the student's transition from the first state to the second state; The student's learning rate is calculated based on the cumulative number of questions, the effective learning time, and the difficulty parameters of the questions in the question database associated with the target knowledge point. In this context, the first state is defined as the percentage of incorrect answers in a consecutive preset number of responses exceeding a first threshold, the second state is defined as the percentage of correct answers in a consecutive preset number of responses exceeding a second threshold, the cumulative number of questions is the number of answer result data included between the first state and the second state, and the effective learning time is calculated based on the answer time in the answer result data included between the first state and the second state.

12. The method according to claim 11, characterized in that, Also includes: Calculate the student's ability value based on the calibrated learning data; A learning profile is generated based on the ability value and the learning rate.

13. An electronic device, characterized in that, The system includes a processor and a memory, the memory storing multiple instructions; the processor loads instructions from the memory to execute the steps of the learning assessment method as described in any one of claims 1 to 12.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to perform the steps of the learning assessment method as described in any one of claims 1 to 12.

15. A learning assessment system, characterized in that, include: The electronic device as claimed in claim 13; The terminal acquisition layer, which is communicatively connected to the electronic device, includes: An uncontrolled acquisition terminal is used to acquire the first type of learning data, which is the answer result data collected in an uncontrolled environment; The controlled acquisition terminal is used to acquire the second type of learning data, which is the answer result data collected in a controlled environment.

Images