Personalized proposition generation system and method based on dynamic modeling of report after examination

Abstract

The invention relates to the technical field of proposition generation, and particularly discloses a personalized proposition generation system and method based on post-examination report dynamic modeling, and the system comprises a modeling module which constructs a personalized to-be-inspected point determination model and a personalized multi-dimensional capability evaluation model; the model output module is used for obtaining the newest personalized to-be-investigated examination site and the newest multi-dimensional ability evaluation result of the examinee in each region based on the newest post-examination report, the personalized to-be-investigated site determination model and the personalized multi-dimensional ability evaluation model; the outlier marking module marks the examine site mastering outlier branch path of the examinees in each region; the test paper generation module obtains personalized test paper of the examinees in each region based on the examination point mastering outlier branch path of the examinees in each region, the latest personalized examination point to be examined, the latest multi-dimensional ability evaluation result and a pre-trained personalized proposition model; therefore, the generated test paper fully considers the regional difference of the learning power levels of the examinees in different regions.

Description

Technical Field

[0001] The present invention relates to the technical field of question generation, and in particular to a personalized question generation system and method based on dynamic modeling of post-exam reports. Background Art

[0002] In education, examinations are a crucial way to assess students' learning outcomes and teachers' teaching effectiveness. Traditional test-setting methods often use standardized exam papers, which make it difficult to account for regional differences in students' academic abilities and accurately meet the learning needs of individual students. With the shift in educational philosophy toward personalized education, developing systems capable of generating personalized exams has become crucial. A personalized question generation system based on dynamic modeling of post-exam reports aims to analyze post-exam reports and dynamically construct models to accurately understand the learning progress of students in each region, thereby generating personalized exams tailored to their specific needs. This system leverages a large number of historical post-exam reports to explore students' unique characteristics in terms of knowledge acquisition and ability development, which is of great significance for improving assessment quality. From a development perspective, with the widespread application of technologies such as big data and artificial intelligence in education, this system is expected to become a key component of future educational examination models, driving education towards greater precision and personalization.

[0003] However, existing question generation methods struggle to accurately analyze the test points and multi-dimensional ability assessment results for each region's candidates based on post-test reports. Regional differences in candidate mastery of these key points are not analyzed in detail enough, making it impossible to generate personalized exams based on effective models that truly meet the needs of candidates in each region, making it difficult to achieve the goal of personalized education.

[0004] Therefore, the present invention proposes a personalized question generation system and method based on dynamic modeling of post-exam reports. Summary of the Invention

[0005] The present invention provides a personalized question generation system and method based on dynamic modeling of post-exam reports. The system fully considers the regional differences in the academic ability levels of candidates in different regions, making the generated test papers highly targeted and effectively helping candidates in different regions consolidate their knowledge, improve their abilities, and enhance the quality of questions and assessment results.

[0006] The present invention provides a personalized question generation system based on dynamic modeling of post-exam reports, comprising: Modeling module, used to build personalized test point determination models and personalized multi-dimensional ability assessment models based on a large number of historical test reports; The model output module is used to determine the model and the personalized multi-dimensional ability assessment model based on the latest post-test report and personalized test points to be assessed, and obtain the latest personalized test points to be assessed and the latest multi-dimensional ability assessment results for candidates in each region; The outlier marking module is used to mark the outlier branch paths of the examinees' mastery of test points in each region based on the time series data of the examinees' mastery of all test points involved in all history examinations of a single academic period and a single subject, and the mastery expansion map of the standard test points of the corresponding subject in the corresponding academic period; The test paper generation module is used to obtain personalized test papers for candidates in each region based on the test points of the candidates in each region, grasp the outlier branch paths, the latest personalized test points to be examined, the latest multi-dimensional ability assessment results and the pre-trained personalized proposition model.

[0007] Optionally, the modeling module includes: The first modeling submodule is used to analyze the official difficulty distribution and macro difficulty distribution of all test points involved in all history exams based on a large number of history exam reports, as well as the differential difficulty distribution of all test points involved in a large number of history exams for each region. It also combines the test points to be tested in each region as reflected in the corresponding scoring results of candidates in each region, as identified by all question setting experts, with a deep learning algorithm to construct a personalized model for determining test points to be tested. The second modeling sub-module is used to construct a personalized multidimensional ability assessment model by utilizing all history test papers in a large number of history post-examination reports, the scoring results of the corresponding history test papers obtained by candidates in each region, the multidimensional ability assessment results of candidates in each region under the corresponding scoring results calibrated by all question experts, and deep learning algorithms.

[0008] Optionally, the first modeling submodule includes: The first difficulty distribution analysis unit is used to analyze the official difficulty distribution of all test points involved in each history test paper based on the evaluation results of all test experts on the test points of all history test questions in each history test paper in a large number of history test post-test reports; A mastery distribution analysis unit is used to treat the ratio of the average score of each history question in each history test paper by all examinees in each region in a large number of history test reports to the full score of the corresponding history question as the mastery of the examinees in the corresponding region on the corresponding history question, and to generate the mastery distribution of each history question for all examinees in all regions based on the mastery of each history question by examinees in all regions; The second difficulty distribution analysis unit is used to analyze the mastery distribution of each history question and the distribution of test points in each history test paper by examinees from all regions in a large number of history test reports, as well as the answer records of each history question by examinees from each region, to obtain the macro difficulty distribution of all test points involved in each history test paper and the differential difficulty distribution of all test points involved in each history test paper by examinees from each region; The examination point determination model modeling unit is used to use deep learning algorithms and the official difficulty distribution of all examination points involved in all history examination papers, the macro difficulty distribution, the differential difficulty distribution of all examination points involved in a large number of history examination papers for candidates in each region, and all the examination points to be examined reflected in the corresponding scoring results of candidates in each region marked by all question experts based on each scoring result, to build a personalized examination point determination model.

[0009] Optionally, the first difficulty distribution analysis unit includes: An examination difficulty determination unit is used to obtain the examination difficulty of each history test paper for each examination point involved, as assessed by each question setter, from the evaluation results of the examination points of all history test questions in each history test paper by all question setters in a large number of history test post-exam reports; The official difficulty distribution generating unit is used to take the average of the examination difficulty of each test point involved in each history test paper evaluated by all question experts as the official difficulty of each test point involved in each history test paper, and generate the official difficulty distribution of all test points involved in each history test paper based on the official difficulty of all test points involved in each history test paper.

[0010] Optionally, the second difficulty distribution analysis unit includes: The test question mastery correction subunit is used to correct the mastery distribution of the corresponding history test questions by the examinees in all regions based on the average answering order and average answering time of the answering records of each history test question of all examinees in each region, and generate a corrected mastery distribution of the corresponding history test questions by the examinees in all regions; The test point mastery determination subunit is used to analyze the actual mastery of each test point involved in each history test paper by candidates in each region based on the corrected mastery distribution of each history test question and the test point distribution of each history test paper; The macro-difficulty distribution determination subunit is used to treat the difference between 1 and the mean of the actual mastery of a single test point by test takers in all regions as the macro-difficulty of the single test point, and to generate a macro-difficulty distribution of all test points involved in each history test paper based on the macro-difficulty of all test points involved in each history test paper. The differential difficulty distribution determination subunit is used to obtain the differential difficulty distribution of all test points involved in each history test paper by candidates in each region based on the actual mastery of each test point involved in each history test paper and the macro difficulty distribution of all test points involved in each history test paper.

[0011] Optionally, the differential difficulty distribution determination subunit includes: The personalized difficulty determination terminal is used to use the difference between 1 and the actual mastery of each test point involved in each history test paper by the examinees in each region as the personalized difficulty of each test point involved in each history test paper by the examinees in each region; The personalized difficulty distribution determination terminal is used to aggregate the personalized difficulty of all test points involved in each history test paper for each region's candidates, and generate a personalized difficulty distribution of all test points involved in each history test paper for each region's candidates; The differential difficulty distribution determination end is used to subtract the personalized difficulty distribution of all test points involved in each history test paper for candidates in each region from the macro difficulty distribution of all test points involved in each history test paper, so as to obtain the differential difficulty distribution of all test points involved in each history test paper for candidates in each region.

[0012] Optionally, the model output module includes: The third difficulty distribution analysis submodule is used to determine the official difficulty distribution of all test points involved in the latest test paper, the macro difficulty distribution, and the differential difficulty distribution of all test points involved in the latest test paper for candidates in each region based on the latest post-test report; The first execution submodule is configured to input the official difficulty distribution of all test points involved in the latest test paper, the macro difficulty distribution, and the differential difficulty distribution of all test points involved in the latest test paper by candidates in each region into the personalized test point determination model to obtain the latest personalized test points to be examined for candidates in each region; The second execution submodule is used to input the scoring results of the latest test papers obtained by candidates in each region in the latest test papers and the latest post-test reports into the personalized multidimensional ability assessment model to obtain the latest multidimensional ability assessment results of candidates in each region.

[0013] Optionally, the outlier marking module includes: The dynamic test point mastery path marking submodule is used to mark the dynamic test point mastery path of candidates in the corresponding region in the standard test point mastery expansion map of the corresponding subject in the corresponding academic period based on the time series data of the mastery of all test points involved in all history examinations of a single academic period and a single subject in each region; A mastery lag factor determination submodule is used to determine the standard mastery moment for each test point based on the standard dynamic test point mastery expansion map for the corresponding subject within the corresponding academic period. Simultaneously, based on the dynamic test point mastery paths of the test takers in each region, the individualized mastery moment for each test point for the test takers in each region is determined. Based on the individualized mastery moment for each test point in each region and the standard mastery moment for the corresponding test point, the mastery lag factor for each test point in the standard dynamic test point mastery path is determined for the test takers in each region. The dynamic fitting submodule is used to dynamically fit the mastery lag factor of each test point in the standard dynamic test point mastery path of the examinees in each region, and obtain the test point mastery outlier branch path of the examinees in each region.

[0014] Optionally, the test paper generation module includes: A comprehensive hysteresis factor calculation submodule is used to calculate the comprehensive hysteresis factor of the examinees in each region based on the outlier branch path of the examinees' mastery of the test points in each region; A supplementary test site determination submodule is used to screen out supplementary test sites for candidates in each region from the test site mastery outlier branch path based on the comprehensive lag factor of candidates in each region; The personalized test paper generation submodule is used to input the latest personalized test points to be examined, supplementary test points to be examined, and the latest multi-dimensional ability assessment results of candidates in each region into the pre-trained personalized proposition model to obtain personalized test papers for candidates in each region.

[0015] The present invention provides a personalized proposition generation method based on dynamic modeling of post-exam reports, comprising: Step S1: Construct a personalized model for determining test points and a personalized multi-dimensional ability assessment model based on a large number of historical post-test reports; Step S2: Based on the latest post-test report, the personalized test point determination model, and the personalized multi-dimensional ability assessment model, the latest personalized test points to be examined and the latest multi-dimensional ability assessment results of the examinees in each region are obtained; Step S3: Based on the time series data of the mastery of all test points involved in all history examinations of a single academic period and a single subject by examinees in each region and the mastery expansion map of the standard test points of the corresponding subject in the corresponding academic period, mark the outlier branch paths of the test point mastery of examinees in each region; Step S4: Based on the test points of the candidates in each region, the outlier branch path, the latest personalized test points to be examined, the latest multi-dimensional ability assessment results and the pre-trained personalized proposition model are mastered to obtain the personalized test papers of the candidates in each region.

[0016] The beneficial effects of the present invention compared to the prior art are as follows: taking regions as units of consideration, building relevant models based on a large number of historical post-test reports, and being able to analyze the learning situation of many candidates in the same region in batches and efficiently, providing targeted learning guidance and test paper generation for candidates in the entire region, compared to individual analysis, improving overall efficiency and meeting the needs of large-scale assessment scenarios. The model output module obtains the latest personalized test points to be examined and multi-dimensional ability assessment results for candidates in each region based on the latest post-test report, personalized test point determination model and personalized multi-dimensional ability assessment model, and can accurately understand the overall characteristics and trends of the candidates in the region in terms of knowledge mastery and ability development. The outlier marking module marks the outlier branch paths of the test point mastery of candidates in each region through the time series data of the candidates' test point mastery and the standard test point mastery expansion map, which can intuitively show the common paths that the candidates in the region deviate from when studying specific test points. The test paper generation module combines the test points of regional candidates to grasp the outlier branch path, the latest personalized test points to be examined, the latest multi-dimensional ability assessment results and the personalized proposition model to generate personalized test papers, ensuring that the test papers can meet the overall learning level and needs of the candidates in the region, and promote the overall improvement of regional assessment quality.

[0017] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in this application document.

[0018] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 Schematic diagram of a personalized question generation system based on dynamic modeling of post-test reports in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0021] The present invention provides an implementation method of a personalized proposition generation system based on dynamic modeling of post-exam reports, such as Figure 1 Shown, including: Modeling module, used to build personalized test point determination models and personalized multi-dimensional ability assessment models based on a large number of historical test reports; The model output module is used to determine the model and the personalized multi-dimensional ability assessment model based on the latest post-test report and personalized test points to be assessed, and obtain the latest personalized test points to be assessed and the latest multi-dimensional ability assessment results for candidates in each region; The outlier marking module is used to mark the outlier branch paths of the examinees' mastery of test points in each region based on the time series data of the examinees' mastery of all test points involved in all history examinations of a single academic period and a single subject, and the mastery expansion map of the standard test points of the corresponding subject in the corresponding academic period; The test paper generation module is used to obtain personalized test papers for candidates in each region based on the test points of the candidates in each region, grasp the outlier branch paths, the latest personalized test points to be examined, the latest multi-dimensional ability assessment results and the pre-trained personalized proposition model.

[0022] In this embodiment: History post-exam report: refers to the report generated after a large number of exams in the past, covering information such as the examination experts' inspection and evaluation of the history exam test points, the scores and records of the candidates' answers in each region, etc., providing data for building relevant models.

[0023] Personalized model for determining test points to be examined: This model is constructed based on a large number of historical post-exam reports, taking into account the official, macro, and regional distribution of test points, as well as the test points to be examined for test takers in different regions as determined by question experts. It is developed with the help of a deep learning algorithm and is used to fully consider the personalized differences in test point mastery among test takers in different regions, and determine more targeted and personalized test points to be examined for test takers in different regions.

[0024] Personalized multi-dimensional ability assessment model: Utilizing a large number of history test reports, combined with history test papers, regional candidate scores, and multi-dimensional ability assessment results calibrated by question experts and constructed using deep learning algorithms, it is used to assess the multi-dimensional abilities of regional candidates and output the latest assessment results. For example, the Chinese language subject can assess the multi-dimensional abilities of regional candidates in reading, writing, etc., or the logical thinking ability, knowledge application ability, and memory ability of regional candidates.

[0025] Among them, regional candidates refer to candidates within a single region.

[0026] Latest post-test report: A report generated after the most recent test, including the latest test locations, answers and scores of candidates in a single region, etc., which is input into the constructed model to obtain the latest personalized test locations and multi-dimensional ability assessment results for candidates in a single region (the latest personalized test locations and multi-dimensional ability assessment results for regional candidates obtained by this model output are more accurate than the test locations and multi-dimensional abilities of regional candidates assessed by question experts based on their answers on the test papers), reflecting the current learning status.

[0027] The latest personalized test points to be examined: The difficulty distribution of the relevant test points in the latest post-test report is input into the personalized test point determination model. It is based on the latest test situation of the candidates in the region and the model analysis. It is targeted and is one of the bases for generating personalized test papers. For example, in the physics subject, if the candidates in the region do not have a good grasp of the circuit test points, the model may list the relevant specific test points as the latest test points to be examined.

[0028] The latest multi-dimensional ability assessment results: The test scores obtained by the regional candidates in the latest test papers and post-test reports are input into the personalized multi-dimensional ability assessment model to reflect the multi-dimensional ability status of the regional candidates after the latest test, helping the question setting experts and regional candidates to understand the performance of the regional candidates in various ability dimensions. For example, the English subject can present the candidates' listening, reading and other abilities.

[0029] Time series data on the mastery of all test points across all history exams for a single subject in a single academic period by examinees in each region: This data records the changes over time in examinees' mastery of each test point across all history exams for a single subject in a specific academic period. For example, by recording examinees' mastery of test points such as functions and geometry across all three years of high school mathematics, this data, which changes over time, can be used to analyze the developmental trends in examinees' mastery of each test point.

[0030] Standardized Exam Point Mastery Map: This map describes the relationships between exam points within a specific academic period, along with standard information such as the sequence (i.e., path) and timing for mastering these points. Like a disciplinary knowledge map, it illustrates the sequence and timing of exam points mastered by students under standard circumstances, providing a reference for analyzing individual regional student mastery.

[0031] Outlier branch paths for test-point mastery: Using time-series data on regional examinees' mastery of all history test points across a single academic period and subject, we plotted their dynamic test-point mastery paths on the standard test-point mastery expansion map. By comparing these with the standard dynamic test-point mastery paths, we determined the mastery lag factor for each test point, and then dynamically fitted these factors. This reveals where examinees in the region deviate from the standard mastery path during their learning process.

[0032] Pre-trained personalized question setting model: This model, trained on a large amount of data, can generate test questions tailored to the specific circumstances of each region's candidates based on their latest personalized test points, supplementary test points, and the latest multi-dimensional ability assessment results. This allows for more accurate assessment of regional candidates' mastery of test points. For example, personalized test papers with appropriate difficulty and question types are generated based on the differences in regional candidates' mastery and ability levels across different test points. Compared to traditional test papers, this approach can better assess regional candidates' current mastery of test points by distinguishing between their different mastery levels.

[0033] Personalized exam papers: Based on the latest personalized test points, supplementary test points, latest personalized test points, and the latest multi-dimensional ability assessment results of regional candidates, a pre-trained personalized question setting model is used to generate personalized exam papers. This fully considers the academic differences of candidates in different regions and meets the specific assessment needs of regional candidates.

[0034] In an alternative embodiment, the modeling module includes: The first modeling submodule is used to analyze the official difficulty distribution and macro difficulty distribution of all test points involved in all history exams based on a large number of history exam reports, as well as the differential difficulty distribution of all test points involved in a large number of history exams for each region. It also combines the test points to be tested in each region as reflected in the corresponding scoring results of candidates in each region, as identified by all question setting experts, with a deep learning algorithm to construct a personalized model for determining test points to be tested. The second modeling sub-module is used to construct a personalized multidimensional ability assessment model by utilizing all history test papers in a large number of history post-examination reports, the scoring results of the corresponding history test papers obtained by candidates in each region, the multidimensional ability assessment results of candidates in each region under the corresponding scoring results calibrated by all question experts, and deep learning algorithms.

[0035] In this embodiment, the official difficulty distribution refers to a distribution generated based on the evaluation results of all question setting experts on the examination points of all history questions in each history test paper in a large number of history test reports.

[0036] Macro-difficulty distribution: This is obtained by analyzing the distribution of mastery of each history question by candidates from all regions in a large number of history test reports, the distribution of test points in each history test paper, and the answer records of each history question by candidates from each region. It represents the distribution of the average difficulty of individual test points in each test paper as perceived by candidates from all regions.

[0037] The differential difficulty distribution of all test points involved in a large number of history test papers by candidates in each region is obtained based on the actual mastery of each test point involved in each history test paper by candidates in each region and the macro-difficulty distribution of all test points involved in each history test paper, reflecting the difference in difficulty of each test point among candidates in each region relative to the overall macro-difficulty.

[0038] Question setting experts identified key areas for further examination based on the corresponding scoring results for candidates in each region: Based on the scoring results of candidates in each region's history exams, and drawing on their experience and understanding of the candidates' learning situations, question setting experts identified key areas for further examination for candidates in each region. For example, if question setting experts analyzed the test papers of candidates in a certain region and found that candidates in that region scored low in the "Application of Trigonometric Function Formulas" section, they would identify this area as a key area for further examination for candidates in that region.

[0039] Based on a large number of history exam post-exam reports, the official and macro-level difficulty distributions of all test points across all history exam papers, as well as the differential difficulty distributions of all test points across each region, are analyzed. Combined with the key points identified by test-setters in each region, as reflected in their corresponding scores, and a deep learning algorithm, a personalized model for determining key points is constructed. This process leverages multiple data points from the history exam post-exam reports to comprehensively analyze test point information from the perspectives of the official difficulty level of test points, the macro-level difficulty level reflected by the overall mastery of test-takers across all regions, and the individual and overall differences in difficulty between test-takers in each region. Combined with the key points identified by test-setters based on the scores of individual test-takers in each region, a deep learning algorithm is applied to the model to learn these complex data relationships, thereby constructing a model capable of determining personalized key points for test-takers in each region. For example, the deep learning algorithm can extract potential connections between knowledge weaknesses of regional test-takers under different difficulty distributions and key points for test-takers, enabling the model to accurately determine key points for test-takers in each region.

[0040] The score for each history exam for each region's candidates is the specific score received by all candidates in a given region in all past history exams. For example, if candidate A scores 80 on a math history exam, this score is the score for that candidate in that region for that history exam.

[0041] The multi-dimensional ability assessment results of examinees in each region, as determined by the test-setting experts based on the corresponding scoring results, are as follows: Based on the performance and scoring results of all examinees in a single region on the history exam, the test-setting experts assess the examinees in that region across multiple ability dimensions. These dimensions may include logical thinking, knowledge application, analytical problem-solving, and other abilities. For example, the test-setting experts discovered through analysis that examinees in a certain region had clear thinking when solving comprehensive application questions, but were prone to making errors in calculations. Consequently, they assessed that the examinees in that region had strong logical thinking skills, but their calculation skills needed improvement. This is the multi-dimensional ability assessment of the examinees in that region based on the test-setting experts' scoring results.

[0042] A personalized multidimensional ability assessment model was constructed using a large dataset of all history exam papers from post-exam reports, the scores for each region's candidates, the multidimensional ability assessment results of each region's candidates based on the corresponding scores as determined by all expert question setters, and a deep learning algorithm. Using these data, the model learns the inherent connections between these data, thereby constructing a model capable of performing multidimensional ability assessments for each region's candidates. The deep learning algorithm can unearth the complex mapping between a region's candidates' performance and scores on the exam and their multidimensional abilities, enabling the model to accurately assess each region's candidates' proficiency across different ability dimensions. For example, after learning from this large amount of data, the model can more accurately assess a region's candidates' actual abilities in areas such as logical thinking and knowledge application based on their scores across different question types.

[0043] In an alternative embodiment, the first modeling submodule includes: The first difficulty distribution analysis unit is used to analyze the official difficulty distribution of all test points involved in each history test paper based on the evaluation results of all test experts on the test points of all history test questions in each history test paper in a large number of history test post-test reports; A mastery distribution analysis unit is used to treat the ratio of the average score of each history question in each history test paper by all examinees in each region in a large number of history test reports to the full score of the corresponding history question as the mastery of the examinees in the corresponding region on the corresponding history question, and to generate the mastery distribution of each history question for all examinees in all regions based on the mastery of each history question by examinees in all regions; The second difficulty distribution analysis unit is used to analyze the mastery distribution of each history question and the distribution of test points in each history test paper by examinees from all regions in a large number of history test reports, as well as the answer records of each history question by examinees from each region, to obtain the macro difficulty distribution of all test points involved in each history test paper and the differential difficulty distribution of all test points involved in each history test paper by examinees from each region; The examination point determination model modeling unit is used to use deep learning algorithms and the official difficulty distribution of all examination points involved in all history examination papers, the macro difficulty distribution, the differential difficulty distribution of all examination points involved in a large number of history examination papers for candidates in each region, and all the examination points to be examined reflected in the corresponding scoring results of candidates in each region marked by all question experts based on each scoring result, to build a personalized examination point determination model.

[0044] In this embodiment, the question setting experts evaluate the examination points of all history questions in each history exam paper. This is the conclusion drawn by the question setting experts based on their evaluation of the relevant examination points for each question in each history exam paper. For example, the question setting experts will evaluate the depth of the examination of a certain examination point in the question, whether it is a simple conceptual understanding or a complex application analysis; they will also evaluate the novelty of the examination point and the importance of the examination point in the overall knowledge system. These evaluation results can help analyze the characteristics of each examination point in different exam papers and provide a basis for subsequent determination of the difficulty distribution of the examination points.

[0045] Generating a mastery distribution for each history question based on the mastery of each history question by examinees across all regions: Numerous history exam reports provide information about examinees' mastery of each history question in each region (i.e., the ratio of each region's score for each history question to the maximum possible score for that question). By collecting mastery data for a particular history question from examinees across all regions and then organizing and analyzing this data, such as by calculating the percentage of regions with different mastery ranges, a mastery distribution for that history question is generated. For example, for an English grammar question, if examinees in 20 of 100 regions have a mastery range of 0-0.2, and in 30 regions have a mastery range of 0.2-0.4, this creates a mastery distribution for that question, which visually demonstrates the overall mastery of the key points covered by examinees across different regions.

[0046] Distribution of test points in a history test paper: It describes the distribution of various test points in a history test paper, that is, the proportion of different test points in each question in this test paper and the correlation between them.

[0047] Answer records of each history question of candidates in each region: This records detailed information about all candidates in a single region in the process of answering each history question, including the order of answering, that is, which question the candidate answered first and which question later; answering time, that is, how long the candidate spent on each question; and answer content, that is, the specific answer written by the candidate.

[0048] In an alternative embodiment, the first difficulty distribution analysis unit includes: An examination difficulty determination unit is used to obtain the examination difficulty of each history test paper for each examination point involved, as assessed by each question setter, from the evaluation results of the examination points of all history test questions in each history test paper by all question setters in a large number of history test post-exam reports; The official difficulty distribution generating unit is used to take the average of the examination difficulty of each test point involved in each history test paper evaluated by all question experts as the official difficulty of each test point involved in each history test paper, and generate the official difficulty distribution of all test points involved in each history test paper based on the official difficulty of all test points involved in each history test paper.

[0049] In this embodiment, the question setting experts assess the difficulty of each test point in each history exam paper: based on their own experience, understanding of the curriculum standards and exam syllabus, and their understanding of the learning level of candidates in a specific region, the question setting experts give a difficulty assessment of each test point in each history exam paper. For example, for a history question about the "Impact of the Industrial Revolution," the question setting experts will comprehensively consider factors such as the depth of the question required of the candidate, the breadth of knowledge required, and the degree of connection with other knowledge points to assess the difficulty of the test point in this exam paper. This difficulty assessment can be expressed as a simple level (such as easy, medium, difficult) or a specific numerical value (such as 1-5 points).

[0050] Based on the official difficulty of all test points in each history exam paper, we generate an official difficulty distribution for all test points in each history exam paper. After obtaining the difficulty assessments of each test point in each history exam paper from all question setters, we calculate the average difficulty of each test point as the official difficulty of that test point. Then, based on this official difficulty data, we organize and rank the official difficulty of different test points to display the distribution of official difficulty of each test point.

[0051] In an alternative embodiment, the second difficulty distribution analysis unit includes: The test question mastery correction subunit is used to correct the mastery distribution of the corresponding history test questions by the examinees in all regions based on the average answering order and average answering time of the answering records of each history test question of all examinees in each region, and generate a corrected mastery distribution of the corresponding history test questions by the examinees in all regions; The test point mastery determination subunit is used to analyze the actual mastery of each test point involved in each history test paper by candidates in each region based on the corrected mastery distribution of each history test question and the test point distribution of each history test paper; The macro-difficulty distribution determination subunit is used to treat the difference between 1 and the mean of the actual mastery of a single test point by test takers in all regions as the macro-difficulty of the single test point, and to generate a macro-difficulty distribution of all test points involved in each history test paper based on the macro-difficulty of all test points involved in each history test paper. The differential difficulty distribution determination subunit is used to obtain the differential difficulty distribution of all test points involved in each history test paper by candidates in each region based on the actual mastery of each test point involved in each history test paper and the macro difficulty distribution of all test points involved in each history test paper.

[0052] In this embodiment, the average answering order and the average answering time: the average answering order refers to the average of the order in which all candidates in a single region answer each history question during the examination, and the answering time refers to the average of the time spent by all candidates in a single region to complete each history question.

[0053] Based on the average answer order and average answer time of each history question for all examinees in each region, the mastery distribution of examinees in all regions for the corresponding history questions was corrected to generate a corrected mastery distribution for each examinee in all regions. Since answer order and answer time can, to a certain extent, reflect examinees' true mastery of the questions, these two factors were used to adjust the original mastery distribution. For example, the first negative correction factor was the ratio of the difference between the answer order value and the question's ranking value in the corresponding exam paper to the total number of questions in the exam paper, and the second negative correction factor was the ratio of the difference between the answer time and the standard answer time to the total exam time. The value obtained by multiplying the examinee's mastery of each question in each region by (1 - first negative correction factor - second negative correction factor) was used as the corrected mastery of the examinee in that region for that question.

[0054] Based on the corrected mastery distribution of each history question and the distribution of test points on each history exam paper across all regions, we analyze the actual mastery of each test point covered by each history exam paper by each region's candidates. Combining the corrected mastery distribution and the test point distribution allows us to accurately analyze the actual mastery of each test point by each region's candidates. For example, the average of the actual mastery of all questions covering a single test point in a single region can be used as the actual mastery of that test point by the candidates in that region.

[0055] The macro-difficulty of a single test point is calculated by subtracting the average of all regions' actual mastery of a single test point from 1. This value reflects the difficulty of a particular test point relative to the general mastery of the test-takers as a whole.

[0056] Based on the macro-difficulty of all test points in each history exam, a macro-difficulty distribution of all test points in each history exam is generated: the macro-difficulty of each test point in a history exam is sorted and presented to form a macro-difficulty distribution. This allows us to intuitively see the differences in macro-difficulty of different test points in a single exam and the overall distribution trend.

[0057] In an alternative embodiment, the differential difficulty distribution determination subunit includes: The personalized difficulty determination terminal is used to use the difference between 1 and the actual mastery of each test point involved in each history test paper by the examinees in each region as the personalized difficulty of each test point involved in each history test paper by the examinees in each region; The personalized difficulty distribution determination terminal is used to aggregate the personalized difficulty of all test points involved in each history test paper for each region's candidates, and generate a personalized difficulty distribution of all test points involved in each history test paper for each region's candidates; The differential difficulty distribution determination end is used to subtract the personalized difficulty distribution of all test points involved in each history test paper for candidates in each region from the macro difficulty distribution of all test points involved in each history test paper, so as to obtain the differential difficulty distribution of all test points involved in each history test paper for candidates in each region.

[0058] In this embodiment, the personalized difficulty scores of all test points in each history exam paper for each region are aggregated to generate a personalized difficulty distribution for all test points in each history exam paper. For each test point in each history exam paper, each region has a personalized difficulty calculated based on their actual mastery (1 minus the region's actual mastery of each test point). These personalized difficulty scores for each test point are collected and organized in a specific manner (for example, by test point category, importance, or other dimensions) to form a personalized difficulty distribution for all test points in that history exam paper for each region's test takers.

[0059] By subtracting the individualized difficulty distribution of all test points in each history exam from the macro-difficulty distribution of all test points in each history exam, we obtain the differential difficulty distribution of all test points in each region. The macro-difficulty distribution reflects the overall difficulty level of all test points in a history exam, while the individualized difficulty distribution reflects the difficulty level of test points perceived by test takers in a particular region. Subtracting the two yields a differential difficulty distribution, which highlights the differences in difficulty levels among test takers in each region relative to the overall level. For example, if test takers in a certain region have an individualized difficulty level of 0.3 for the test point "Impact of the Industrial Revolution," and a macro-difficulty level of 0.2 for the same test point, the differential difficulty is 0.1, indicating that test takers in that region perceive this test point as more difficult than the overall level. By performing this calculation and summarizing all test points, we obtain the differential difficulty distribution of all test points in that history exam. This distribution allows test setters to identify test points where test takers in each region differ significantly from the overall level.

[0060] In an alternative embodiment, the model output module includes: The third difficulty distribution analysis submodule is used to determine the official difficulty distribution of all test points involved in the latest test paper, the macro difficulty distribution, and the differential difficulty distribution of all test points involved in the latest test paper for candidates in each region based on the latest post-test report; The first execution submodule is configured to input the official difficulty distribution of all test points involved in the latest test paper, the macro difficulty distribution, and the differential difficulty distribution of all test points involved in the latest test paper by candidates in each region into the personalized test point determination model to obtain the latest personalized test points to be examined for candidates in each region; The second execution submodule is used to input the scoring results of the latest test papers obtained by candidates in each region in the latest test papers and the latest post-test reports into the personalized multidimensional ability assessment model to obtain the latest multidimensional ability assessment results of candidates in each region.

[0061] In this embodiment, the official difficulty distribution, macro-difficulty distribution, and differential difficulty distribution for all test points in the latest exam are determined based on the latest post-exam report. The latest post-exam report contains a lot of detailed information about the latest exam, and this information is used to determine the difficulty distribution in different dimensions. The official difficulty distribution is determined similarly to the method used to determine the official difficulty distribution from historical post-exam reports. The assessment results of the difficulty of each test point in the latest exam are collected by the question-setting experts, and the mean difficulty of each test point is calculated. This generates the official difficulty distribution for all test points, which reflects the difficulty level set by the question-setting experts for each test point in the latest exam.

[0062] To determine the macro-difficulty distribution, we use the latest post-exam reports of candidates from all regions on each question on the latest exam paper. We then adjust the mastery of each question by individual regions, taking into account factors such as answering order and time, to create a corrected mastery distribution. We then use this corrected mastery distribution and the distribution of test points on the latest exam paper to calculate the actual mastery of each test point for each region. The average of the actual mastery of each test point across all regions is subtracted from 1 to determine the macro-difficulty of each test point. This results in a macro-difficulty distribution for all test points, reflecting the overall difficulty level of each test point on the latest exam paper for candidates.

[0063] To calculate the differential difficulty distribution of all test points in the latest exam paper for candidates in each region, we need to first determine the personalized difficulty of each test point for candidates in each region (1 minus the actual mastery of each test point by candidates in a single region), summarize it to form a personalized difficulty distribution, and then subtract it from the macro difficulty distribution. The result is the differential difficulty distribution, which shows the difference in difficulty of each test point for candidates in each region relative to the overall candidates.

[0064] Latest exam paper: This refers to the most recently administered exam paper, containing a series of questions designed to assess a candidate's knowledge. These questions cover multiple key areas within a specific subject, providing a valuable platform for candidates to demonstrate their learning and for teachers to understand their progress. For example, a recent math exam paper might contain questions covering topics from various areas, such as algebra and geometry. By answering these questions, a candidate's mastery of each key area is reflected in their answers, which in turn serve as the foundational data for the latest post-exam report and subsequent analysis.

[0065] In an alternative embodiment, the outlier labeling module includes: The dynamic test point mastery path marking submodule is used to mark the dynamic test point mastery path of candidates in the corresponding region in the standard test point mastery expansion map of the corresponding subject in the corresponding academic period based on the time series data of the mastery of all test points involved in all history examinations of a single academic period and a single subject in each region; A mastery lag factor determination submodule is used to determine the standard mastery moment for each test point based on the standard dynamic test point mastery expansion map for the corresponding subject within the corresponding academic period. Simultaneously, based on the dynamic test point mastery paths of the test takers in each region, the individualized mastery moment for each test point for the test takers in each region is determined. Based on the individualized mastery moment for each test point in each region and the standard mastery moment for the corresponding test point, the mastery lag factor for each test point in the standard dynamic test point mastery path is determined for the test takers in each region. The dynamic fitting submodule is used to dynamically fit the mastery lag factor of each test point in the standard dynamic test point mastery path of the examinees in each region, and obtain the test point mastery outlier branch path of the examinees in each region.

[0066] In this embodiment, based on the time-series data of each region's examinee's mastery of all test points across all history exams for a single subject in a single academic period, the examinee's dynamic test point mastery path for that region is marked on a standard test point mastery expansion map for that subject within that academic period. The time-series data of each region's examinee's mastery of all test points across all history exams for a single subject in that single academic period records how the examinee's mastery of each test point changes over time across all exams for that subject in that academic period. The standard test point mastery expansion map depicts information such as the relationships between test points within that academic period and the normal process of mastering those test points. By combining the examinee's test point mastery time-series data with the standard map, and marking the examinee's mastery of each test point over time on the map, the examinee's dynamic test point mastery path for that region is formed. For example, on the standard test point mastery expansion map for high school mathematics, based on the examinee's mastery of test points such as functions and geometry across all math exams starting from the first year of high school, the examinee's dynamic mastery of these test points is marked, demonstrating their mastery trajectory over time.

[0067] Dynamic Test Point Mastery Path: This is a personalized mastery path presented on the Standard Test Point Mastery Expansion Map, based on the changes in the mastery of each test point by candidates in a single region during the course of a single subject in a single academic period. It reflects dynamic information such as the order in which candidates in a single region mastered each test point, the fluctuations in their mastery of the test points, and other dynamic information when studying the subject within that academic period. For example, a candidate's dynamic test point mastery path in physics in a certain region might show that they first mastered the mechanics test points well in the first semester of their first year of high school, and then gradually improved their mastery of the electromagnetism test points in the second semester of their second year of high school, clearly demonstrating the dynamic process of the region's candidates' mastery of physics test points.

[0068] The Standard Dynamic Key Point Mastery Path, present in the Standard Key Point Mastery Expansion Map for the corresponding subject within the corresponding academic period, represents the ideal or common order and pace in which students should master each key point. It is a normative and guiding key point mastery path model, providing a reference standard for evaluating individual students' learning progress. For example, in the Standard Dynamic Key Point Mastery Path for junior high Chinese, students typically begin by mastering the foundational key points of vocabulary, then gradually move on to more advanced key points such as reading comprehension and writing, progressing in an orderly manner as the semester progresses.

[0069] Based on the standard dynamic test point mastery path in the standard test point mastery expansion map for the corresponding subject within the corresponding academic period, the standard mastery time for each test point is determined: Based on the standard dynamic test point mastery path, it can be clearly stated at what point in time, under standard circumstances, the examinee should master a specific test point. This point in time is the standard mastery time for each test point. For example, in the standard dynamic test point mastery path for the elementary school English subject, according to the syllabus and general teaching progress, it may be set that the examinee should master the spelling test point of basic words in the first semester of the third grade. In this case, the first semester of the third grade is the standard mastery time for this test point, which reflects the time standard for test point mastery under the normal teaching rhythm.

[0070] Based on the dynamic test point mastery path of the candidates in each region, the personalized mastery moment of each test point for the candidates in each region is determined: By analyzing the dynamic test point mastery path of the candidates in a single region, it is possible to determine the specific time point when the candidates in the region actually mastered each test point. This is the personalized mastery moment of each test point for the candidates in the region. It reflects the actual time that all candidates in the region spend on average to master each test point during the learning process, which may differ from the standard mastery moment. For example, when a candidate in a certain region was studying junior high school chemistry, it was discovered through their dynamic test point mastery path that they only truly mastered the test point of writing chemical equations in the second semester of the second year of junior high school. This is the personalized mastery moment of this test point for the candidates in the region, which shows the uniqueness of the individual learning rhythm of candidates in different regions.

[0071] Based on each region's individualized mastery time for each test point and the corresponding standard mastery time for the test point, a mastery lag factor for each test point in the standard dynamic test point mastery path is determined. This factor compares the individualized mastery time for each test point with the corresponding standard mastery time. A numerical value is calculated using a specific method (such as the ratio of the time difference to the standard mastery time). This value is the mastery lag factor for each test point in the standard dynamic test point mastery path for each region. It quantifies the degree to which a region's individual test point mastery lags relative to the standard progress.

[0072] Dynamic fitting is performed on the mastery lag factors of each test point in the standard dynamic test point mastery path for each region's candidates to obtain the test point mastery outlier branch path for each region's candidates. The mastery lag factors of each test point in a single region are treated as a series of data points. Dynamic fitting mathematical methods (such as curve fitting) are used to connect these discrete data points into a continuous curve or path. This path is the test point mastery outlier branch path for each region's candidates, and it can intuitively demonstrate in which test points the candidates' mastery progress in a single region deviates from the standard dynamic test point mastery path, as well as the degree and trend of the deviation. For example, dynamic fitting revealed that the path formed by the mastery lag factors of multiple test points in mathematics, such as functions and series, for candidates in a certain region deviates significantly from the standard path.

[0073] In an alternative embodiment, the test paper generation module includes: A comprehensive hysteresis factor calculation submodule is used to calculate the comprehensive hysteresis factor of the examinees in each region based on the outlier branch path of the examinees' mastery of the test points in each region; A supplementary test site determination submodule is used to screen out supplementary test sites for candidates in each region from the test site mastery outlier branch path based on the comprehensive lag factor of candidates in each region; The personalized test paper generation submodule is used to input the latest personalized test points to be examined, supplementary test points to be examined, and the latest multi-dimensional ability assessment results of candidates in each region into the pre-trained personalized proposition model to obtain personalized test papers for candidates in each region.

[0074] In this embodiment, the comprehensive hysteresis factor of the examinees in each region is calculated based on the test point mastery outlier branch path of the examinees in each region: the test point mastery outlier branch path reflects the deviation of the examinees in the region from the standard mastery path at each test point. The comprehensive hysteresis factor is a comprehensive quantitative indicator of this deviation. When calculating, the mastery hysteresis factor of each test point on the path may be comprehensively considered, such as by taking a weighted average of the mastery hysteresis factors of all test points (the weight is determined according to the importance and frequency of the test points set by the teacher, for example, the ratio of the importance of the test point to the sum of the importance of all test points in the test point mastery outlier branch path is used as the weight of a single test point), and obtaining a value that can represent the overall test point mastery hysteresis degree of the examinees in the region. This value is the comprehensive hysteresis factor.

[0075] The comprehensive lag factor for candidates in each region: This is a generalized value used to comprehensively measure the degree to which all candidates in a particular region lag behind in their progress in mastering a particular subject compared to the standard. This factor comprehensively considers the difference between the time it takes for candidates in a single region to master each test point and the standard mastery time, allowing question setting experts to quickly understand the relative lag of candidates in a single region in their mastery of the entire subject knowledge system. For example, the higher the value of the comprehensive lag factor, the more significantly the overall test point mastery progress of candidates in a single region lags behind the standard progress; the lower the value, the smaller the degree of lag. It provides an important reference indicator for subsequent personalized question setting.

[0076] Based on the comprehensive lag factor of each region's candidates, additional test points for each region are screened from the outlier path of test point mastery. The comprehensive lag factor reflects the overall lag in test point mastery among candidates in a particular region. Based on this factor, test points within the outlier path of test point mastery can be identified as requiring specific attention and further examination, i.e., supplementary test points. For example, a high comprehensive lag factor indicates that candidates in that region are lagging in mastery of a large number of test points. In this case, test points within the outlier path of test point mastery with large lag factors and significant impact on the subject knowledge system can be screened as supplementary test points. Therefore, the comprehensive lag factor can be used to search a preset list of comprehensive lag factor-mastery lag factor thresholds to determine the corresponding mastery lag factor threshold. All test points within the outlier path of test point mastery with a mastery lag factor no less than the corresponding mastery lag factor threshold can then be selected as supplementary test points. This helps test designers design more targeted questions to help candidates address knowledge gaps.

[0077] Supplementary test points to be examined: These test points are selected from the outlier branch paths of test point mastery based on the comprehensive lag factors of candidates in a single region. They are to further clarify the weak links in the current knowledge mastery of candidates in a single region. Different from regular test points to be examined, supplementary test points to be examined focus more on the parts that candidates in a single region may not understand deeply or apply well due to the lagging progress in mastery. For example, for a candidate in a region with a high comprehensive lag factor in physics, by analyzing the outlier branch paths of test point mastery, it is found that his mastery of the "electromagnetic induction" test point is significantly lagging behind. In this case, the test points related to "electromagnetic induction" can be used as supplementary test points to be examined and focused on in subsequent exams to help candidates in the region improve their mastery of this part of knowledge.

[0078] Pre-trained personalized question setting model: This model is trained on a large amount of data and aims to generate personalized exam questions based on the specific circumstances of regional candidates. During training, the model learns the relationship between the knowledge proficiency characteristics of candidates in different regions (such as the latest personalized test points to be tested, supplementary test points to be tested, and the latest multi-dimensional ability assessment results) and appropriate test questions. For example, for candidates in regions with strong proficiency, the model may generate more challenging questions; for candidates in regions where proficiency in certain test points is lagging, the model will focus on generating consolidation questions related to these test points. By continuously adjusting the model parameters, it can generate test papers that meet the knowledge level and learning needs of candidates in each region, achieving personalized exam setting.

[0079] The present invention provides an implementation of a personalized question generation method based on dynamic modeling of post-exam reports, comprising: Step S1: Construct a personalized model for determining test points and a personalized multi-dimensional ability assessment model based on a large number of historical post-test reports; Step S2: Based on the latest post-test report, the personalized test point determination model, and the personalized multi-dimensional ability assessment model, the latest personalized test points to be examined and the latest multi-dimensional ability assessment results of the examinees in each region are obtained; Step S3: Based on the time series data of the mastery of all test points involved in all history examinations of a single academic period and a single subject by examinees in each region and the mastery expansion map of the standard test points of the corresponding subject in the corresponding academic period, mark the outlier branch paths of the test point mastery of examinees in each region; Step S4: Based on the test points of the candidates in each region, the outlier branch path, the latest personalized test points to be examined, the latest multi-dimensional ability assessment results and the pre-trained personalized proposition model are mastered to obtain the personalized test papers of the candidates in each region.

[0080] The implementation method of the personalized proposition generation method based on dynamic modeling of post-test reports has many beneficial effects. First, through step S1, a personalized test point determination model and a personalized multi-dimensional ability assessment model are constructed based on a large number of historical post-test reports, providing a solid foundation for subsequent analysis and using historical data to mine the potential learning patterns of candidates. Then, step S2 obtains the latest personalized test points to be examined and multi-dimensional ability assessment results of candidates in each region based on the latest post-test report and the constructed model, accurately grasps the current learning status of candidates in each region, and provides targeted direction for assessment. Subsequently, step S3 uses the candidate's test point mastery time series data and the standard test point mastery expansion map to mark the test point mastery outlier branch path, which can clearly show the common situation of candidates in each region deviating from the norm in learning specific test points. Finally, step S4 comprehensively combines the test point mastery outlier branch path, the latest personalized test points to be examined, the latest multi-dimensional ability assessment results and the personalized proposition model to generate personalized test papers for candidates in each region, improving the quality of assessment questions and assessment results.

[0081] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is intended to include these modifications and variations.

Claims

1. A personalized proposition generation system based on dynamic modeling of post-exam reports, characterized by: include: Modeling module, used to build personalized test point determination models and personalized multi-dimensional ability assessment models based on a large number of historical test reports; The model output module is used to determine the model and the personalized multi-dimensional ability assessment model based on the latest post-test report and personalized test points to be assessed, and obtain the latest personalized test points to be assessed and the latest multi-dimensional ability assessment results for candidates in each region; The outlier marking module is used to mark the outlier branch paths of the examinees' mastery of test points in each region based on the time series data of the examinees' mastery of all test points involved in all history examinations of a single academic period and a single subject, and the mastery expansion map of the standard test points of the corresponding subject in the corresponding academic period; The test paper generation module is used to obtain personalized test papers for candidates in each region based on the test points of the candidates in each region, grasp the outlier branch paths, the latest personalized test points to be examined, the latest multi-dimensional ability assessment results and the pre-trained personalized proposition model.

2. The personalized proposition generation system based on dynamic modeling of post-exam reports according to claim 1 is characterized in that: Modeling modules, including: The first modeling submodule is used to analyze the official difficulty distribution and macro difficulty distribution of all test points involved in all history exams based on a large number of history exam reports, as well as the differential difficulty distribution of all test points involved in a large number of history exams for each region. It also combines the test points to be tested in each region as reflected in the corresponding scoring results of candidates in each region, as identified by all question setting experts, with a deep learning algorithm to construct a personalized model for determining test points to be tested. The second modeling sub-module is used to construct a personalized multidimensional ability assessment model by utilizing all history test papers in a large number of history post-examination reports, the scoring results of the corresponding history test papers obtained by candidates in each region, the multidimensional ability assessment results of candidates in each region under the corresponding scoring results calibrated by all question experts, and deep learning algorithms.

3. The personalized proposition generation system based on dynamic modeling of post-exam reports according to claim 1 is characterized in that: The first modeling submodule includes: The first difficulty distribution analysis unit is used to analyze the official difficulty distribution of all test points involved in each history test paper based on the evaluation results of all test experts on the test points of all history test questions in each history test paper in a large number of history test post-test reports; A mastery distribution analysis unit is used to treat the ratio of the average score of each history question in each history test paper by all examinees in each region in a large number of history test reports to the full score of the corresponding history question as the mastery of the examinees in the corresponding region on the corresponding history question, and to generate the mastery distribution of each history question for all examinees in all regions based on the mastery of each history question by examinees in all regions; The second difficulty distribution analysis unit is used to analyze the mastery distribution of each history question and the distribution of test points in each history test paper by examinees from all regions in a large number of history test reports, as well as the answer records of each history question by examinees from each region, to obtain the macro difficulty distribution of all test points involved in each history test paper and the differential difficulty distribution of all test points involved in each history test paper by examinees from each region; The examination point determination model modeling unit is used to use deep learning algorithms and the official difficulty distribution of all examination points involved in all history examination papers, the macro difficulty distribution, the differential difficulty distribution of all examination points involved in a large number of history examination papers for candidates in each region, and all the examination points to be examined reflected in the corresponding scoring results of candidates in each region marked by all question experts based on each scoring result, to build a personalized examination point determination model.

4. The personalized proposition generation system based on dynamic modeling of post-exam reports according to claim 3 is characterized in that: The first difficulty distribution analysis unit includes: An examination difficulty determination unit is used to obtain the examination difficulty of each history test paper for each examination point involved, as assessed by each question setter, from the evaluation results of the examination points of all history test questions in each history test paper by all question setters in a large number of history test post-exam reports; The official difficulty distribution generating unit is used to take the average of the examination difficulty of each test point involved in each history test paper evaluated by all question experts as the official difficulty of each test point involved in each history test paper, and generate the official difficulty distribution of all test points involved in each history test paper based on the official difficulty of all test points involved in each history test paper.

5. The personalized proposition generation system based on dynamic modeling of post-exam reports according to claim 3 is characterized in that: The second difficulty distribution analysis unit includes: The test question mastery correction subunit is used to correct the mastery distribution of the corresponding history test questions by the examinees in all regions based on the average answering order and average answering time of the answering records of each history test question of all examinees in each region, and generate a corrected mastery distribution of the corresponding history test questions by the examinees in all regions; The test point mastery determination subunit is used to analyze the actual mastery of each test point involved in each history test paper by candidates in each region based on the corrected mastery distribution of each history test question and the test point distribution of each history test paper; The macro-difficulty distribution determination subunit is used to treat the difference between 1 and the mean of the actual mastery of a single test point by test takers in all regions as the macro-difficulty of the single test point, and to generate a macro-difficulty distribution of all test points involved in each history test paper based on the macro-difficulty of all test points involved in each history test paper. The differential difficulty distribution determination subunit is used to obtain the differential difficulty distribution of all test points involved in each history test paper by candidates in each region based on the actual mastery of each test point involved in each history test paper and the macro difficulty distribution of all test points involved in each history test paper.

6. The personalized proposition generation system based on dynamic modeling of post-exam reports according to claim 5 is characterized in that: Differential difficulty distribution determines subunits, including: The personalized difficulty determination terminal is used to use the difference between 1 and the actual mastery of each test point involved in each history test paper by the examinees in each region as the personalized difficulty of each test point involved in each history test paper by the examinees in each region; The personalized difficulty distribution determination terminal is used to aggregate the personalized difficulty of all test points involved in each history test paper for each region's candidates, and generate a personalized difficulty distribution of all test points involved in each history test paper for each region's candidates; The differential difficulty distribution determination end is used to subtract the personalized difficulty distribution of all test points involved in each history test paper for candidates in each region from the macro difficulty distribution of all test points involved in each history test paper, so as to obtain the differential difficulty distribution of all test points involved in each history test paper for candidates in each region.

7. The personalized proposition generation system based on dynamic modeling of post-exam reports according to claim 1 is characterized in that: Model output module, including: The third difficulty distribution analysis submodule is used to determine the official difficulty distribution of all test points involved in the latest test paper, the macro difficulty distribution, and the differential difficulty distribution of all test points involved in the latest test paper for candidates in each region based on the latest post-test report; The first execution submodule is configured to input the official difficulty distribution of all test points involved in the latest test paper, the macro difficulty distribution, and the differential difficulty distribution of all test points involved in the latest test paper by candidates in each region into the personalized test point determination model to obtain the latest personalized test points to be examined for candidates in each region; The second execution submodule is used to input the scoring results of the latest test papers obtained by candidates in each region in the latest test papers and the latest post-test reports into the personalized multidimensional ability assessment model to obtain the latest multidimensional ability assessment results of candidates in each region.

8. The personalized proposition generation system based on dynamic modeling of post-exam reports according to claim 1 is characterized in that: Outlier labeling module, including: The dynamic test point mastery path marking submodule is used to mark the dynamic test point mastery path of candidates in the corresponding region in the standard test point mastery expansion map of the corresponding subject in the corresponding academic period based on the time series data of the mastery of all test points involved in all history examinations of a single academic period and a single subject in each region; A mastery lag factor determination submodule is used to determine the standard mastery moment for each test point based on the standard dynamic test point mastery expansion map for the corresponding subject within the corresponding academic period. Simultaneously, based on the dynamic test point mastery paths of the test takers in each region, the individualized mastery moment for each test point for the test takers in each region is determined. Based on the individualized mastery moment for each test point in each region and the standard mastery moment for the corresponding test point, the mastery lag factor for each test point in the standard dynamic test point mastery path is determined for the test takers in each region. The dynamic fitting submodule is used to dynamically fit the mastery lag factor of each test point in the standard dynamic test point mastery path of the examinees in each region, and obtain the test point mastery outlier branch path of the examinees in each region.

9. The personalized proposition generation system based on dynamic modeling of post-exam reports according to claim 1 is characterized in that: Test paper generation module, including: A comprehensive hysteresis factor calculation submodule is used to calculate the comprehensive hysteresis factor of the examinees in each region based on the outlier branch path of the examinees' mastery of the test points in each region; A supplementary test site determination submodule is used to screen out supplementary test sites for candidates in each region from the test site mastery outlier branch path based on the comprehensive lag factor of candidates in each region; The personalized test paper generation submodule is used to input the latest personalized test points to be examined, supplementary test points to be examined, and the latest multi-dimensional ability assessment results of candidates in each region into the pre-trained personalized proposition model to obtain personalized test papers for candidates in each region.

10. A personalized proposition generation method based on dynamic modeling of post-exam reports, characterized in that: include: Step S1: Constructing a personalized model for determining test points and a personalized multi-dimensional ability assessment model based on a large number of historical post-test reports; Step S2: Based on the latest post-test report, the personalized test point determination model, and the personalized multi-dimensional ability assessment model, the latest personalized test points to be examined and the latest multi-dimensional ability assessment results of the examinees in each region are obtained; Step S3: Based on the time series data of the mastery of all test points involved in all history examinations of a single academic period and a single subject by examinees in each region and the mastery expansion map of the standard test points of the corresponding subject in the corresponding academic period, mark the outlier branch paths of the test point mastery of examinees in each region; Step S4: Based on the test points of the candidates in each region, the outlier branch path, the latest personalized test points to be examined, the latest multi-dimensional ability assessment results and the pre-trained personalized proposition model are mastered to obtain the personalized test papers of the candidates in each region.

Images