Adaptive group test method and device based on multi-dimensional weight, equipment and storage medium

Abstract

The application provides a kind of adaptive group test method and device based on multi-dimensional weight, equipment and storage medium, by clustering division to global knowledge point, and combining knowledge point serial number and knowledge point cluster scale calculation knowledge point weight, to determine seed knowledge point set, from the source, improve the balance of knowledge point selection.Further, by establishing the coverage state pool including the pool to be covered and the covered pool, and according to the coverage state pool, the second type of question is dynamically evaluated, the real-time feedback and dynamic adjustment of knowledge point coverage state in the process of group test are realized, so that the selection of subsequent questions preferentially covers the unexamined knowledge points and inhibits the repeated knowledge points.Compared with the existing random group test mode, the application can effectively improve the knowledge point coverage rate of the test paper, reduce the knowledge point redundancy, enhance the adaptive ability of the group test process to different question bank distribution characteristics, realize more balanced and reasonable knowledge point coverage under limited number of questions, improve the group test quality and examination effect of target test paper.

Description

Technical Field

[0001] This invention relates to the field of intelligent test paper generation technology, and in particular to an adaptive test paper generation method, apparatus, device and storage medium based on multi-dimensional weights. Background Technology

[0002] With the continuous development of applications in online education, vocational skills training, corporate assessments, and qualification certifications, automated test paper generation technology has been widely applied in various intelligent examination systems. Existing automated test paper generation methods typically select questions from a question bank using random or weighted random sampling based on preset requirements for the number of question types, difficulty levels, and knowledge point distribution, to generate a target test paper that conforms to preset rules. This type of method is simple to implement, has high test paper generation efficiency, and can meet the automated test paper generation needs of general examination scenarios.

[0003] However, existing automated test paper generation methods still have significant limitations. Since the number of questions corresponding to different knowledge points in the question bank often varies considerably, when using random sampling for test paper generation, knowledge points with a large number of questions are more likely to be repeatedly sampled, while equally important knowledge points with fewer questions are easily overlooked, leading to uneven knowledge point coverage in the test paper. Furthermore, for complex questions that simultaneously relate to multiple knowledge points, existing test paper generation methods typically only filter based on question attributes, lacking a dynamic feedback mechanism for the knowledge point coverage of the generated test papers. This easily results in a large number of duplicate knowledge points, reducing the efficiency and information density of the test paper's knowledge point coverage. Simultaneously, existing test paper generation processes mostly employ static sampling strategies, failing to dynamically adjust the selection strategy for subsequent questions based on the current knowledge point coverage status of the test paper. This makes it difficult to further improve the knowledge point coverage with a limited number of questions, and to balance knowledge point coverage balance, knowledge point redundancy control, and test paper quality.

[0004] In summary, the shortcomings of the existing technology urgently need to be addressed. Summary of the Invention

[0005] This invention provides an adaptive test paper generation method, apparatus, device, and storage medium based on multi-dimensional weights to overcome the deficiencies in the prior art and achieve higher knowledge point balance and test paper generation quality in the generated target test paper.

[0006] This invention provides an adaptive test paper generation method based on multidimensional weights, comprising: Obtain the knowledge point information associated with each question in the target question bank, and divide the knowledge points of the entire domain into multiple knowledge point clusters based on the knowledge point information; The knowledge points within each knowledge point cluster are rearranged in sequence, and the weight of each knowledge point is calculated based on the rearranged sequence number and the size of the knowledge point cluster. The seed knowledge point set corresponding to the test paper assembly task is determined based on the knowledge point weight. A coverage state pool is established based on the set of seed knowledge points. The coverage state pool includes a pool to be covered and a pool that has been covered. The pool to be covered is used to store seed knowledge points that have not yet been covered by the test questions, and the pool that has been covered is used to store knowledge points that have been covered by the test questions included in the exam. Select a first target test question from the first type of questions in the target question bank, and update the knowledge points associated with the first target test question to the coverage state pool; Based on the pool to be covered and the pool already covered, the second type of questions in the target question bank are evaluated to determine the second target test paper questions; The target test paper is determined based on the first target test paper questions and the second target test paper questions.

[0007] According to the adaptive test paper generation method based on multidimensional weights provided by the present invention, the step of dividing the global knowledge points into multiple knowledge point clusters based on the knowledge point information specifically includes: Obtain the name, description, and associated tags corresponding to each knowledge point, and construct the text features of the knowledge points; Generate corresponding semantic vectors for the knowledge points based on the textual features of the knowledge points. Calculate the similarity between the semantic vectors of each knowledge point; The knowledge points are clustered based on the similarity to obtain multiple knowledge point clusters.

[0008] According to the adaptive test paper generation method based on multidimensional weights provided by the present invention, the step of reordering the knowledge points within each knowledge point cluster, calculating the knowledge point weight corresponding to each knowledge point based on the reordered sequence number and the size of the knowledge point cluster, and determining the seed knowledge point set corresponding to the test paper generation task based on the knowledge point weights specifically includes: Obtain the business importance of each knowledge point within each knowledge point cluster, and generate a corresponding random perturbation value for each knowledge point. The re-ranking score corresponding to each knowledge point is calculated based on the business importance and the random disturbance value, and the knowledge points are sorted according to the re-ranking score to determine the sequence number of each knowledge point. The weight of each knowledge point is calculated based on its corresponding serial number and the size of the knowledge point cluster to which it belongs. Generate corresponding random sorting key values ​​based on the knowledge point weights of each knowledge point, and determine the seed knowledge point set corresponding to the test paper assembly task based on each random sorting key value.

[0009] According to the adaptive test paper generation method based on multi-dimensional weights provided by the present invention, the step of selecting a first target test paper question from the first type of questions in the target question bank and updating the knowledge points associated with the first target test paper question to the coverage state pool specifically includes: Select the first target set of questions from the first type of questions in the target question bank; Obtain the set of knowledge points associated with the questions in the first target test paper; Iterate through each knowledge point in the set of knowledge points; When the corresponding knowledge point exists in the pool to be covered, remove the knowledge point from the pool to be covered and add it to the covered pool; When the corresponding knowledge point does not exist in the pool to be covered, the knowledge point is added to the pool that has already been covered.

[0010] According to the adaptive test paper generation method based on multi-dimensional weights provided by the present invention, the step of evaluating the second type of questions in the target question bank based on the pool to be covered and the pool already covered, and determining the second target test paper generation questions, specifically includes: Obtain the set of knowledge points associated with each second type of question in the target question bank; Based on the correspondence between each set of knowledge points and the pool to be covered, determine the incremental revenue value corresponding to each second type of question; Based on the correspondence between each set of knowledge points and the covered pool, determine the redundancy suppression value corresponding to each type of question; The dynamic value score for each type of question is determined based on the incremental return value, redundancy suppression value, and dynamic adjustment parameters corresponding to each type of question. Based on the dynamic value scores corresponding to each type of question in the second category, the questions for the second target test paper are determined.

[0011] According to the adaptive test case assembly method based on multidimensional weights provided by the present invention, the incremental gain value and the redundancy suppression value are determined by the following method: Determine the number of knowledge points that intersect with the knowledge point set corresponding to each second type of question and the pool to be covered, and determine the number of intersecting knowledge points as the incremental revenue value for the corresponding second type of question; Determine the number of knowledge points that intersect with the knowledge point set corresponding to each type of question and the covered pool, and use the number of intersecting knowledge points as the redundancy suppression value for the corresponding type of question.

[0012] According to the adaptive test paper generation method based on multi-dimensional weights provided by the present invention, the step of determining the dynamic value score corresponding to each second type of question based on the incremental benefit value, redundancy suppression value, and dynamic adjustment parameter corresponding to each second type of question specifically includes: Get the number of remaining knowledge points in the knowledge point pool to be covered and the total number of knowledge points in the seed knowledge point set; The incentive coefficient and the redundancy penalty coefficient are determined based on the ratio between the number of remaining knowledge points in the pool to be covered and the total number of knowledge points in the seed knowledge point set. The dynamic value score corresponding to each type of question is determined based on the incentive coefficient, the redundancy penalty coefficient, the incremental benefit value, and the redundancy suppression value.

[0013] The present invention also provides an adaptive paper generation device based on multidimensional weights, comprising: The knowledge segmentation module is used to obtain the knowledge point information associated with each question in the target question bank, and to divide the knowledge points in the entire domain into multiple knowledge point clusters based on the knowledge point information; The sequence rearrangement module is used to rearrange the knowledge points within each knowledge point cluster, calculate the weight of each knowledge point based on the rearranged sequence number and the size of the knowledge point cluster, and determine the seed knowledge point set corresponding to the test paper assembly task based on the knowledge point weight. The state pool module is used to establish a coverage state pool based on the seed knowledge point set. The coverage state pool includes a pending coverage pool and a covered pool. The pending coverage pool is used to store seed knowledge points that have not yet been covered by the test questions, and the covered pool is used to store knowledge points that have been covered by the test questions included in the exam. The first test paper module is used to select the first target test paper questions from the first type of questions in the target question bank, and update the knowledge points associated with the first target test paper questions to the coverage state pool. The second test paper module is used to evaluate the second type of questions in the target question bank based on the pool to be covered and the pool already covered, and to determine the second target test paper questions; The test paper determination module is used to determine the target test paper based on the first target test paper questions and the second target test paper questions.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the adaptive volume composition method based on multidimensional weights as described above.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the adaptive volume-building method based on multidimensional weights as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the adaptive volume composition method based on multidimensional weights as described above.

[0017] The present invention provides an adaptive test paper generation method, apparatus, device, and storage medium based on multi-dimensional weights. First, it clusters all knowledge points in the question bank and calculates the weight of each knowledge point based on its index and cluster size. This weight is then used to determine the seed knowledge point set, reducing sampling bias caused by different sizes of knowledge point clusters in the question bank and improving the distribution balance of seed knowledge points. Furthermore, a coverage status pool, including a pool to be covered and a pool already covered, is established based on the seed knowledge point set. This pool is updated in real-time during the first type of question generation process. The second type of questions are dynamically evaluated based on the pool to be covered and the pool already covered, prioritizing questions that can add uncovered knowledge points and reduce duplicate knowledge points for test paper generation. This allows subsequent question selection to be adjusted in real-time according to the knowledge point coverage status, forming a dynamic feedback closed-loop test paper generation mechanism. Therefore, under the same question quantity conditions, the present invention can effectively improve the knowledge point coverage of the test paper, reduce the knowledge point repetition rate, increase the information density and assessment coverage of the test paper, and enhance the adaptability of the test paper generation method to different question bank distribution characteristics, so that the generated target test paper has higher knowledge point balance and test paper generation quality. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the adaptive test paper generation method based on multidimensional weights provided by the present invention. Figure 2 This is a schematic diagram of the structure of the adaptive paper-generating device based on multidimensional weights provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] To address the problems in existing technologies, this invention proposes an adaptive test paper generation method based on multidimensional weights to achieve higher knowledge point balance and overall test paper quality in the generated target test papers. The following describes this adaptive test paper generation method based on multidimensional weights, as follows: Figure 1 As shown, including but not limited to the following steps: Step 110: Obtain the knowledge point information associated with each question in the target question bank, and divide the knowledge points of the entire domain into multiple knowledge point clusters based on the knowledge point information.

[0022] In step 110, the system first obtains the knowledge point information corresponding to each question in the target question bank. This knowledge point information represents the knowledge content tested in each question; a question can be associated with one or more knowledge points. Subsequently, the system classifies all knowledge points based on the relationships between them, grouping highly correlated knowledge points into the same knowledge point cluster, resulting in multiple knowledge point clusters. By clustering all knowledge points, the system can manage knowledge points with similar attributes or relationships in a unified manner, providing foundational data for subsequent knowledge point weight calculations and seed knowledge point selection.

[0023] Step 120: Rearrange the knowledge points within each knowledge point cluster, calculate the weight of each knowledge point based on the rearranged sequence number and the size of the knowledge point cluster, and determine the seed knowledge point set corresponding to the test paper assembly task based on the knowledge point weight.

[0024] In step 120, the system reorders the knowledge points within each knowledge point cluster to obtain the corresponding sequence number of each knowledge point within its cluster, and calculates the weight of each knowledge point based on the size of its cluster. Subsequently, the system filters all knowledge points based on their weights to determine the set of seed knowledge points that meet the requirements of the current test paper assembly task. By introducing knowledge point weights, the impact of differences in knowledge point cluster size on the selection results can be reduced, improving the distribution balance of seed knowledge points and providing a more reasonable foundation for knowledge point coverage in subsequent test paper assembly.

[0025] Step 130: Establish a coverage state pool based on the seed knowledge point set. The coverage state pool includes a pending coverage pool and a covered pool. The pending coverage pool is used to store seed knowledge points that have not yet been covered by the test questions, and the covered pool is used to store knowledge points that have been covered by the test questions included in the exam.

[0026] In step 130, the system establishes a coverage status pool based on the seed knowledge point set determined in step 120, and divides the coverage status pool into a pending coverage pool and a covered pool. The pending coverage pool records seed knowledge points that have not yet been covered by questions during the current test paper assembly process, while the covered pool records knowledge points that have already been covered by the included test questions. At the start of test paper assembly, all knowledge points in the seed knowledge point set are initialized and stored in the pending coverage pool, and the covered pool is initialized to empty. The coverage status pool records the coverage status of knowledge points throughout the entire test paper assembly process, providing a status basis for subsequent question evaluation and test paper assembly decisions.

[0027] Step 140: Select the first target test question from the first type of questions in the target question bank, and update the knowledge points associated with the first target test question to the coverage state pool.

[0028] In step 140, the system selects a first target test paper question that meets the test paper assembly requirements from the first type of questions in the target question bank and adds the first target test paper question to the target test paper. Subsequently, the system obtains the knowledge points associated with the first target test paper question and updates the coverage status pool based on these knowledge points, updating the knowledge points already covered by the first target test paper question from the pending coverage state to the covered state, thus reflecting the current knowledge point coverage status of the target test paper in real time. By updating the coverage status pool in real time, the system can provide the latest knowledge point coverage information for subsequent question evaluation.

[0029] Step 150: Based on the pool to be covered and the pool already covered, evaluate the second type of questions in the target question bank to determine the second target test paper questions.

[0030] In step 150, the system evaluates the second type of questions in the target question bank based on the knowledge point coverage status recorded in the current to-be-covered pool and the already-covered pool. It analyzes the ability of each second type of question to supplement uncovered knowledge points and the degree of repetition of covered knowledge points, and determines the second target test paper questions based on the evaluation results. Subsequently, the determined second target test paper questions are added to the target test paper, and the coverage status pool is updated according to the knowledge points associated with the second target test paper questions, ensuring that the coverage status pool continuously reflects the latest knowledge point coverage status, thereby achieving dynamic adjustment during the test paper assembly process.

[0031] Step 160: Determine the target test paper based on the first target test paper questions and the second target test paper questions.

[0032] In step 160, the system combines the first target test paper questions determined in step 140 and the second target test paper questions determined in step 150 to form a target test paper that meets the requirements of the current test paper assembly task. When the target test paper meets the preset test paper assembly conditions, the test paper assembly process ends and the target test paper is output. By using the first type of questions to establish basic knowledge point coverage, and then combining it with the dynamic supplementation method of the second type of questions, the balance of knowledge point coverage of the target test paper can be improved, the knowledge point repetition rate can be reduced, and thus the quality of test paper assembly can be improved.

[0033] As a further optional embodiment, the step of dividing the entire domain knowledge points into multiple knowledge point clusters based on the knowledge point information specifically includes: Obtain the name, description, and associated tags corresponding to each knowledge point, and construct the text features of the knowledge points; Generate corresponding semantic vectors for the knowledge points based on the textual features of the knowledge points. Calculate the similarity between the semantic vectors of each knowledge point; The knowledge points are clustered based on the similarity to obtain multiple knowledge point clusters.

[0034] In this embodiment, the system first obtains the name, description information, and associated tags corresponding to each knowledge point, and then concatenates or combines the name, description information, and associated tags to construct the corresponding knowledge point text features. The name represents the topic content of the knowledge point, the description information represents the detailed semantic information of the knowledge point, and the associated tags represent auxiliary information such as the course, chapter, ability requirement, or classification attribute to which the knowledge point belongs. By integrating multiple types of information, the semantic features of each knowledge point can be more completely reflected.

[0035] Subsequently, the system generates corresponding semantic vectors for the knowledge points based on their textual features. Specifically, pre-trained language models, text embedding models, or other semantic encoding models can be used to encode the textual features of the knowledge points, converting the textual information into fixed-dimensional vector representations. This ensures that semantically similar knowledge points are close in the vector space, thus providing a unified data representation for subsequent similarity calculations between knowledge points.

[0036] After obtaining the semantic vectors of each knowledge point, the system calculates the similarity between any two knowledge point semantic vectors to characterize the degree of semantic association between the two knowledge points. The similarity can be calculated using cosine similarity, Euclidean distance transformation, Pearson correlation coefficient, or other methods that can characterize the degree of vector similarity; this embodiment does not limit this method.

[0037] Subsequently, the system clusters all knowledge points based on the similarity between them, grouping knowledge points with high semantic relevance into the same knowledge point cluster and those with low semantic relevance into different knowledge point clusters, ultimately resulting in multiple knowledge point clusters. The clustering algorithm can be hierarchical clustering, K-Means clustering, density clustering, or other clustering algorithms, and this embodiment does not limit this to any particular algorithm.

[0038] The above method can automatically construct knowledge point clusters by utilizing the semantic relationships between knowledge points, avoiding the problem of insufficient semantic relationships caused by relying solely on manual classification or chapter structure to divide knowledge points. This allows knowledge points with similar semantic features to belong to the same knowledge point cluster, thereby improving the accuracy of subsequent knowledge point weight calculation and knowledge point coverage analysis during the test paper assembly process, and providing a reliable data foundation for achieving balanced knowledge point coverage.

[0039] As a further optional embodiment, the step of rearranging the sequence of knowledge points within each knowledge point cluster, calculating the weight of each knowledge point based on the rearranged sequence number and the size of the knowledge point cluster, and determining the seed knowledge point set corresponding to the test paper assembly task based on the knowledge point weights specifically includes: Obtain the business importance of each knowledge point within each knowledge point cluster, and generate a corresponding random perturbation value for each knowledge point. The re-ranking score corresponding to each knowledge point is calculated based on the business importance and the random disturbance value, and the knowledge points are sorted according to the re-ranking score to determine the sequence number of each knowledge point. The weight of each knowledge point is calculated based on its corresponding serial number and the size of the knowledge point cluster to which it belongs. Generate corresponding random sorting key values ​​based on the knowledge point weights of each knowledge point, and determine the seed knowledge point set corresponding to the test paper assembly task based on each random sorting key value.

[0040] In this embodiment, the system first obtains the business importance of each knowledge point within each knowledge point cluster. The business importance is used to characterize the assessment value or priority of the knowledge point in the current test paper generation task. Simultaneously, to reduce the impact of factors such as the original arrangement order and fixed sorting rules of knowledge points on the knowledge point selection results, the system generates a corresponding random perturbation value for each knowledge point, thereby adding a certain degree of randomness to the knowledge points while maintaining the overall trend of business importance.

[0041] Subsequently, the system calculates the corresponding rearrangement score based on the business importance of each knowledge point and the random perturbation value, and sorts the knowledge points within the same knowledge point cluster according to the rearrangement score to determine the sequence number of each knowledge point in its respective knowledge point cluster. By rearranging the sequence of knowledge points, knowledge points in different knowledge point clusters can participate in subsequent weight calculations based on unified rules, avoiding the situation where some knowledge points occupy a high selection priority for a long time due to the order of knowledge point entry or fixed arrangement.

[0042] After determining the sequence number of each knowledge point, the system further calculates the knowledge point weight by combining the size of the knowledge point cluster to which each knowledge point belongs with its corresponding sequence number. The size of the knowledge point cluster reflects the number of knowledge points contained within it, while the knowledge point sequence number reflects the ranking position of the knowledge point within its cluster. By simultaneously considering both the knowledge point cluster size and the knowledge point sequence number, the knowledge point weight can reflect the importance of the knowledge point itself and the distribution of its knowledge point cluster, thereby reducing the impact of large-scale knowledge point clusters on the seed knowledge point selection process.

[0043] Furthermore, the system generates corresponding random sorting key values ​​based on the weights of each knowledge point, and sorts them according to these random sorting key values. From the sorting results, it determines a set of seed knowledge points that meet the requirements of the test paper assembly task. The set of seed knowledge points represents the target knowledge points that need to be covered in the current test paper assembly task, providing a basis for the subsequent establishment of a coverage state pool and the test question selection process.

[0044] Through the above methods, this embodiment can adjust the knowledge point selection process in multiple dimensions based on the size of the knowledge point cluster, the ranking position of the knowledge points, and the weight of the knowledge points. This ensures that knowledge points in knowledge point clusters of different sizes have a reasonable probability of participation, avoids the concentration of test question resources in a few knowledge point clusters, improves the coverage balance of the seed knowledge point set, and provides a foundation for generating a target test paper with a more reasonable distribution of knowledge points.

[0045] As a further optional embodiment, the step of selecting a first target test paper question from the first type of questions in the target question bank and updating the knowledge points associated with the first target test paper question to the coverage state pool specifically includes: Select the first target set of questions from the first type of questions in the target question bank; Obtain the set of knowledge points associated with the questions in the first target test paper; Iterate through each knowledge point in the set of knowledge points; When the corresponding knowledge point exists in the pool to be covered, remove the knowledge point from the pool to be covered and add it to the covered pool; When the corresponding knowledge point does not exist in the pool to be covered, the knowledge point is added to the pool that has already been covered.

[0046] In this embodiment, the system first selects the first target test paper questions from the first category of questions in the target question bank. These first category of questions are used to prioritize the construction of the basic knowledge point coverage structure of the target test paper. Based on the question type requirements, number of questions, and knowledge point coverage requirements corresponding to the current test paper assembly task, the system determines the first target test paper questions to be added to the target test paper from the first category of questions.

[0047] After determining the first target test question, the system obtains the set of knowledge points corresponding to the first target test question. The set of knowledge points represents all the knowledge points tested in the question. Since different questions may correspond to one or more knowledge points, the system iterates through each knowledge point in the set of knowledge points to determine the state of each knowledge point in the current coverage state pool.

[0048] Specifically, when a knowledge point is found in the pool to be covered, it indicates that the knowledge point belongs to the target knowledge point that the current test paper assembly task requires to be covered but has not yet been covered by the test questions. The system removes the knowledge point from the pool to be covered and adds it to the pool that has been covered, indicating that the knowledge point has been covered and updated by the first target test paper questions.

[0049] When a knowledge point encountered during the traversal does not exist in the pool to be covered, it indicates that the knowledge point is already covered, or that the knowledge point does not belong to the seed knowledge point set set by the current test paper creation task. At this time, the system adds the knowledge point to the covered pool to record all the knowledge point information that the target test paper has already covered.

[0050] Through the above method, the system updates the coverage status pool in real time while selecting the first target test paper questions. This allows the coverage status pool to dynamically reflect the knowledge point coverage of the current target test paper, providing real-time status information for the evaluation and selection of subsequent second-type questions. Compared to traditional test paper generation methods that rely solely on fixed rules for question selection, this embodiment introduces a knowledge point coverage status feedback mechanism, enabling subsequent question selection to be dynamically adjusted based on existing test paper generation results, thus improving the knowledge point coverage effect with a limited number of questions.

[0051] As a further optional embodiment, the step of evaluating the second type of questions in the target question bank based on the pool to be covered and the pool already covered, and determining the second target test paper questions, specifically includes: Obtain the set of knowledge points associated with each second type of question in the target question bank; Based on the correspondence between each set of knowledge points and the pool to be covered, determine the incremental revenue value corresponding to each second type of question; Based on the correspondence between each set of knowledge points and the covered pool, determine the redundancy suppression value corresponding to each type of question; The dynamic value score for each type of question is determined based on the incremental return value, redundancy suppression value, and dynamic adjustment parameters corresponding to each type of question. Based on the dynamic value scores corresponding to each type of question in the second category, the questions for the second target test paper are determined.

[0052] In this embodiment, after selecting the first target test questions and updating the coverage state pool, the system further dynamically evaluates the second type of questions in the target question bank to select second target test questions from the remaining questions that can improve the coverage of knowledge points. The second type of questions typically relate to multiple knowledge points. Through comprehensive evaluation of the second type of questions, it is possible to further supplement uncovered knowledge points within a limited number of questions and reduce the introduction of duplicate knowledge points.

[0053] Specifically, the system first obtains the set of knowledge points corresponding to each type of question in the target question bank to determine the scope of examination for each type of question. Then, the system performs correlation analysis between the set of knowledge points corresponding to each type of question and the pool to be covered and the pool already covered in the current coverage status pool.

[0054] The system determines the incremental benefit value of each second-type question based on the correspondence between the knowledge point set of the second-type questions and the knowledge pool to be covered. The incremental benefit value characterizes the ability of the second-type question to supplement currently uncovered knowledge points. The higher the incremental benefit value, the more knowledge points the second-type question can cover, and the greater its value in improving the knowledge point coverage of the target test paper.

[0055] Simultaneously, based on the correspondence between the knowledge point set of the second type of questions and the covered pool, the system determines the redundancy suppression value for the corresponding second type of questions. This redundancy suppression value characterizes the degree to which the second type of question incorporates covered knowledge points. By analyzing the redundancy suppression value, the waste of test resources caused by repeatedly testing existing knowledge points can be reduced.

[0056] Furthermore, the system comprehensively evaluates each type of question based on its incremental benefit value, redundancy suppression value, and dynamic adjustment parameters, obtaining a corresponding dynamic value score. The dynamic adjustment parameters are used to adjust the weighting relationship between incremental benefit and redundancy control according to the knowledge point coverage status during the current test paper assembly process, enabling the selection process of type II questions to dynamically change as the test paper assembly progresses.

[0057] Finally, based on the dynamic value scores corresponding to each type of question, the system identifies the second target test paper questions with higher dynamic value scores from multiple types of questions and adds them to the target test paper. Through this method, the system can utilize real-time feedback information from the coverage state pool to dynamically filter questions covering multiple knowledge points. This allows the selected second-type questions to prioritize supplementing uncovered knowledge points while reducing knowledge point duplication, thereby improving the balance of knowledge point coverage and the overall quality of the test paper.

[0058] As a further optional embodiment, the incremental gain value and the redundancy suppression value are determined by the following method: Determine the number of knowledge points that intersect with the knowledge point set corresponding to each second type of question and the pool to be covered, and determine the number of intersecting knowledge points as the incremental revenue value for the corresponding second type of question; Determine the number of knowledge points that intersect with the knowledge point set corresponding to each type of question and the covered pool, and use the number of intersecting knowledge points as the redundancy suppression value for the corresponding type of question.

[0059] In this embodiment, when the system performs dynamic value evaluation on the second type of questions, it analyzes the relationship between the knowledge points associated with the second type of questions and the current coverage state pool, and determines the incremental benefit value and redundancy suppression value of each second type of question, so as to quantify the contribution of the second type of questions to the optimization of knowledge point coverage.

[0060] Specifically, the system obtains the knowledge point set corresponding to each type of question in the second category, and matches the knowledge point set with the knowledge points in the coverage pool to determine the number of overlapping knowledge points. The overlapping knowledge points represent target knowledge points in the current test paper generation task that are not yet covered by existing questions but can be supplemented and covered by these second-category questions. The system determines the number of overlapping knowledge points as the incremental benefit value for the corresponding second-category question. The larger the incremental benefit value, the more uncovered knowledge points the second-category question can supplement, and the higher its contribution to improving the knowledge point coverage of the test paper.

[0061] Furthermore, the system matches the set of knowledge points corresponding to each type of question with the knowledge points in the already covered pool to determine the number of overlapping knowledge points. These overlapping knowledge points represent the knowledge points included in the type of question that have already been covered by the current target exam paper. The system determines the number of overlapping knowledge points as the redundancy suppression value for the corresponding type of question. The larger the redundancy suppression value, the more repeatedly tested knowledge points the type of question contains, and the more significant its impact on the exam paper's knowledge point coverage efficiency.

[0062] By calculating the intersection between the second type of questions and the pool to be covered and the pool already covered, this embodiment can quantitatively characterize the new coverage capability and the degree of knowledge point repetition of the second type of questions. This provides a data basis for determining the dynamic value score based on the incremental benefit value and the redundancy suppression value, making the selection of the second type of questions more in line with the test paper assembly goal of maximizing knowledge point coverage and minimizing redundancy.

[0063] As a further optional embodiment, the step of determining the dynamic value score corresponding to each type of question based on the incremental revenue value, redundancy suppression value, and dynamic adjustment parameters corresponding to each type of question specifically includes: Get the number of remaining knowledge points in the knowledge point pool to be covered and the total number of knowledge points in the seed knowledge point set; The incentive coefficient and the redundancy penalty coefficient are determined based on the ratio between the number of remaining knowledge points in the pool to be covered and the total number of knowledge points in the seed knowledge point set. The dynamic value score corresponding to each type of question is determined based on the incentive coefficient, the redundancy penalty coefficient, the incremental benefit value, and the redundancy suppression value.

[0064] In this embodiment, the system dynamically adjusts the evaluation strategy for the second type of questions based on the knowledge point coverage status during the current test paper assembly process, so that the selection target of the second type of questions at different test paper assembly stages can adapt to the current knowledge point coverage requirements.

[0065] Specifically, the system obtains the number of remaining knowledge points in the current knowledge point pool to be covered and the total number of knowledge points in the initial seed knowledge point set. The number of remaining knowledge points in the knowledge point pool reflects the number of knowledge points that the current target test paper has not yet covered, and the total number of knowledge points in the seed knowledge point set represents the scale of knowledge points that the current test paper assembly task needs to focus on covering. By obtaining these parameters, the system can determine the degree of knowledge point coverage corresponding to the current test paper assembly stage.

[0066] Subsequently, the system determines the incentive coefficient and redundancy penalty coefficient based on the ratio between the number of remaining knowledge points in the coverage pool and the total number of knowledge points in the seed knowledge point set. Specifically, in the initial stage of test paper generation, because the number of remaining knowledge points in the coverage pool is relatively large, the target test paper needs to prioritize expanding the coverage of knowledge points. Therefore, the system increases the incentive coefficient, giving higher selection priority to second-type questions that can supplement more uncovered knowledge points. As the test paper generation process progresses, the number of remaining knowledge points in the coverage pool gradually decreases, and the system gradually increases the redundancy penalty coefficient, making the selection of second-type questions more focused on the repetition of knowledge points, thereby reducing the introduction of redundant questions.

[0067] Furthermore, the system comprehensively calculates the corresponding dynamic value score for each type of question based on the incentive coefficient, the redundancy penalty coefficient, the incremental benefit value of each type of question, and the redundancy suppression value. The dynamic value score comprehensively reflects the contribution of the type of question to the current test paper creation task. A higher score indicates that the type of question can provide higher new coverage value under the current knowledge point coverage state, while having a lower degree of knowledge point repetition.

[0068] Through the above method, this embodiment can dynamically adjust the evaluation criteria of the second type of questions according to the progress of knowledge point coverage in the test paper assembly process. This allows the early stage of test paper assembly to focus on rapidly expanding the scope of knowledge point coverage, while the later stage focuses on reducing knowledge point repetition, thereby improving the rationality of the overall knowledge point distribution in the test paper and achieving adaptive optimization in the test paper assembly process.

[0069] The adaptive paper generation device based on multidimensional weights provided by this invention is described below, such as... Figure 2 As shown, the adaptive paper-building device based on multidimensional weights described below and the adaptive paper-building method based on multidimensional weights described above can be referred to in correspondence.

[0070] An adaptive paper generation device based on multidimensional weights, comprising: The knowledge segmentation module 210 is used to obtain the knowledge point information associated with each question in the target question bank, and to divide the knowledge points of the entire domain into multiple knowledge point clusters based on the knowledge point information; The sequence rearrangement module 220 is used to rearrange the knowledge points within each knowledge point cluster, calculate the weight of each knowledge point based on the rearranged sequence number and the size of the knowledge point cluster, and determine the seed knowledge point set corresponding to the test paper assembly task based on the knowledge point weight. The state pool module 230 is used to establish a coverage state pool based on the seed knowledge point set. The coverage state pool includes a pending coverage pool and a covered pool. The pending coverage pool is used to store seed knowledge points that have not yet been covered by the test questions, and the covered pool is used to store knowledge points that have been covered by the test questions included in the exam. The first test paper module 240 is used to select a first target test paper question from the first type of questions in the target question bank, and update the knowledge points associated with the first target test paper question to the coverage state pool. The second test paper module 250 is used to evaluate the second type of questions in the target question bank based on the pool to be covered and the pool already covered, and to determine the second target test paper questions; The test paper determination module 260 is used to determine the target test paper based on the first target test paper questions and the second target test paper questions.

[0071] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute an adaptive volume-building method based on multi-dimensional weights, the method including: Obtain the knowledge point information associated with each question in the target question bank, and divide the knowledge points of the entire domain into multiple knowledge point clusters based on the knowledge point information; The knowledge points within each knowledge point cluster are rearranged in sequence, and the weight of each knowledge point is calculated based on the rearranged sequence number and the size of the knowledge point cluster. The seed knowledge point set corresponding to the test paper assembly task is determined based on the knowledge point weight. A coverage state pool is established based on the set of seed knowledge points. The coverage state pool includes a pool to be covered and a pool that has been covered. The pool to be covered is used to store seed knowledge points that have not yet been covered by the test questions, and the pool that has been covered is used to store knowledge points that have been covered by the test questions included in the exam. Select a first target test question from the first type of questions in the target question bank, and update the knowledge points associated with the first target test question to the coverage state pool; Based on the pool to be covered and the pool already covered, the second type of questions in the target question bank are evaluated to determine the second target test paper questions; The target test paper is determined based on the first target test paper questions and the second target test paper questions.

[0072] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0073] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the adaptive volume-building method based on multi-dimensional weights provided by the above methods, the method comprising: Obtain the knowledge point information associated with each question in the target question bank, and divide the knowledge points of the entire domain into multiple knowledge point clusters based on the knowledge point information; The knowledge points within each knowledge point cluster are rearranged in sequence, and the weight of each knowledge point is calculated based on the rearranged sequence number and the size of the knowledge point cluster. The seed knowledge point set corresponding to the test paper assembly task is determined based on the knowledge point weight. A coverage state pool is established based on the set of seed knowledge points. The coverage state pool includes a pool to be covered and a pool that has been covered. The pool to be covered is used to store seed knowledge points that have not yet been covered by the test questions, and the pool that has been covered is used to store knowledge points that have been covered by the test questions included in the exam. Select a first target test question from the first type of questions in the target question bank, and update the knowledge points associated with the first target test question to the coverage state pool; Based on the pool to be covered and the pool already covered, the second type of questions in the target question bank are evaluated to determine the second target test paper questions; The target test paper is determined based on the first target test paper questions and the second target test paper questions.

[0074] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the adaptive volume-building method based on multidimensional weights provided by the methods described above, the method comprising: Obtain the knowledge point information associated with each question in the target question bank, and divide the knowledge points of the entire domain into multiple knowledge point clusters based on the knowledge point information; The knowledge points within each knowledge point cluster are rearranged in sequence, and the weight of each knowledge point is calculated based on the rearranged sequence number and the size of the knowledge point cluster. The seed knowledge point set corresponding to the test paper assembly task is determined based on the knowledge point weight. A coverage state pool is established based on the set of seed knowledge points. The coverage state pool includes a pool to be covered and a pool that has been covered. The pool to be covered is used to store seed knowledge points that have not yet been covered by the test questions, and the pool that has been covered is used to store knowledge points that have been covered by the test questions included in the exam. Select a first target test question from the first type of questions in the target question bank, and update the knowledge points associated with the first target test question to the coverage state pool; Based on the pool to be covered and the pool already covered, the second type of questions in the target question bank are evaluated to determine the second target test paper questions; The target test paper is determined based on the first target test paper questions and the second target test paper questions.

[0075] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0076] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive test paper generation method based on multidimensional weights, characterized in that, include: Obtain the knowledge point information associated with each question in the target question bank, and divide the knowledge points of the entire domain into multiple knowledge point clusters based on the knowledge point information; The knowledge points within each knowledge point cluster are rearranged in sequence, and the weight of each knowledge point is calculated based on the rearranged sequence number and the size of the knowledge point cluster. The seed knowledge point set corresponding to the test paper assembly task is determined based on the knowledge point weight. A coverage state pool is established based on the set of seed knowledge points. The coverage state pool includes a pool to be covered and a pool that has been covered. The pool to be covered is used to store seed knowledge points that have not yet been covered by the test questions, and the pool that has been covered is used to store knowledge points that have been covered by the test questions included in the exam. Select a first target test question from the first type of questions in the target question bank, and update the knowledge points associated with the first target test question to the coverage state pool; Based on the pool to be covered and the pool already covered, the second type of questions in the target question bank are evaluated to determine the second target test paper questions; The target test paper is determined based on the first target test paper questions and the second target test paper questions.

2. The adaptive test paper generation method based on multidimensional weights according to claim 1, characterized in that, The step of dividing the entire domain of knowledge points into multiple knowledge point clusters based on the knowledge point information specifically includes: Obtain the name, description, and associated tags corresponding to each knowledge point, and construct the text features of the knowledge points; Generate corresponding semantic vectors for the knowledge points based on the textual features of the knowledge points. Calculate the similarity between the semantic vectors of each knowledge point; The knowledge points are clustered based on the similarity to obtain multiple knowledge point clusters.

3. The adaptive test paper generation method based on multidimensional weights according to claim 1, characterized in that, The step of rearranging the sequence of knowledge points within each knowledge point cluster, calculating the weight of each knowledge point based on the rearranged sequence number and the size of the knowledge point cluster, and determining the seed knowledge point set corresponding to the test paper assembly task based on the knowledge point weights specifically includes: Obtain the business importance of each knowledge point within each knowledge point cluster, and generate a corresponding random perturbation value for each knowledge point. The re-ranking score corresponding to each knowledge point is calculated based on the business importance and the random disturbance value, and the knowledge points are sorted according to the re-ranking score to determine the sequence number of each knowledge point. The weight of each knowledge point is calculated based on its corresponding serial number and the size of the knowledge point cluster to which it belongs. Generate corresponding random sorting key values ​​based on the knowledge point weights of each knowledge point, and determine the seed knowledge point set corresponding to the test paper assembly task based on each random sorting key value.

4. The adaptive test paper generation method based on multidimensional weights according to claim 1, characterized in that, The step of selecting a first target test paper question from the first type of questions in the target question bank and updating the knowledge points associated with the first target test paper question to the coverage state pool specifically includes: Select the first target set of questions from the first type of questions in the target question bank; Obtain the set of knowledge points associated with the questions in the first target test paper; Iterate through each knowledge point in the set of knowledge points; When the corresponding knowledge point exists in the pool to be covered, remove the knowledge point from the pool to be covered and add it to the covered pool; When the corresponding knowledge point does not exist in the pool to be covered, the knowledge point is added to the pool that has already been covered.

5. The adaptive test paper generation method based on multidimensional weights according to claim 1, characterized in that, The step of evaluating the second type of questions in the target question bank based on the pool to be covered and the pool already covered, and determining the second target test paper questions, specifically includes: Obtain the set of knowledge points associated with each second type of question in the target question bank; Based on the correspondence between each set of knowledge points and the pool to be covered, determine the incremental revenue value corresponding to each second type of question; Based on the correspondence between each set of knowledge points and the covered pool, determine the redundancy suppression value corresponding to each type of question; The dynamic value score for each type of question is determined based on the incremental return value, redundancy suppression value, and dynamic adjustment parameters corresponding to each type of question. Based on the dynamic value scores corresponding to each type of question in the second category, the questions for the second target test paper are determined.

6. The adaptive test paper generation method based on multidimensional weights according to claim 5, characterized in that, The incremental gain value and the redundancy suppression value are determined by the following method: Determine the number of knowledge points that intersect with the knowledge point set corresponding to each second type of question and the pool to be covered, and determine the number of intersecting knowledge points as the incremental revenue value for the corresponding second type of question; Determine the number of knowledge points that intersect with the knowledge point set corresponding to each type of question and the covered pool, and use the number of intersecting knowledge points as the redundancy suppression value for the corresponding type of question.

7. The adaptive test paper generation method based on multidimensional weights according to claim 5, characterized in that, The step of determining the dynamic value score for each type of question based on the incremental return value, redundancy suppression value, and dynamic adjustment parameters for each type of question specifically includes: Get the number of remaining knowledge points in the knowledge point pool to be covered and the total number of knowledge points in the seed knowledge point set; The incentive coefficient and the redundancy penalty coefficient are determined based on the ratio between the number of remaining knowledge points in the pool to be covered and the total number of knowledge points in the seed knowledge point set. The dynamic value score corresponding to each type of question is determined based on the incentive coefficient, the redundancy penalty coefficient, the incremental benefit value, and the redundancy suppression value.

8. An adaptive paper generation device based on multidimensional weights, characterized in that, include: The knowledge segmentation module is used to obtain the knowledge point information associated with each question in the target question bank, and to divide the knowledge points in the entire domain into multiple knowledge point clusters based on the knowledge point information; The sequence rearrangement module is used to rearrange the knowledge points within each knowledge point cluster, calculate the weight of each knowledge point based on the rearranged sequence number and the size of the knowledge point cluster, and determine the seed knowledge point set corresponding to the test paper assembly task based on the knowledge point weight. The state pool module is used to establish a coverage state pool based on the seed knowledge point set. The coverage state pool includes a pending coverage pool and a covered pool. The pending coverage pool is used to store seed knowledge points that have not yet been covered by the test questions, and the covered pool is used to store knowledge points that have been covered by the test questions included in the exam. The first test paper module is used to select the first target test paper questions from the first type of questions in the target question bank, and update the knowledge points associated with the first target test paper questions to the coverage state pool. The second test paper module is used to evaluate the second type of questions in the target question bank based on the pool to be covered and the pool already covered, and to determine the second target test paper questions; The test paper determination module is used to determine the target test paper based on the first target test paper questions and the second target test paper questions.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the adaptive paper generation method based on multidimensional weights as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the adaptive volume composition method based on multidimensional weights as described in any one of claims 1 to 7.

Images