Test question recommendation method and device, electronic device, and storage medium

Abstract

The application provides a test question recommendation method and device, an electronic device and a storage medium. The method comprises: generating recommended test questions corresponding to target test questions based on the target test questions corresponding to a target student and test question recommendation strategies corresponding to the target student; wherein the test question recommendation strategies are determined based on a first test question recommendation strategy and a reference test question recommendation strategy, the first test question recommendation strategy comprises a test question recommendation strategy of a target region to which the target student belongs, and the reference test question recommendation strategy comprises at least one of a test question recommendation strategy of an associated region of the target region and a general test question recommendation strategy. The technical solution of the application can improve the test question recommendation capability of the test question recommendation strategy without sharing the test question library of the associated region, and can improve the accuracy of test question recommendation while ensuring the data privacy and security of the test question resources of each region by combining the test question recommendation strategies determined by the test question library of the associated region and / or the general test question library of the region.

Description

Technical Field

[0001] This application relates to the field of natural language processing technology, and in particular to a test item recommendation method, apparatus, electronic device, and storage medium. Background Technology

[0002] In education, practice questions can solidify students' grasp of knowledge points. Therefore, when recommending practice questions, it's essential to tailor them to each student's needs. For example, based on a student's incorrect answers, recommend questions of the same type to help them master frequently missed concepts. However, in traditional education models, teachers struggle to quickly identify questions that match students' needs from a vast pool of available resources.

[0003] As traditional education moves towards personalization and intelligence, deep learning-based test recommendation models are increasingly being applied in the education field. Typically, to ensure the data privacy and security of their respective test resources, regions do not upload their independent test question banks to servers for sharing. Existing test recommendation methods analyze the questions in each region's independent test question bank to derive a recommendation strategy for that region, and then use this strategy to recommend questions to students based on their region's requirements. However, most regions have a limited number of questions in their independent test question banks, resulting in low recommendation capabilities and consequently, low accuracy. Summary of the Invention

[0004] Based on the defects and shortcomings of the prior art, this application proposes a test question recommendation method, apparatus, electronic device and storage medium, which can improve the accuracy of test question recommendation.

[0005] The technical solution proposed in this application is as follows:

[0006] According to a first aspect of the embodiments of this application, a test question recommendation method is provided, including:

[0007] Obtain the target test questions corresponding to the target students;

[0008] Based on the target test question and the test question recommendation strategy corresponding to the target student, generate recommended test questions corresponding to the target test question;

[0009] The test question recommendation strategy is determined based on a first test question recommendation strategy and a reference test question recommendation strategy. The first test question recommendation strategy includes the test question recommendation strategy for the target region to which the target student belongs. The reference test question recommendation strategy includes at least one of a second test question recommendation strategy and a third test question recommendation strategy. The second test question recommendation strategy includes the test question recommendation strategy for the related regions of the target region. The third test question recommendation strategy includes a general test question recommendation strategy.

[0010] Optionally, the test question recommendation strategy includes test question recommendation operation parameters; the first test question recommendation strategy includes test question recommendation operation parameters for the target region to which the target student belongs, the reference test question recommendation strategy includes at least one of a second test question recommendation strategy and a third test question recommendation strategy, the second test question recommendation strategy includes test question recommendation operation parameters for the associated regions of the target region, and the third test question recommendation strategy includes general test question recommendation operation parameters;

[0011] Based on the target test question and the test question recommendation strategy corresponding to the target student, recommended test questions corresponding to the target test question are generated, including:

[0012] Based on the target test question, test question recommendation calculation is performed according to the test question recommendation calculation parameters to obtain recommended test questions corresponding to the target test question.

[0013] Optionally, based on the target test question, a test question recommendation operation is performed according to the test question recommendation operation parameters to obtain recommended test questions corresponding to the target test question, including:

[0014] The target question is input into the first question recommendation model corresponding to the target region to obtain the recommended question corresponding to the target question;

[0015] The first question recommendation model is used to perform question recommendation calculations according to the question recommendation calculation parameters.

[0016] Optionally, the training process of the first test item recommendation model includes:

[0017] The cloud server determines the first model parameters of the first test question recommendation model, and obtains at least one of the first reference parameters and the second reference parameters; the first reference parameters include the test question recommendation operation parameters of the test question recommendation model corresponding to the associated region of the target region, and the second reference parameters include the test question recommendation operation parameters of the general test question recommendation model;

[0018] The cloud server determines the updated model parameters based on the first model parameters and at least one of the first reference parameters and the second reference parameters;

[0019] The local server in the target region trains the first question recommendation model by using questions from the question bank in the target region based on the updated model parameters.

[0020] Optionally, the cloud server determines the first model parameters of the first test question recommendation model, and obtains at least one of the first reference parameters and the second reference parameters, including:

[0021] The cloud server obtains the model parameters of the first question recommendation model after the current iteration of training as the first model parameters, and determines the first training iteration number of the first question recommendation model; wherein, the first question recommendation model uses questions in the question bank of the target region for question recommendation training in the current iteration of training;

[0022] The cloud server obtains the model parameters of the test question recommendation model corresponding to the associated region after training for the first number of training iterations as the first reference parameter, and / or obtains the model parameters of the general test question recommendation model after training for the first number of training iterations as the second reference parameter.

[0023] Optionally, the training process of the first question recommendation model for the target region further includes:

[0024] The cloud server uses the first influence parameters of the test recommendation model in each region on the general test recommendation model, and weights and aggregates the model parameters of the general test recommendation model after training the first number of training iterations and the model parameters of the test recommendation models in each region after training the first number of training iterations to obtain the updated general model parameters of the general test recommendation model.

[0025] Based on the updated general model parameters, the cloud server trains the general test item recommendation model using test items from the regional general test item library.

[0026] Optionally, based on the first model parameters and the first reference parameters, the updated model parameters are determined, including:

[0027] Based on the correlation between the target region and the associated regions, the first model parameters and the first reference parameters corresponding to each associated region are weighted and aggregated to obtain updated model parameters.

[0028] Optionally, based on the first model parameters and the second reference parameters, the updated model parameters are determined, including:

[0029] Based on the second influence parameter of the first test item recommendation model according to the general test item recommendation model, the first model parameters and the second reference parameters corresponding to the general test item recommendation model are weighted and aggregated to obtain the updated model parameters.

[0030] Optionally, the training process of the first question recommendation model for the target region further includes:

[0031] After stopping iterative training, if the question update coverage rate of the question bank in the target region reaches the preset update coverage rate, or if the question update coverage rate of the question bank in the associated region reaches the preset update coverage rate, then the first question recommendation model for the target region continues to be trained for question recommendation.

[0032] According to a second aspect of the embodiments of this application, a test question recommendation device is provided, comprising:

[0033] The acquisition module is used to acquire the target test questions corresponding to the target students;

[0034] The recommendation module is used to generate recommended test questions corresponding to the target test questions based on the target test questions and the test question recommendation strategy corresponding to the target students.

[0035] The test question recommendation strategy is determined based on a first test question recommendation strategy and a reference test question recommendation strategy. The first test question recommendation strategy includes the test question recommendation strategy for the target region to which the target student belongs. The reference test question recommendation strategy includes at least one of a second test question recommendation strategy and a third test question recommendation strategy. The second test question recommendation strategy includes the test question recommendation strategy for the related regions of the target region. The third test question recommendation strategy includes a general test question recommendation strategy.

[0036] According to a third aspect of the embodiments of this application, an electronic device is provided, including: a memory and a processor;

[0037] The memory is connected to the processor and is used to store programs;

[0038] The processor is used to implement the above-mentioned test question recommendation method by running the program in the memory.

[0039] According to a fourth aspect of the embodiments of this application, a storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the above-described test question recommendation method.

[0040] The test question recommendation method proposed in this application includes: obtaining target test questions corresponding to target students; and generating recommended test questions corresponding to the target test questions based on the target test questions and a test question recommendation strategy corresponding to the target students. The test question recommendation strategy is determined based on a first test question recommendation strategy and a reference test question recommendation strategy. The first test question recommendation strategy includes the test question recommendation strategy for the target region to which the target student belongs. The reference test question recommendation strategy includes at least one of a second test question recommendation strategy and a third test question recommendation strategy. The second test question recommendation strategy includes the test question recommendation strategy for related regions of the target region. The third test question recommendation strategy includes a general test question recommendation strategy. By adopting the technical solution of this application, at least one of the test question recommendation strategies for related regions and a general test question recommendation strategy can be combined with the test question recommendation strategy for the target region. This eliminates the need for a shared test question bank among related regions and allows for the combined use of test question recommendation strategies determined by the test question bank of related regions and / or the general test question bank of the region. While ensuring the data privacy and security of test question resources in each region, this method improves the test question recommendation capability of the test question recommendation strategy, thereby improving the accuracy of test question recommendations. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application 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 only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating a test question recommendation method provided in an embodiment of this application;

[0043] Figure 2 This is a schematic diagram of the training structure of the test question recommendation model provided in the embodiments of this application;

[0044] Figure 3 This is a schematic diagram of the processing flow of the training question recommendation model provided in the embodiments of this application;

[0045] Figure 4 This is a schematic diagram of the structure of a test question recommendation device provided in an embodiment of this application;

[0046] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0047] The technical solution of this application is applicable to the application scenario of test question recommendation. By adopting the technical solution of this application, the test question recommendation strategy can be improved while ensuring the data privacy and security of test question resources in various regions, thereby improving the accuracy of test question recommendation.

[0048] In the field of education, recommending relevant test questions to students based on their needs can improve their mastery of the corresponding knowledge points. For example, recommending questions of the same type as knowledge points that students have just learned can help them master the knowledge points more solidly. Recommending questions of the same type as the questions they got wrong can help them master the knowledge points that are prone to errors.

[0049] To improve the speed of question recommendation, a question recommendation strategy can be pre-determined and used for question recommendation. For example, a deep learning-based question recommendation model can be employed. Each region has its own manually constructed regional question bank; however, due to constraints such as manpower, most regions have a limited number of questions in their regional question banks. Determining a question recommendation strategy requires a large amount of sample data to ensure its accuracy. For instance, when using a deep learning-based question recommendation model, its effectiveness is limited by the amount of training data; a deep learning model trained solely on a region's independent question bank is unlikely to achieve optimal recommendation results. In existing technologies, to increase the sample data for determining the question recommendation strategy, each region uploads its independent regional question bank to a cloud server, thereby utilizing questions from all regions to determine the recommendation strategy and improve its effectiveness. However, due to concerns about the privacy and data security of their local test question resources, each region tends to keep its regional test question bank locally rather than sharing it with the cloud server, which makes it impossible to obtain a test question recommendation strategy with high accuracy.

[0050] Therefore, how to improve the test recommendation capability and accuracy of test recommendation strategies while ensuring the data privacy and security of test resources in various regions is a technical problem that urgently needs to be solved by those skilled in the art.

[0051] Based on this, this application proposes a test question recommendation method. This technical solution can combine at least one of the test question recommendation strategies of related regions and general test question recommendation strategies with the test question recommendation strategy of the target region. It does not require the related regions to share a test question bank, and it can combine the test question recommendation strategies determined by using the test question bank of related regions and / or the general test question bank of the region. While ensuring the data privacy and security of test question resources in each region, it can improve the test question recommendation capability of the test question recommendation strategy, thereby improving the accuracy of test question recommendation.

[0052] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0053] Exemplary methods

[0054] This application proposes a test question recommendation method, which can be executed by an electronic device. This electronic device can be any device with data and instruction processing capabilities, such as a computer, smart terminal, or server. See also... Figure 1 As shown, the method includes:

[0055] S101. Obtain the target test questions corresponding to the target students.

[0056] Specifically, this embodiment can recommend test questions based on the target student's learning data or the current teaching content of the teacher for that target student. Test questions related to the target student's learning data or the teacher's current teaching content will be used as the target student's target test questions. Specifically, test questions related to the target student's learning data (i.e., the target student's learning situation) can be the target student's incorrect answers, test questions corresponding to question types with high error rates, or test questions corresponding to question types that the target student has practiced less. Test questions related to the teacher's current teaching content can be example questions that the teacher is currently explaining, or test questions that contain the knowledge points that the teacher is currently explaining.

[0057] S102. Based on the target test questions and the test question recommendation strategy corresponding to the target students, generate recommended test questions that correspond to the target test questions.

[0058] In this embodiment, different regions store different regional question banks, and the question recommendation strategy is derived from question recommendation training and analysis using questions from these regional question banks. Therefore, the question recommendation strategy used will differ depending on the region to which the target student belongs. This embodiment first needs to determine the question recommendation strategy corresponding to the target student based on the target region, and then use this strategy to generate recommended questions corresponding to the target questions. The division of different regions can be based on district division (different districts are considered different regions), school division (different schools are considered different regions), city division (different cities are considered different regions), or province division (different provinces are considered different regions), etc.

[0059] Specifically, determining the test recommendation strategy for the target students involves three steps: First, identifying the test recommendation strategy for the target region to which the target students belong as the first test recommendation strategy; second, identifying the test recommendation strategies for related regions as the second test recommendation strategy; and third, obtaining a universal test recommendation strategy applicable to all regions. The target region's test recommendation strategy is developed using test questions from the target region's regional test question bank as sample data through test recommendation training and analysis. The related region's test recommendation strategy is developed using test questions from the related region's regional test question bank as sample data through test recommendation training and analysis. The universal test recommendation strategy is developed using test questions from a pre-stored universal regional test question bank as sample data through test recommendation training and analysis. To determine whether a region is a related region of the target region, it can be based on whether there is a correlation between the teaching content of the two regions. If a correlation exists, the region is identified as a related region of the target region. Furthermore, the relevance to the target region varies from region to region. For example, if two regions use the same version of textbooks, the differences in teaching content between them are likely small, resulting in a high degree of relevance. Conversely, if two regions use different versions of textbooks, the differences in teaching content are significant, leading to a lower degree of relevance. Similarly, if two regions belong to the same examination area, the relevance is high; if they do not belong to the same examination area, the relevance is low.

[0060] Then, at least one of the second and third test item recommendation strategies is used as a reference test item recommendation strategy, that is, the reference test item recommendation strategy includes the second and / or the third test item recommendation strategy.

[0061] Finally, based on the first and reference test item recommendation strategies, the corresponding test item recommendation strategy for the target students is determined. For example, when the reference test item recommendation strategy includes a second strategy, the first and reference strategies are combined to integrate the test item recommendation capabilities of the target region and related regions. When the reference strategy includes a third strategy, the first and reference strategies are combined to integrate the test item recommendation capabilities of the target region and those trained from a regionally common test item bank. When the reference strategy includes both the second and third strategies, the first and reference strategies are combined to integrate the test item recommendation capabilities of the target region, related regions, and those trained from a regionally common test item bank. In this way, there is no need to link regional shared question banks. Instead, the question recommendation strategy determined by linking regional question banks and / or regional general question banks can be combined to improve the question recommendation capability of the strategy while ensuring the data privacy and security of question resources in each region, thereby improving the accuracy of question recommendation.

[0062] Furthermore, the question recommendation strategy includes corresponding question recommendation operation parameters. Specifically, the question recommendation strategy for the target student includes question recommendation operation parameters corresponding to the target student; the first question recommendation strategy includes question recommendation operation parameters for the target region to which the target student belongs; the reference question recommendation strategy includes at least one of a second and a third question recommendation strategy; the second question recommendation strategy includes question recommendation operation parameters for the associated regions of the target region; and the third question recommendation strategy includes general question recommendation operation parameters. In this embodiment, based on the target question, question recommendation operation is performed according to the question recommendation operation parameters corresponding to the target student to obtain recommended questions corresponding to the target question. The test question recommendation operation parameters can be parameters of the test question recommendation operation formula. These parameters are configured in the formula, and the formula is used to recommend target test questions. For example, the feature vector of the target test question is extracted, and then the similarity between the feature vector of the target test question and the feature vectors of each test question in the regional test question bank of the target region is calculated. Test questions whose feature vectors have a similarity to the feature vector of the target test question reaching a preset similarity threshold are extracted from the regional test question bank of the target region as recommended test questions. Test questions whose feature vectors have a similarity to the feature vector of the target test question reaching the preset similarity threshold are considered to have a high similarity to the target test question. If a test question has a high information similarity to the target test question, it indicates that the test question and the target test question belong to the same type of test question (i.e., test questions that examine the same knowledge points). The test item recommendation operation parameters can also be the model parameters of a deep learning-based test item recommendation model. By setting the model parameters of the test item recommendation model as the test item recommendation operation parameters, the first test item recommendation model corresponding to the target region is obtained. The target test item is input into the first test item recommendation model, and the first test item recommendation model performs test item recommendation, thereby obtaining the recommended test item corresponding to the target test item.

[0063] Furthermore, in this embodiment, different regions can use the questions in the regional question bank of that region as training samples to train the question recommendation model. Therefore, different regions correspond to different question recommendation models. In order to recommend the target questions for the target student, it is necessary to obtain the first question recommendation model corresponding to the target region to which the target student belongs, and then input the target questions of the target student into the first question recommendation model for question recommendation.

[0064] When the first test question recommendation model for the target region of the target student recommends test questions based on the input target test questions, it first extracts the feature vector of the target test question, then calculates the similarity between the feature vector of the target test question and the feature vectors of each test question in the test question bank of the target region. Test questions whose feature vectors are similar to the feature vectors of the target test question information and reach a preset similarity threshold are extracted from the test question bank of the target region as recommended test questions. Among them, test questions whose feature vectors are similar to the feature vectors of the target test question and reach the preset similarity threshold are test questions with high similarity to the target test question information. If a test question has high similarity to the target test question, it means that the test question and the target test question belong to the same type of test question (i.e., test questions that test the same knowledge points).

[0065] In this embodiment, the first question recommendation model corresponding to the target region is obtained by training question recommendation using questions from the question bank of the target region, based on at least one of the first reference parameter and the second reference parameter, and the first model parameter of the first question recommendation model. The first reference parameter is the model parameter of the question recommendation model corresponding to the associated region of the target region, i.e., the question recommendation operation parameter of the associated region; the second reference parameter is the model parameter of the general question recommendation model, i.e., the general question recommendation operation parameter.

[0066] Further, see Figure 2 As shown, this embodiment includes a cloud server 20, which stores a regionally universal question bank 21 usable by various regions. The questions in the regionally universal question bank 21 conform to the teaching content of each region and are usable by all regions. The cloud server 20 can use the questions in the regionally universal question bank 21 to train a question recommendation model, thus obtaining a universal question recommendation model. That is, the training samples for the universal question recommendation model are the questions in the regionally universal question bank 21 stored in the cloud server 20. Each region also has a local server 10, which stores its own regionally independent question bank 11. The question recommendation model for each region is trained by the local server 10 using the region's independent regional question bank 11. In other words, the training samples for the question recommendation model for each region are the questions in the independent regional question bank stored in the local server 10 of each region.

[0067] Each region's question recommendation model is trained using questions from the region's question bank, based on at least one of the model parameters of the question recommendation models for related regions and the general question recommendation model, as well as the model parameters of the question recommendation model for that region. Figure 2In this embodiment, cloud server 20 can receive model parameters of the test question recommendation models corresponding to each region uploaded by local servers 10 of regions 1 to n. Cloud server 20 can also obtain model parameters of the general test question recommendation model. Then, the association between regions 1 to n determines the updated model parameters. For example, regions associated with region 1 are region 2, region 3, and region n. The test question recommendation model parameters corresponding to region 2, region 3, and region n are the first reference parameters of region 1, and the model parameters of the general test question recommendation model are the second reference parameters of region 1. Cloud server 20 can combine the model parameters of the test question recommendation model corresponding to region 1, as well as at least one of the first and second reference parameters of region 1, to obtain the updated model parameters of region 1. Cloud server 20 then transmits the updated model parameters of region 1 to the local server of region 1 as the model parameters of the test question recommendation model corresponding to region 1. The updating of model parameters of test question recommendation models corresponding to other regions is the same as that of region 1, and will not be described in detail in this embodiment. In practical applications, to recommend test questions to a target student, the target region for that student is determined from Region 1 to Region n. Each region trains its test question recommendation model on its local server using its own independent regional test question bank, without uploading it to a cloud server. This ensures the data privacy and security of the test question resources in each region's independent regional test question bank. Although the regional test question bank itself is not uploaded, the model parameters of the test question recommendation model trained using its independent regional test question bank are uploaded. This allows the model parameters of the test question recommendation model for that region to be combined with the model parameters of the test question recommendation models for related regions and / or the model parameters of a general test question recommendation model, thereby improving the test question recommendation capability of the model.

[0068] As described above, the test question recommendation method proposed in this application obtains the target test questions corresponding to the target students, and generates recommended test questions corresponding to the target test questions based on the target test questions and the test question recommendation strategy corresponding to the target students. The test question recommendation strategy is determined based on a first test question recommendation strategy and a reference test question recommendation strategy. The first test question recommendation strategy includes the test question recommendation strategy for the target region to which the target students belong. The reference test question recommendation strategy includes at least one of a second test question recommendation strategy and a third test question recommendation strategy. The second test question recommendation strategy includes the test question recommendation strategy for related regions of the target region, and the third test question recommendation strategy includes a general test question recommendation strategy. By adopting the technical solution of this embodiment, at least one of the test question recommendation strategies for related regions and the general test question recommendation strategy can be combined with the test question recommendation strategy for the target region. This eliminates the need for a shared test question bank for related regions and allows for the combined use of the test question recommendation strategies determined by the test question bank for related regions and / or the general test question bank for the region. While ensuring the data privacy and security of test question resources in each region, this method improves the test question recommendation capability of the test question recommendation strategy, thereby improving the accuracy of test question recommendations.

[0069] As an optional implementation, see [link to relevant documentation]. Figure 3 As shown in another embodiment of this application, the training process of the first question recommendation model for the target region is disclosed, which may specifically include the following steps:

[0070] S301, The cloud server determines the first model parameters of the first test question recommendation model, and obtains at least one of the first reference parameter and the second reference parameter.

[0071] Specifically, such as Figure 2 As shown, cloud server 20 needs to determine the first model parameters of the first test question recommendation model corresponding to the target region, and obtain at least one of the model parameters of the test question recommendation model uploaded by the local server 10 of the associated region (i.e., the test question recommendation operation parameters of the associated region) and the model parameters of the general test question recommendation model trained by cloud server 20 using test questions from the regional general test question library (i.e., the general test question recommendation operation parameters). The specific steps are as follows:

[0072] First, the cloud server obtains the model parameters of the first test question recommendation model after the current iteration of training as the first model parameters, and determines the first training iteration number of the first test question recommendation model.

[0073] The local server 10 of the target region stores a regional test question bank 11, which is the target region test question bank. The local server 10 of the target region can use this test question bank to train the first test question recommendation model corresponding to the target region. Then, the model parameters of the first test question recommendation model after the current iteration of training are the first model parameters of the first test question recommendation model. The local server 10 of the target region uploads the first model parameters to the cloud server 20, and the cloud server 20 obtains the first model parameters after the current iteration of training. In addition, the cloud server 20 records the number of iterations of the first test question recommendation model after the current iteration of training as the first training iteration number. For example, if the current iteration of training of the first test question recommendation model is the second iteration of training of the first recommendation model, then the first training iteration number recorded by the cloud server 20 is 2. In this embodiment, if the target region is the kth region among regions 1 to n, i.e., region k, and the current iteration of training is the tth iteration of training, i.e., the first training iteration number is t, then the first model parameters of the first test question recommendation model corresponding to the target region can be expressed as W. t,k .

[0074] When training the first question recommendation model using questions from the target region's question bank, questions can be input into the first question recommendation model. The first question recommendation model uses a mask to perform context prediction on the input question. Based on the cross-entropy loss function between the prediction result and the real token, the model parameters of the first question recommendation model are optimized and adjusted. In this embodiment, the training of the question recommendation model can adopt supervised training, unsupervised training, or a combination of supervised and unsupervised training. This embodiment does not impose specific limitations, and supervised training, unsupervised training, or a combination of supervised and unsupervised training are existing training methods, which will not be specifically described in this embodiment.

[0075] Second, the cloud server obtains the model parameters of the test question recommendation model corresponding to the associated region after training for the first number of training iterations as the first reference parameter, and / or obtains the model parameters of the general test question recommendation model after training for the first number of training iterations as the second reference parameter.

[0076] When obtaining the first and second reference parameters, the cloud server 20 needs to obtain model parameters with the same number of iterations as the first training iteration number of the first question recommendation model. This can improve the training effect and speed of the first question recommendation model. If the first training iteration number of the first question recommendation model is 50, but the model parameters obtained when obtaining the first and / or second reference parameters are those obtained after 2 iterations, the accuracy of the parameters after 2 iterations will inevitably be lower than that after 50 iterations. Therefore, combining the first and / or second reference parameters in this case will affect the accuracy of the model parameters of the first question recommendation model, thus hindering the training of the first question recommendation model.

[0077] If the cloud server 20 obtains the model parameters of the test question recommendation model corresponding to the associated region after the first training iteration, and the target region is the kth region among regions 1 to n (i.e., region k), and the current iteration is the tth iteration (i.e., the first training iteration is t), then the model parameters (i.e., the first reference parameters) of the test question recommendation model corresponding to the associated region after the first training iteration can be expressed as W. t,i Where i represents the identifier of the associated region, that is, the i-th region among regions 1 to n is the associated region of the target region k.

[0078] If the model parameters of the general test item recommendation model after the first training iteration are obtained from cloud server 20, and the current iteration is the t-th iteration (i.e., the first training iteration is t), the model parameters (i.e., the second reference parameters) of the general test item recommendation model after the first training iteration can be expressed as W. t,c Where 'c' represents the identifier of the general test item recommendation model.

[0079] In this embodiment, the update of model parameters and training of the test question recommendation model for regions other than the target region are the same as those for the target region, and will not be described in detail in this embodiment.

[0080] Furthermore, since the questions in the regional general question bank contain general features applicable to all regions, the question recommendation models corresponding to all regions will affect the general question recommendation model. Moreover, the degree of influence of the question recommendation models corresponding to different regions on the general question recommendation model is different. In this embodiment, the first influence parameter of the question recommendation models corresponding to each region on the general question recommendation model will be preset in advance.

[0081] The method for determining the model parameters of the general test item recommendation model after the first training iteration is to combine the model parameters of the test item recommendation models corresponding to each region interacting with the cloud server after the previous iteration of the first training iteration with the model parameters of the general test item recommendation model after the previous iteration of the first training iteration. This results in the updated general model parameters of the general test item recommendation model after the previous iteration of the first training iteration. Then, based on these updated general model parameters, the general test item recommendation model is trained using test items from the regional general test item database (i.e., the training corresponding to the first training iteration), thereby obtaining the model parameters after the first training iteration (i.e., the second reference parameters).

[0082] If the number of the first training iterations is t... Among them, W′ t―1,c W represents the updated general model parameters after the previous training iteration (i.e., the (t-1)th training iteration). t,c To utilize the test items in the regionally shared test item bank, the model parameter is W′. t―1,c The model parameters are obtained after training the general test question recommendation model (i.e., the training corresponding to the first training iteration), where n represents the total number of regions interacting with the cloud server, and c... i W represents the first influence parameter of the test item recommendation model corresponding to the i-th region on the general test item recommendation model. t―1,i W represents the model parameters of the test item recommendation model corresponding to the i-th region after the previous iteration (i.e., the (t-1)-th iteration) following the first training iteration. t―1,c This represents the model parameters of the general test item recommendation model after the previous iteration (i.e., the (t-1)th iteration) of the first training iteration.

[0083] S302. The cloud server determines the updated model parameters based on the first model parameters and at least one of the first reference parameters and the second reference parameters.

[0084] When the cloud server 20 only obtains the first model parameters and the first reference parameters, since the target region and each associated region may have different degrees of association, this embodiment pre-constructs a correlation matrix between each region. The correlation matrix records the correlation degree (i.e., the correlation influence coefficient) between every two regions. The correlation matrix A is shown below:

[0085]

[0086] Where α represents the correlation coefficient between region i and region k. When there is no correlation between region i and region k, α is... i,k=0.

[0087] The correlation between the target region and related regions (i.e., the correlation influence coefficient of the related regions on the target region) is used as the weight of the model parameters of the test item recommendation model corresponding to the related regions. The model parameters of the test item recommendation models corresponding to all related regions of the target region are weighted and aggregated. Then, the weighted aggregated model parameters are aggregated with the first model parameters of the first test item recommendation model corresponding to the target region to obtain the updated model parameters. The calculation formula for the updated model parameters is as follows:

[0088]

[0089] Where k represents the identifier of the target region, that is, the k-th region is the target region, W′ t,k W represents the updated model parameters of the first question recommendation model corresponding to the target region after training for the first training iteration number t. t,i This represents the model parameters of the test item recommendation model corresponding to the i-th region after training for the first training iteration number t. Although the formula above weights and aggregates the model parameters of the test item recommendation models corresponding to all regions after training for the first training iteration number t, the model parameters of the test item recommendation models corresponding to regions unrelated to the target region are not actually included because the correlation coefficient between regions unrelated to the target region and the target region is 0. When i = k, α i,k =1, W t,i That is, the first model parameters W of the first question recommendation model corresponding to the target region after training for the first training iteration number t. t,k Therefore, the above formula reflects a weighted aggregation of the model parameters (i.e., the first reference parameters) corresponding to the associated region and the first model parameters corresponding to the target region.

[0090] When the cloud server 20 only obtains the first model parameters and the second reference parameters, since the questions in the regional general question bank contain general features applicable to all regions, the general question recommendation model trained using the regional general question bank can influence the question recommendation models corresponding to all regions. Furthermore, the second influence parameter of the general question recommendation model differs for the question recommendation models corresponding to different regions. In this embodiment, an influence parameter matrix B of the general question recommendation model on the question recommendation models corresponding to different regions is pre-constructed, B = [β1 β2 … β n ],β i ∈[0,1], where β i This represents the second influence parameter of the general test item recommendation model on the i-th region.

[0091] The second influence parameter of the general test item recommendation model on the first test item recommendation model corresponding to the target region is used as the weight of the model parameter (i.e., the second reference parameter) of the general test item recommendation model. The model parameter (i.e., the second reference parameter) of the general test item recommendation model is then weighted and aggregated with the first model parameter of the first test item recommendation model corresponding to the target region to obtain the updated model parameter. The calculation formula for the updated model parameter is as follows:

[0092] W′ t,k =W t,k +β k W t,c

[0093] Where k represents the identifier of the target region, that is, the k-th region is the target region, W′ t,k W represents the updated model parameters of the first question recommendation model corresponding to the target region after training for the first training iteration number t. t,k β represents the first model parameter of the first question recommendation model corresponding to the target region after training for the first training iteration number t. k W represents the second influence parameter of the general test item recommendation model on the k-th region (target region). t,c This represents the model parameters (i.e., the second reference parameters) of the general test item recommendation model after training for the first training iteration number t.

[0094] When cloud server 20 obtains the first model parameters, the first reference parameters, and the second reference parameters, it uses the correlation between the target region and related regions (i.e., the correlation influence coefficient of the related regions on the target region) as the weight of the model parameters (i.e., the first reference parameters) of the test item recommendation model corresponding to the related regions. It uses the second influence parameter of the general test item recommendation model on the first test item recommendation model corresponding to the target region as the weight of the model parameters (i.e., the second reference parameters) of the general test item recommendation model. It then performs a weighted aggregation of the model parameters of the test item recommendation models corresponding to all related regions of the target region. Finally, it aggregates the weighted aggregated model parameters with the first model parameters of the first test item recommendation model corresponding to the target region, and then performs a weighted aggregation with the model parameters (i.e., the second reference parameters) of the general test item recommendation model to obtain the updated model parameters. The calculation formula for the updated model parameters is as follows:

[0095]

[0096] S303. The local server in the target area updates the model parameters and uses the questions in the question bank of the target area to train the first question recommendation model.

[0097] Specifically, in this embodiment, the cloud server 20 transmits the updated model parameters obtained by combining the model parameters of the first question recommendation model corresponding to the target region with the model parameters of the question recommendation model corresponding to the associated region and / or the model parameters of the general question recommendation model to the local server 10 of the target region. The local server 10 of the target region lays out the updated model parameters in the first question recommendation model corresponding to the target region, and then uses the questions in the question bank of the target region to train the first question recommendation model for question recommendation. The model parameters of the trained first question recommendation model are used as the first model parameters for the next iteration of training. t+1,k Let t represent the number of the first training iteration, and k represent the identifier of the target region, i.e., the kth region is the target region.

[0098] As an optional implementation, another embodiment of this application discloses the training process of the first question recommendation model for the target region, which further includes the following steps:

[0099] First, the cloud server uses the first influence parameters of the test recommendation models in each region on the general test recommendation model. It then performs a weighted aggregation of the model parameters of the general test recommendation model after the first training iteration and the model parameters of the test recommendation models in each region after the first training iteration to obtain the updated general model parameters of the general test recommendation model.

[0100] Since the questions in the regional general question bank contain general features applicable to all regions, the question recommendation models corresponding to all regions will affect the general question recommendation model. Moreover, the degree of influence of the question recommendation models corresponding to different regions on the general question recommendation model is different. In this embodiment, the first influence parameter of the question recommendation models corresponding to each region on the general question recommendation model will be preset in advance.

[0101] In this embodiment, the first influence parameter of the test item recommendation model corresponding to each region on the general test item recommendation model is used as the weight of the model parameters of the test item recommendation model corresponding to each region. The model parameters of the test item recommendation models corresponding to each region after training for the first training iteration number t are weighted and aggregated. The weighted aggregated parameters are then aggregated with the model parameters of the general test item recommendation model after training for the first training iteration number t, thereby obtaining the updated general model parameters of the general test item recommendation model after training for the first training iteration number t. The calculation formula for the updated general model parameters is as follows:

[0102]

[0103] Among them, W′ t,cc represents the updated parameters of the general test item recommendation model after training for the first training iteration t. i W represents the first influence parameter of the test item recommendation model corresponding to the i-th region on the general test item recommendation model. t,i W represents the model parameters of the question recommendation model corresponding to the i-th region after training for the first training iteration number t. t,c This represents the model parameters of the general test item recommendation model after training for the first training iteration number t.

[0104] Second, the cloud server updates the general model parameters and uses questions from the regional general question bank to train the general question recommendation model for question recommendation.

[0105] Cloud server 20 utilizes questions from a regionally shared question bank to update the general model parameter W′ for layout. t,c The general test item recommendation model is trained (i.e., trained in the (t+1)th iteration) to obtain the model parameters W of the general test item recommendation model after the (t+1)th iteration. t+1,c The model parameters are used as the model parameters (also the second reference parameters) of the general test item recommendation model for the first test item recommendation model in the next iteration training (i.e., the t+1th iteration training) corresponding to the first training iteration number t.

[0106] In a specific embodiment, before performing the first iteration training on the test item recommendation models for each region and the general test item recommendation model, the cloud server 20 first needs to use the test items in the regional general test item library to perform test item recommendation training to obtain a baseline test item recommendation model, and then transmit the baseline test item recommendation model to the local server of each region.

[0107] During the first iteration of training, the target region uses the baseline item recommendation model as the first item recommendation model. The local server 10 of the target region uses items from the item library of the target region to train the first item recommendation model, and obtains the model parameters of the first item recommendation model as the first model parameters W for the first iteration of training. 1,k The target region's associated regions use the baseline item recommendation model as the corresponding item recommendation model. The local server 10 of the associated regions uses items from its regional item library to train the item recommendation model corresponding to the associated regions, obtaining the model parameters of the associated region's item recommendation model, which are then used as the first reference parameter W for the first iteration of training. 1,i(Where, i represents the identifier of the associated region, i.e., the i-th region is an associated region). Cloud server 20 uses the baseline test item recommendation model as the general test item recommendation model, and uses the test items in the regional general test item library to train the general test item recommendation model, obtaining the model parameters of the general test item recommendation model as the second reference parameter W for the first iteration of training. 1,c .

[0108] The cloud server 20 will train the first model parameters W in the first iteration. 1,k The first reference parameter W in the first iteration of training 1,i And the second reference parameter W during the first iteration of training. 1,c By performing weighted aggregation, the updated model parameters W′ from the first training iteration were obtained. 1,k Cloud server 20 will update model parameter W′. 1,k The data is transmitted to the local server 10 in the target area, so that the local server 10 in the target area can update the layout model parameters W using the questions in the question bank of the target area. ′ 1,k The first item recommendation model is trained in a second iteration, and the model parameters of the trained first item recommendation model are used as the first model parameters W for the second iteration. 2,k The cloud server 20 also performs a weighted aggregation of the model parameters from the first iteration of training of the question recommendation model for each region and the model parameters from the first iteration of training of the general question recommendation model, to obtain the updated general model parameters W from the first iteration of training. ′ 1,c And by using the questions in the regional general question bank, the layout can be updated with the general model parameter W. ′ 1,c The general test item recommendation model is trained in a second iteration, and the model parameters of the trained general test item recommendation model are used as the second reference parameter W for the second iteration training. 2,c The iterative training of the test item recommendation model for other regions is the same as the iterative training of the first test item recommendation model for the target region, thus enabling the acquisition of the model parameters W for the second iteration of the test item recommendation model for other regions. 2,i (Model parameters of the test item recommendation model corresponding to the i-th region). The above iterative training steps are repeated in subsequent iterative training, which will not be described in detail in this embodiment.

[0109] As an optional implementation, another embodiment of this application discloses the training process of the first question recommendation model for the target region, which further includes:

[0110] After each iteration of training the first question recommendation model for the target region, it is checked whether the number of the first training iterations has reached a preset iteration threshold. If the preset iteration threshold is reached, training of the first question recommendation model for the target region is stopped, and the final first question recommendation model is deployed to the local server in the target region for application.

[0111] Alternatively, after each iteration of training the first question recommendation model for the target region, the convergence of the first question recommendation model is checked. Convergence can be determined by the stability of the loss function. If the loss function is stable with minimal fluctuation, the model has converged; if the loss function is unstable with significant fluctuation, the model has not converged. If convergence is detected, training of the first question recommendation model is stopped, and the final first question recommendation model is deployed to the local server in the target region for application.

[0112] Alternatively, after each iteration of training the first question recommendation model for the target region, it is checked whether the number of the first training iterations has reached a preset iteration threshold, and whether the first question recommendation model has converged. If the number of the first training iterations has reached the preset iteration threshold, or if the first question recommendation model has converged, then the training of the first question recommendation model is stopped, and the final first question recommendation model is deployed to the local server in the target region for application.

[0113] As an optional implementation, another embodiment of this application discloses the training process of the first question recommendation model for the target region, which further includes:

[0114] After ceasing iterative training of the first item recommendation model, it is necessary to check the item update coverage rate of the target region's item library or the item update coverage rate of the related regions' item libraries. If the item update coverage rate of the target region's item library reaches a preset update coverage rate, or the item update coverage rate of the related regions' item libraries reaches a preset update coverage rate, then the first item recommendation model for the target region continues to be trained. This allows for iterative training of the model using the updated item library, improving the model's item recommendation capability. The item update coverage rate is the proportion of updated items in the item library to the total number of original items.

[0115] Exemplary device

[0116] Corresponding to the above-mentioned test question recommendation method, this application also discloses a test question recommendation device, see [link to relevant documentation]. Figure 4 As shown, the device includes:

[0117] Module 100 is used to retrieve the target test questions corresponding to the target students.

[0118] Recommendation module 110 is used to generate recommended test questions corresponding to the target test questions based on the target test questions and the test question recommendation strategy corresponding to the target students;

[0119] The test item recommendation strategy is determined based on the first test item recommendation strategy and the reference test item recommendation strategy. The first test item recommendation strategy includes the test item recommendation strategy for the target area to which the target student belongs. The reference test item recommendation strategy includes at least one of the second and third test item recommendation strategies. The second test item recommendation strategy includes the test item recommendation strategy for the related areas of the target area. The third test item recommendation strategy includes the general test item recommendation strategy.

[0120] The test question recommendation device proposed in this application includes an acquisition module 100 that acquires target test questions corresponding to target students, and a recommendation module 110 that generates recommended test questions corresponding to the target test questions based on the target test questions and a test question recommendation strategy corresponding to the target students. The test question recommendation strategy is determined based on a first test question recommendation strategy and a reference test question recommendation strategy. The first test question recommendation strategy includes the test question recommendation strategy for the target region to which the target student belongs. The reference test question recommendation strategy includes at least one of a second test question recommendation strategy and a third test question recommendation strategy. The second test question recommendation strategy includes the test question recommendation strategy for related regions of the target region, and the third test question recommendation strategy includes a general test question recommendation strategy. By adopting the technical solution of this application, at least one of the test question recommendation strategies for related regions and a general test question recommendation strategy can be combined with the test question recommendation strategy for the target region. This eliminates the need for a shared test question bank for related regions and allows for the combined use of test question recommendation strategies determined by the test question bank for related regions and / or the general test question bank for the region. While ensuring the data privacy and security of test question resources in each region, this improves the test question recommendation capability of the test question recommendation strategy, thereby increasing the accuracy of test question recommendations.

[0121] As an optional implementation, another embodiment of this application discloses a test question recommendation device, which includes a test question recommendation strategy, comprising test question recommendation operation parameters; a first test question recommendation strategy including test question recommendation operation parameters for the target region to which the target student belongs; a reference test question recommendation strategy including at least one of a second test question recommendation strategy and a third test question recommendation strategy; the second test question recommendation strategy including test question recommendation operation parameters for the associated regions of the target region; and the third test question recommendation strategy including general test question recommendation operation parameters.

[0122] The recommendation module 110 is specifically used to perform test question recommendation calculations based on the target test question and according to the test question recommendation calculation parameters to obtain recommended test questions corresponding to the target test question.

[0123] As an optional implementation, another embodiment of this application discloses a recommendation module 110, which is further used to input the target test question into a first test question recommendation model corresponding to the target region to obtain recommended test questions corresponding to the target test question; wherein, the first test question recommendation model is used to perform test question recommendation calculation according to the test question recommendation calculation parameters.

[0124] As an optional implementation, another embodiment of this application discloses that the test question recommendation device further includes: a parameter acquisition module, a first parameter update module, and a first model training module. The parameter acquisition module and the first parameter update module are modules in a cloud server, and the first model training module is a module in a local server in the target region.

[0125] The parameter acquisition module is used to determine the first model parameters of the first test item recommendation model, and to acquire at least one of the first reference parameters and the second reference parameters; the first reference parameters include the test item recommendation operation parameters of the test item recommendation model corresponding to the associated region of the target region, and the second reference parameters include the test item recommendation operation parameters of the general test item recommendation model;

[0126] The first parameter update module is used to determine the updated model parameters based on the first model parameters and at least one of the first reference parameters and the second reference parameters;

[0127] The first model training module is used to train the first item recommendation model by updating the model parameters and using items from the item bank in the target area.

[0128] As an optional implementation, another embodiment of this application also discloses a parameter acquisition module, specifically used for:

[0129] The model parameters of the first test item recommendation model after the current iteration of training are obtained as the first model parameters, and the first training iteration number of the first test item recommendation model is determined; wherein, the first test item recommendation model uses test items in the test item bank of the target region for test item recommendation training in the current iteration of training;

[0130] The model parameters of the test question recommendation model corresponding to the associated region after training for the first number of training iterations are obtained as the first reference parameters, and / or the model parameters of the general test question recommendation model after training for the first number of training iterations are obtained as the second reference parameters.

[0131] As an optional implementation, another embodiment of this application also discloses that the test question recommendation model further includes: a second parameter update module and a second model training module. The second parameter update module and the second model training module are modules located in a cloud server.

[0132] The second parameter update module is used to perform a weighted aggregation of the model parameters of the general test question recommendation model after training the general test question recommendation model for the first number of training iterations and the model parameters of the test question recommendation models for each region after training the first number of training iterations, to obtain the updated general model parameters of the general test question recommendation model.

[0133] The second model training module is used to train the general test item recommendation model by updating the general model parameters and using test items from the regional general test item bank.

[0134] As an optional implementation, another embodiment of this application also discloses a first parameter update module, which is specifically used to weight and aggregate the first model parameters and the first reference parameters corresponding to each associated region based on the correlation between the target region and the associated region to obtain updated model parameters.

[0135] As an optional implementation, another embodiment of this application also discloses a first parameter update module, which is specifically used to perform weighted aggregation of the first model parameters and the second reference parameters corresponding to the general test item recommendation model based on the second influence parameter of the first test item recommendation model, to obtain updated model parameters.

[0136] As an optional implementation, another embodiment of this application also discloses that the test item recommendation model further includes: a test item bank update detection module, which is used to continue test item recommendation training on the first test item recommendation model of the target area if the test item update coverage rate of the test item bank in the target area reaches a preset update coverage rate, or if the test item update coverage rate of the test item bank in the associated area reaches a preset update coverage rate after stopping iterative training.

[0137] The test question recommendation device provided in this embodiment belongs to the same application concept as the test question recommendation method provided in the above embodiments of this application. It can execute the test question recommendation method provided in any of the above embodiments of this application and has the corresponding functional modules and beneficial effects for executing the test question recommendation method. Technical details not described in detail in this embodiment can be found in the specific processing content of the test question recommendation method provided in the above embodiments of this application, and will not be repeated here.

[0138] Exemplary electronic devices, storage media, and computing products

[0139] Corresponding to the above-mentioned test question recommendation method, this application also discloses an electronic device, see [link to relevant documentation]. Figure 5 As shown, the electronic device includes:

[0140] Memory 200 and processor 210;

[0141] The memory 200 is connected to the processor 210 and is used to store programs;

[0142] The processor 210 is configured to implement the test item recommendation method disclosed in any of the above embodiments by running a program stored in the memory 200.

[0143] Specifically, the aforementioned electronic device may further include: a bus, a communication interface 220, an input device 230, and an output device 240.

[0144] The processor 210, memory 200, communication interface 220, input device 230, and output device 240 are interconnected via a bus. Among them:

[0145] A bus can include a pathway for transmitting information between various components of a computer system.

[0146] The processor 210 can be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the execution of the program of the present application. It can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0147] Processor 210 may include a main processor, as well as a baseband chip, modem, etc.

[0148] The memory 200 stores a program for executing the technical solution of this application, and may also store an operating system and other critical business functions. Specifically, the program may include program code, which includes computer operation instructions. More specifically, the memory 200 may include read-only memory (ROM), other types of static storage devices capable of storing static information and instructions, random access memory (RAM), other types of dynamic storage devices capable of storing information and instructions, disk storage, flash memory, etc.

[0149] Input device 230 may include a device for receiving user input data and information, such as a keyboard, mouse, camera, scanner, light pen, voice input device, touch screen, pedometer, or gravity sensor.

[0150] Output device 240 may include devices that allow information to be output to a user, such as a display screen, printer, speaker, etc.

[0151] The communication interface 220 may include a device that uses any transceiver to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0152] The processor 210 executes the program stored in the memory 200 and calls other devices, which can be used to implement the various steps of the test question recommendation method provided in the above embodiments of this application.

[0153] Another embodiment of this application provides a storage medium storing a computer program, which, when executed by a processor, implements the various steps of the test question recommendation method provided in any of the above embodiments.

[0154] For the foregoing method embodiments, in order to simplify the description, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0155] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For apparatus embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0156] The steps in the methods of the various embodiments of this application can be adjusted, merged, or deleted in order according to actual needs, and the technical features described in each embodiment can be replaced or combined.

[0157] The modules and sub-modules in the apparatus and terminal in the various embodiments of this application can be merged, divided, and deleted according to actual needs.

[0158] It should be understood, in the several embodiments provided in this application, that the disclosed terminals, devices, and methods can be implemented in other ways. For example, the terminal embodiments described above are merely illustrative; for instance, the division of modules or sub-modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple sub-modules or modules may be combined or integrated into another module, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0159] The modules or submodules described as separate components may or may not be physically separate. The components that constitute a module or submodule may or may not be physical modules or submodules; that is, they may be located in one place or distributed across multiple network modules or submodules. Some or all of the modules or submodules can be selected to achieve the purpose of this embodiment's solution, depending on actual needs.

[0160] Furthermore, the functional modules or sub-modules in the various embodiments of this application can be integrated into one processing module, or each module or sub-module can exist physically separately, or two or more modules or sub-modules can be integrated into one module. The integrated modules or sub-modules described above can be implemented in hardware or in the form of software functional modules or sub-modules.

[0161] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0162] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software unit executed by a processor, or a combination of both. The software unit can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0163] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0164] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A test item recommendation method, characterized in that, include: Obtain the target test questions corresponding to the target students; Based on the target test question and the test question recommendation strategy corresponding to the target student, generate recommended test questions corresponding to the target test question; The test question recommendation strategy is determined based on a first test question recommendation strategy and a reference test question recommendation strategy. The test question recommendation strategy includes test question recommendation operation parameters. The first test question recommendation strategy includes test question recommendation operation parameters for the target region to which the target student belongs. The reference test question recommendation strategy includes at least one of a second and a third test question recommendation strategy. The second test question recommendation strategy includes test question recommendation operation parameters for related regions of the target region. The third test question recommendation strategy includes general test question recommendation operation parameters. The target region is the region to which the target student belongs after being divided according to a regional division method, which includes any one of: district division, school division, and city division. The related regions of the target region include regions whose teaching content is related to the target region. The test item recommendation operation parameters include: model parameters of the deep learning-based test item recommendation model; Based on the target test question and the test question recommendation strategy corresponding to the target student, recommended test questions corresponding to the target test question are generated, including: The model parameters of the test question recommendation model are set as test question recommendation operation parameters to obtain the first test question recommendation model corresponding to the target region. The target test question is input into the first test question recommendation model corresponding to the target region to obtain the recommended test question corresponding to the target test question. The training process of the first test item recommendation model includes: The cloud server determines the first model parameters of the first test question recommendation model, and obtains at least one of the first reference parameters and the second reference parameters; the first reference parameters include the test question recommendation operation parameters of the test question recommendation model corresponding to the associated region of the target region, and the second reference parameters include the test question recommendation operation parameters of the general test question recommendation model; The cloud server determines the updated model parameters based on the first model parameters and at least one of the first reference parameters and the second reference parameters; the updated model parameters are obtained by updating the model parameters of the deep learning-based test recommendation model. The local server in the target region trains the first question recommendation model by using questions from the question bank in the target region based on the updated model parameters. Based on the first model parameters and the first reference parameters, the updated model parameters are determined, including: Based on the correlation between the target region and the associated regions, the first model parameters and the first reference parameters corresponding to each associated region are weighted and aggregated to obtain updated model parameters.

2. The method according to claim 1, characterized in that, The cloud server determines the first model parameters of the first test question recommendation model, and obtains at least one of the first reference parameters and the second reference parameters, including: The cloud server obtains the model parameters of the first question recommendation model after the current iteration of training as the first model parameters, and determines the first training iteration number of the first question recommendation model; wherein, the first question recommendation model uses questions in the question bank of the target region for question recommendation training in the current iteration of training; The cloud server obtains the model parameters of the test question recommendation model corresponding to the associated region after training for the first number of training iterations as the first reference parameter, and / or obtains the model parameters of the general test question recommendation model after training for the first number of training iterations as the second reference parameter.

3. The method according to claim 2, characterized in that, The training process of the first question recommendation model for the target region also includes: The cloud server uses the first influence parameters of the test recommendation model in each region on the general test recommendation model, and weights and aggregates the model parameters of the general test recommendation model after training the first number of training iterations and the model parameters of the test recommendation models in each region after training the first number of training iterations to obtain the updated general model parameters of the general test recommendation model. Based on the updated general model parameters, the cloud server trains the general test item recommendation model using test items from the regional general test item library.

4. The method according to claim 1, characterized in that, Based on the first model parameters and the second reference parameters, the updated model parameters are determined, including: Based on the second influence parameter of the first test item recommendation model according to the general test item recommendation model, the first model parameters and the second reference parameters corresponding to the general test item recommendation model are weighted and aggregated to obtain the updated model parameters.

5. The method according to claim 1, characterized in that, The training process of the first question recommendation model for the target region also includes: After stopping iterative training, if the question update coverage rate of the question bank in the target region reaches the preset update coverage rate, or if the question update coverage rate of the question bank in the associated region reaches the preset update coverage rate, then the first question recommendation model for the target region continues to be trained for question recommendation.

6. A test question recommendation device, characterized in that, include: The acquisition module is used to acquire the target test questions corresponding to the target students. The recommendation module is used to generate recommended test questions corresponding to the target test questions based on the target test questions and the test question recommendation strategy corresponding to the target students; The test question recommendation strategy is determined based on a first test question recommendation strategy and a reference test question recommendation strategy. The test question recommendation strategy includes test question recommendation operation parameters. The first test question recommendation strategy includes test question recommendation operation parameters for the target region to which the target student belongs. The reference test question recommendation strategy includes at least one of a second and a third test question recommendation strategy. The second test question recommendation strategy includes test question recommendation operation parameters for related regions of the target region. The third test question recommendation strategy includes general test question recommendation operation parameters. The target region is the region to which the target student belongs after being divided according to a regional division method, which includes any one of: district division, school division, and city division. The related regions of the target region include regions whose teaching content is related to the target region. The test item recommendation operation parameters include: model parameters of the deep learning-based test item recommendation model; The recommendation module is specifically used for: The model parameters of the test question recommendation model are set as test question recommendation operation parameters to obtain the first test question recommendation model corresponding to the target region. The target test question is input into the first test question recommendation model corresponding to the target region to obtain the recommended test question corresponding to the target test question. The parameter acquisition module is used to determine the first model parameters of the first test question recommendation model, and to acquire at least one of the first reference parameters and the second reference parameters; the first reference parameters include the test question recommendation operation parameters of the test question recommendation model corresponding to the associated region of the target region, and the second reference parameters include the test question recommendation operation parameters of the general test question recommendation model; The first parameter update module is used to determine the updated model parameters based on the first model parameters and at least one of the first reference parameters and the second reference parameters; the updated model parameters are obtained by updating the model parameters of the deep learning-based test recommendation model. The first model training module is used to train the first question recommendation model by using questions from the question bank in the target area based on the updated model parameters. The first parameter update module is specifically used to perform weighted aggregation of the first model parameters and the first reference parameters corresponding to each of the associated regions based on the correlation degree between the target region and the associated regions, so as to obtain updated model parameters.

7. An electronic device, characterized in that, include: Memory and processor; The memory is connected to the processor and is used to store programs; The processor is configured to implement the test item recommendation method as described in any one of claims 1 to 5 by running the program in the memory.

8. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the test item recommendation method as described in any one of claims 1 to 5.

Images