Personalized dynamic question setting method and system based on large language model

Abstract

The invention discloses a personalized dynamic question setting method and system based on a large language model. The method comprises the following steps: S1, collecting multi-source data such as historical answer records, subject scores and learning directions of students, and constructing structured feature vectors; s2, evaluating the mastery degree of the students on the complete knowledge system through the pre-training model, and calculating the knowledge point coverage; s3, a default test question set is extracted from the question bank, knowledge point mastering scores of the students are converted into weights through an attention mechanism, question difficulty distribution is dynamically adjusted, and weak link training is strengthened; and S4, fusing the student characteristics, the grasp score, the coverage degree and the adjusted difficulty distribution, and generating and pushing test questions through a personalized question setting model finely adjusted by a large language model. According to the method, the powerful semantic comprehension and generation capability of the large language model is utilized, the learning track of students can be analyzed, the question difficulty, question type distribution and the knowledge point coverage range are dynamically adjusted, and therefore precise and personalized question recommendation is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of generative artificial intelligence, and in particular relates to a personalized dynamic question-setting method and system based on a large language model. Background Art

[0002] With the rapid development of online education and intelligent learning systems, personalized teaching based on artificial intelligence has become a research hotspot in the field of education. The traditional way of setting questions mainly relies on teachers to manually compile or fix question banks, which is difficult to adapt to students' personalized needs in real time, resulting in low matching of learning content and uncontrollable difficulty, making it difficult to effectively improve students' learning efficiency and learning results.

[0003] At present, some intelligent education systems try to use knowledge graphs, rule matching and other technologies to achieve personalized question setting, but there are still many limitations: 1) Inaccurate personalized matching: Most existing systems set questions based on static rules, and fail to fully combine students' historical learning data, cognitive abilities, interest preferences and other multi-dimensional information for dynamic adjustment, resulting in insufficient accuracy of question recommendations. 2) Limited ability to adjust question difficulty: Traditional methods are difficult to dynamically optimize question difficulty based on students' learning progress and answering performance, which may lead to excessively high question difficulty, increasing learning pressure, or too low difficulty, affecting challenge, and ultimately reducing learning effects. 3) Weak feedback and optimization mechanism: Existing intelligent question setting systems mainly rely on simple accuracy statistics for adjustment, lack advanced algorithms for in-depth optimization, and are difficult to provide accurate personalized learning feedback, limiting the continuous improvement of learning effects.

[0004] In recent years, the rapid development of large language models (LLMs) has provided a new technical path for personalized question setting. Pre-trained language models can more accurately match students' learning needs with deep semantic understanding, conditional generation strategies and attention mechanisms, and dynamically adjust the difficulty of questions and the distribution of knowledge points in combination with their learning trajectories and goals. However, the existing question setting methods based on large language models still face the following challenges: 1) Insufficient use of personalized data: Most systems only rely on students' historical answer records and fail to fully integrate multi-dimensional learning data, such as learning habits, interest preferences and goal planning, resulting in limited accuracy of personalized recommendations. 2) Imperfect dynamic question setting strategies: Existing methods lack efficient adaptive optimization mechanisms, making it difficult to adjust question difficulty, question type distribution and knowledge point coverage based on real-time feedback, affecting the personalized learning experience. 3) Insufficient multimodal information fusion capabilities: The current system mainly relies on text data for question setting, while ignoring the enhancement of learning effects by multimodal information such as images and voices, resulting in a single way of question presentation, which is difficult to meet the needs of different learning scenarios. Summary of the invention

[0005] The purpose of the present invention is to solve the problems existing in the prior art and disclose a personalized dynamic question-setting method and system based on a large language model.

[0006] The specific technical solutions adopted by the present invention are as follows:

[0007] In a first aspect, the present invention provides a personalized dynamic question-setting method based on a large language model, comprising:

[0008] S1. Obtain multi-dimensional information data of target student users, including historical answer records, historical scores of various subjects, and learning direction guidance information, and construct them into structured student feature vectors through feature engineering;

[0009] S2. Input the student feature vector into the pre-trained knowledge point mastery evaluation model to obtain the mastery score of the target student user on each knowledge point in the complete knowledge point distribution, and then calculate the knowledge point coverage based on the mastery score of the target student user on all knowledge points;

[0010] S3. Extract the default test question set to be recommended to the target student from the current question bank, and analyze the difficulty distribution of the default test question set on all knowledge points; convert the mastery scores of the target student users on all knowledge points into attention weights for each knowledge point through the attention mechanism, and then dynamically adjust the difficulty distribution of the questions using the attention weights, increase the difficulty of the questions corresponding to the knowledge points whose attention weights exceed the weight threshold, and decrease the difficulty of the questions corresponding to the knowledge points whose attention weights are lower than the weight threshold, to obtain the adjusted difficulty distribution of the questions;

[0011] S4. The default test questions of the target students, the mastery scores of the target student users on all knowledge points, the knowledge point coverage, and the adjusted question difficulty distribution are combined with the prompt word template to construct a prompt word text. The prompt word text is input into a personalized question setting model obtained by fine-tuning the large language model (LLM) on the question setting task. The personalized question setting model outputs a new test question set and pushes it to the target student users.

[0012] As a preferred embodiment of the first aspect, the learning direction guidance information includes the subjects that need to be studied emphatically and / or the knowledge points that need to be studied emphatically in each subject given to the target student user.

[0013] As a preferred embodiment of the above-mentioned first aspect, the knowledge point mastery degree assessment model is obtained by supervised training of a multi-layer perceptron, a random forest, a support vector machine or a convolutional neural network on a labeled data set.

[0014] As a preferred embodiment of the above-mentioned first aspect, the knowledge point coverage is calculated based on the expected distribution of knowledge point mastery levels predefined by the teacher user and the mastery levels of the target student users on all knowledge points. The calculation method is: for each knowledge point in the complete knowledge point distribution, the expected mastery level score corresponding to the knowledge point in the expected distribution of knowledge point mastery levels is compared with the mastery level score of the target student user on the knowledge point, and the smaller one is taken as the representation value, and then the ratio of the sum of the representation values ​​of all knowledge points to the sum of the expected mastery levels of all knowledge points is calculated, and the calculated ratio is used as the current knowledge point coverage of the target student user.

[0015] As a preferred embodiment of the first aspect, when the large language model (LLM) is fine-tuned on the question-setting task, the model parameters of the large language model are optimized with the maximization of the sum of knowledge point coverage and knowledge point difficulty matching as the optimization goal to obtain a personalized question-setting model.

[0016] As a preferred embodiment of the first aspect, the knowledge point difficulty matching degree is the mean square error between the adjusted question difficulty distribution of the input model and the question difficulty distribution of the new test question set output by the model.

[0017] As a preferred embodiment of the first aspect mentioned above, in actual application, the personalized question-setting model needs to construct an incremental learning data set by collecting feedback data, with maximizing the sum of knowledge point coverage and knowledge point difficulty matching as the optimization goal, and regularly updating and optimizing the model parameters of the personalized question-setting model through the Bayesian optimization algorithm.

[0018] In a second aspect, the present invention provides a personalized dynamic question setting system based on a large language model, which includes:

[0019] Information collection module, used to record multi-dimensional information data of each student user on the online learning platform;

[0020] The target user selection module is used for users to select target student users who need to receive personalized test questions.

[0021] The personalized push module is used to perform personalized test question push for the target student users selected by the user according to the multi-dimensional information data recorded in the information collection module and the personalized dynamic question-setting method based on the large language model as described in any of the schemes of the first aspect above.

[0022] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the personalized dynamic question-setting method based on a large language model as described in any of the schemes of the first aspect above is implemented.

[0023] In a fourth aspect, the present invention provides a computer electronic device comprising a memory and a processor;

[0024] The memory is used to store computer programs;

[0025] The processor is used to implement the personalized dynamic question-setting method based on a large language model as described in any solution of the first aspect above when executing the computer program.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The present invention provides a personalized dynamic question-setting method and system based on a large language model. The present invention constructs an intelligent and personalized dynamic question-setting model through a dynamic question-setting mechanism driven by multi-dimensional data, combining multi-source information such as students' historical answer records, historical scores in various subjects, and guidance on learning directions. The method uses the powerful semantic understanding and generation capabilities of the large language model to analyze students' learning trajectories, dynamically adjust the difficulty of questions, question type distribution, and knowledge point coverage, thereby achieving precise and personalized question recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A schematic diagram of the steps of a personalized dynamic question-setting method based on a large language model;

[0029] Figure 2 This is a schematic diagram of the module composition of a personalized dynamic question-setting system based on a large language model;

[0030] Figure 3 A schematic diagram of the components of computer electronic equipment. DETAILED DESCRIPTION

[0031] In order to make the above-mentioned purpose, features and advantages of the present invention more obvious and easy to understand, the specific implementation mode of the present invention is described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. The technical features in each embodiment of the present invention can be combined accordingly without conflicting with each other.

[0032] In the description of the present invention, it should be understood that the terms "first" and "second" are only used for the purpose of distinguishing descriptions, and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features.

[0033] The present invention provides a personalized dynamic question-setting method based on a large language model, which can be used to recommend personalized test questions to students on an online learning platform to meet the students' own learning progress and learning needs, thereby maximizing the effect of online learning.

[0034] like Figure 1 As shown, in a preferred embodiment of the present invention, the personalized dynamic question-setting method based on a large language model includes steps S1 to S4, and the specific implementation of each step is described in detail below.

[0035] S1. Obtain multi-dimensional information data of target student users, including historical answer records, historical scores in various subjects, and learning direction guidance information, and construct them into structured student feature vectors through feature engineering.

[0036] It should be noted that the target student users refer to the objects that need to be asked questions by the current personalized dynamic question-setting method on the online learning platform. Historical answer records refer to the answer records of student users for exercises on the platform, including answer time, answer accuracy, answer ideas, and answer time. The historical grades of each subject refer to the historical performance data of students in each subject, such as unit tests, midterm and final exams, and mock exams. Learning direction guidance information refers to the situation in which teachers or parents and other personnel guide students' next learning focus, such as learning suggestions, tutoring content, learning habit guidance and other information. It needs to include the subjects that need to be focused on for the target student users, and can also include the knowledge points that need to be focused on in each subject. It can be adjusted according to actual needs.

[0037] In an embodiment of the present invention, historical answering situations, subject score data, learning guidance, future goals and plans, and other steps can be collected to collect multi-dimensional information data for student users. Historical answering situation collection: collect students' answering records in the past learning process, including answering time, answering accuracy, answering ideas (using natural language processing technology to extract key information in the answering text) and answering time and other data, and convert them into structured data storage. Subject score data collection: obtain students' historical score data in various subjects from the score management system of the online learning platform, including unit tests, mid-term and final exams, and mock exams. The score data is normalized so that it can be integrated with other dimensional data. Learning guidance situation collection: collect teachers and parents' learning guidance for students through questionnaires, interviews or online feedback systems, including learning suggestions, tutoring content, learning habit guidance and other information. This information is stored after extracting key features through text analysis technology. Future goal and planning collection: collect students' learning goals and future plans, including short-term goals (such as improving the scores of a certain subject) and long-term goals (such as applying for target colleges or majors). These goals are converted into structured knowledge point requirements through natural language processing technology. The data collected in the above learning guidance and future goals and plans need to extract core keywords, that is, the subjects that need to be focused on and the knowledge points that need to be focused on in each subject, so as to serve as key features, so as to be applicable to the recommendation tasks of the present invention. Among them, information such as the target colleges or majors in the long-term goals can be converted into key features through the key subjects that different colleges and majors focus on.

[0038] In addition, when constructing the structured student feature vector through feature engineering, information of different dimensions needs to be vectorized according to their respective data characteristics. Numerical data can be directly used after standardization or normalization. Type features can be converted into feature vectors through One-Hot Encoding or Label Encoding. Text data needs to be preprocessed first, such as word segmentation, stop word removal, keyword extraction, etc., and then converted into numerical vectors through word vector models. Therefore, for all data, it can eventually be divided into dimensions to form vectors, and then structured to form the final student feature vector. ,in represents the eigenvalue of the i-th dimension, and n represents the total number of dimensions.

[0039] S2. Input the student feature vector into the pre-trained knowledge point mastery evaluation model to obtain the target student user's mastery score on each knowledge point in the complete knowledge point distribution, and then calculate the knowledge point coverage based on the target student user's mastery score on all knowledge points.

[0040] It should be noted that the above-mentioned complete knowledge point distribution includes all the knowledge points in the course, and the knowledge point distribution is generally determined in the course outline and teaching goal evaluation in the course design stage, and can be specified by the teacher user. The present invention needs to use the knowledge point mastery evaluation model to predict the mastery score of the target student user on each knowledge point based on the student feature vector. The mastery score can adopt a set of discrete score labels, and the specific score range can be designed according to actual needs. For example, in one embodiment, a mastery score with a score range of 0 to 5 points can be set, 0 points represents no mastery at all, 5 points represents complete mastery, and the middle 1, 2, 3, and 4 points represent the mastery degree in increasing order.

[0041] In an embodiment of the present invention, the above-mentioned knowledge point mastery evaluation model can be obtained by training through a simple linear transformation model or a nonlinear transformation model. For example, a multi-layer perceptron, a random forest, a support vector machine or a convolutional neural network model can be used, and after initialization, supervised training is performed on a labeled data set. The loss used for training can be the mean square error (MSE) between the predicted value and the labeled value of the mastery score distribution of all knowledge points.

[0042] In addition, in an embodiment of the present invention, the above-mentioned knowledge point coverage is calculated based on the expected distribution of knowledge point mastery predefined by the teacher user and the mastery score of the target student user on all knowledge points, wherein the expected distribution of knowledge point mastery can be predefined by the teacher user, wherein it is necessary to set an expected mastery score for each knowledge point according to its importance. For example, assuming that the complete knowledge point distribution contains 4 knowledge points A, B, C, and D, the teacher user can set the expected mastery score that the students need to achieve according to their respective importance in the course. For example, knowledge points A and B are extremely important and must be fully mastered, and the expected mastery score can be set to 5 points. Knowledge point C is generally important, and the expected mastery score of both can be set to 3 points. Knowledge point D is not very important, and the expected mastery score can be set to 2 points. Thus, the expected distribution of knowledge point mastery corresponding to the complete knowledge point distribution [A, B, C, D] is [5, 5, 3, 2].

[0043] In an embodiment of the present invention, the knowledge point coverage currently mastered by a target student user can be calculated based on the predefined expected distribution of knowledge point mastery and the mastery score of the target student user on all knowledge points given by the knowledge point mastery evaluation model. The specific calculation method is as follows: for each knowledge point in the complete knowledge point distribution, the expected mastery score corresponding to the knowledge point in the expected distribution of knowledge point mastery is compared with the mastery score of the target student user on the knowledge point, and the smaller one is taken as the characterization value, and then the sum of the characterization values ​​of all knowledge points and the sum of the expected mastery scores of all knowledge points are calculated as the ratio, and the calculated ratio is used as the current knowledge point coverage of the target student user. The calculation method of the knowledge point coverage can be expressed by the formula:

[0044]

[0045] Where: K is the expected distribution of the predefined knowledge point mastery level, is the mastery score distribution of the target student users on all knowledge points, and m is the total number of knowledge points in the complete knowledge point distribution. is the expected score of the mastery level corresponding to the i-th knowledge point, Score the target student user's mastery of the i-th knowledge point. In the above formula, Used to calculate the expected distribution K of knowledge point mastery and the distribution of student knowledge mastery scores The overlapping part reflects the students’ actual mastery of each knowledge point. It represents the expected total score of all knowledge points, and is used as a measurement standard to ensure that the knowledge point coverage is between [0, 1]. The higher it is, the more the question content can meet students' learning needs on their weak knowledge points.

[0046] S3. Extract the default test question set to be recommended to the target students from the current question bank, and analyze the difficulty distribution of the default test question set on all knowledge points; convert the target student user's mastery score on all knowledge points into an attention weight for each knowledge point through an attention mechanism, and then dynamically adjust the question difficulty distribution using the attention weight, increase the question difficulty corresponding to the knowledge point whose attention weight exceeds the weight threshold, and decrease the question difficulty corresponding to the knowledge point whose attention weight is lower than the weight threshold, to obtain the adjusted question difficulty distribution.

[0047] It should be noted that when extracting the default set of test questions to be recommended to the target students from the current question bank, the extraction can be random or selected according to the original recommendation rules of the platform, and there is no specific limitation.

[0048] It should also be noted that the current question bank in the present invention refers to the question bank that can be recommended to students on the online learning platform. The question bank contains a series of test questions, and each test question has a corresponding knowledge point and question difficulty. And all the test questions in the question bank have been created or compiled by the teacher user or administrator in advance to form a test question set. In the traditional online learning platform recommendation mechanism, the test question set pushed to each student user at a certain stage is the same. However, in the present invention, it is necessary to adjust the test question difficulty and knowledge point distribution according to the current personalized status of the student, and then form a new test question set.

[0049] In the embodiment of the present invention, it is assumed that the student feature vector of the target student user is X, and the knowledge point mastery evaluation model ) The student feature vector Mapped to the target students' mastery of each knowledge point After that, i=1,2,…,m, and the attention weight of each knowledge point can be further calculated through the attention mechanism. , to reflect the relative importance of this knowledge point to students. The calculation formula of the attention mechanism is as follows:

[0050]

[0051] Then we can use the above attention weights (i=1,2,...,m) dynamically adjusts the difficulty distribution of the default test question set on all knowledge points. Since the difficulty of the test questions is generally a discrete label, the present invention can preset a single difficulty adjustment amount To check the original difficulty of each test question Adjust the difficulty of the questions on the i-th knowledge point It can be dynamically adjusted as follows: :

[0052]

[0053] in, is the original question difficulty of the i-th knowledge point, is the difficulty of the adjusted question on the i-th knowledge point, is the student’s attention weight on knowledge point i. It has been normalized, so its relative size can reflect the students' mastery of the knowledge point. When it is higher, it shows that the students have a higher degree of mastery of the knowledge point, and the difficulty can be appropriately increased on the basis of the original difficulty to facilitate consolidation and advanced learning. When it is lower, it shows that the students have a weaker grasp of the knowledge point, so the difficulty can be appropriately reduced on the basis of the original difficulty to avoid the test questions being too difficult, so that students can study on simpler test questions and gradually master the knowledge point.

[0054] Above The specific value can be adjusted according to the actual situation. For example, if the highest difficulty score of the question on the knowledge point is set to 5 points and the lowest to 1 point, then It can be set to 1 point, that is, adjust the difficulty level by 1 point each time. It can also be set to an unfixed form. Set to Adjust the difficulty level of each adjustment by Decide.

[0055] Therefore, for the default test set, the difficulty distribution of questions in all knowledge points is , get the adjusted question difficulty for each knowledge point in the complete knowledge point distribution After that, the adjusted question difficulty distribution can be formed. , this distribution will serve as the input of the subsequent large language model.

[0056] S4. The default test questions of the target students, the mastery scores of the target student users on all knowledge points, the knowledge point coverage, and the adjusted question difficulty distribution are combined with the prompt word template to construct a prompt word text. The prompt word text is input into a personalized question setting model obtained by fine-tuning the large language model (LLM) on the question setting task. The personalized question setting model outputs a new test question set and pushes it to the target student users.

[0057] It should be noted that in the process of outputting a new set of test questions by the personalized question-setting model, it is preferred to let the model re-extract questions of different difficulty levels and related knowledge points from the existing question bank to replace the existing questions in the default test question set. Of course, based on the generative output capability of the large language model itself, the large language model can also automatically generate questions that meet the requirements to replace the existing questions in the default test question set, but it is necessary to ensure that the large language model has the ability to generate correct and reasonable questions. The large language model should preferably be further fine-tuned on the task of generating test questions.

[0058] It should be noted that the specific practice of setting the prompt word template above is a common practice in the prompt word (Prompt) project. The prompt word template can be designed according to the task requirements, and a placeholder can be preset in the prompt word template for each personalized input content, so that the program can automatically replace the placeholder and generate the prompt word text input by the large language model (LLM). An exemplary prompt word template is as follows: "Please perform a task of recommending test questions for students. The current default test question is [placeholder 1], the current student's mastery score on all knowledge points is [placeholder 2], the current student's corresponding knowledge point coverage is [placeholder 3], and the adjusted question difficulty distribution for the current student should be [placeholder 4]. Based on the student's default test question, according to the student's mastery score on all knowledge points, knowledge point coverage and adjusted question difficulty distribution, extract questions from the existing question bank to adjust the question composition in the default test question, and obtain new test questions that meet the student's personalized learning needs."

[0059] [Placeholder 1], [Placeholder 2], [Placeholder 3], and [Placeholder 4] in the above prompt word template correspond to the default test questions of the target students, the mastery scores of the target student users on all knowledge points, the coverage of knowledge points, and the adjusted distribution of question difficulty. The replacement of the placeholders in the prompt word template can be automated by a script program to facilitate the automation of the present invention.

[0060] In addition, it should be noted that the above S1~S4 describe the practice of the personalized dynamic question-setting method in actual application, but those skilled in the art should know that the above-mentioned large language model (LLM) needs to be fine-tuned in advance before it can be used as a personalized question-setting model. In an embodiment of the present invention, when fine-tuning the large language model (LLM) on the question-setting task, the model parameters of the large language model can be optimized with the maximization of the sum of knowledge point coverage and knowledge point difficulty matching as the optimization goal to obtain a personalized question-setting model. In the concept of this embodiment, the knowledge point difficulty matching can be set to the mean square error between the adjusted question difficulty distribution of the input model and the question difficulty distribution of the new test question set output by the model.

[0061] The optimization objective can be expressed as follows:

[0062]

[0063]

[0064] Among them, Match means calculating the matching degree between two distributions; T is a learning cycle, and test questions need to be recommended to students at each moment in the cycle. and They represent the corresponding and . and are K and . are the learnable parameters of the large language model, It is the optimized learnable parameters of the large language model.

[0065] Among the above optimization goals, The larger the better, but based on the general loss form of model optimization, the loss value often needs to be as small as possible. Therefore, in actual fine-tuning, the above optimization objective can be replaced by minimizing the following loss function:

[0066]

[0067]

[0068] Therefore, the questions generated by the present invention not only cover the students' weak knowledge points, but also expand their knowledge in combination with future goals, thereby improving the students' learning effects.

[0069] In addition, in the actual application process of the above-mentioned personalized question-setting model, it is necessary to construct an incremental learning data set by collecting feedback data, with maximizing the sum of knowledge point coverage and knowledge point difficulty matching as the optimization goal, and regularly update and optimize the model parameters of the personalized question-setting model through the Bayesian optimization algorithm.

[0070] In an embodiment of the present invention, the learning progress of students can be monitored in real time through the background management system on the online learning platform, including the number of completed questions, learning time, knowledge points mastered, etc., and the learning progress data can be fed back to the dynamic adjustment module. At the same time, students' feedback on the questions can be collected, including the correct answer rate, answer time, subjective satisfaction, etc. These data can be used to construct an incremental learning data set, and the parameters of the personalized question model can be optimized through the Bayesian Optimization algorithm. Conduct incremental training to maximize student learning outcomes According to the optimized model parameters , you can dynamically adjust the question-setting strategy, update the question difficulty and knowledge point distribution, ensure that the question-setting model can adapt to students' learning changes in real time, and improve the accuracy of personalized question-setting.

[0071] Similarly, based on the same inventive concept, Figure 2 As shown, another preferred embodiment of the present invention further provides a personalized dynamic question setting system based on a large language model, which includes:

[0072] Information collection module, used to record multi-dimensional information data of each student user on the online learning platform;

[0073] The target user selection module is used for users to select target student users who need to receive personalized test questions.

[0074] The personalized push module is used to push personalized test questions for the target student users selected by the user according to the multi-dimensional information data recorded in the information collection module and the personalized dynamic question-setting method based on the large language model as described in the above embodiment.

[0075] It should be noted that the above-mentioned information collection module can be built in the background server of the online learning platform to record the historical learning records of student users in real time. The target user selection module can be located on the front-end interface of the online learning platform and provide specified functions through buttons and the like. Of course, the target user can also be set by default in the target user selection module. If the management user does not modify the default setting, the target student user selection is performed according to the default setting. Every student user on the platform can be used as a personalized question object. The push method in the personalized push module can be implemented through a fixed module or display logic in the user interface, or through system messages, pop-up windows, or emails, text messages, and the like.

[0076] It should also be noted that the personalized dynamic question-setting method based on a large language model described in S1 to S4 in the above embodiments can essentially be executed through a computer program or module.

[0077] Therefore, similarly, based on the same inventive concept, another preferred embodiment of the present invention also provides a computer program product corresponding to the personalized dynamic question-setting method based on a large language model provided in the above embodiment, including a computer program / instruction. When the computer program / instruction is executed by a processor, the personalized dynamic question-setting method based on a large language model in the above embodiment can be implemented.

[0078] Similarly, based on the same inventive concept, Figure 3 As shown, another preferred embodiment of the present invention further provides a computer electronic device corresponding to the personalized dynamic question-setting method based on a large language model provided in the above embodiment, which includes a memory and a processor;

[0079] The memory is used to store computer programs;

[0080] The processor is used to implement the personalized dynamic question-setting method based on a large language model in the above embodiment when executing the computer program.

[0081] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0082] Therefore, based on the same inventive concept, another preferred embodiment of the present invention also provides a computer-readable storage medium corresponding to the personalized dynamic question-setting method based on a large language model provided in the above embodiment, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the personalized dynamic question-setting method based on a large language model in the above embodiment can be implemented.

[0083] It is understandable that the above storage medium may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage. The storage medium may also be a U disk, a mobile hard disk, a magnetic disk or an optical disk, etc., which can store program codes.

[0084] It is understandable that the above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0085] It should also be noted that those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the various embodiments provided in this application, the division of steps or modules in the system and method is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules or steps can be combined or integrated together, and a module or step can also be split.

[0086] The above-described embodiments are only some preferred implementations of the present invention, but are not intended to limit the present invention. A person skilled in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A personalized dynamic question-setting method based on a large language model, characterized in that: include: S1. Obtain multi-dimensional information data of target student users, including historical answer records, historical scores of various subjects, and learning direction guidance information, and construct them into structured student feature vectors through feature engineering; S2. Input the student feature vector into the pre-trained knowledge point mastery evaluation model to obtain the mastery score of the target student user on each knowledge point in the complete knowledge point distribution, and then calculate the knowledge point coverage based on the mastery score of the target student user on all knowledge points; S3. Extract the default test question set to be recommended to the target student from the current question bank, and analyze the difficulty distribution of the default test question set on all knowledge points; convert the mastery scores of the target student users on all knowledge points into attention weights for each knowledge point through the attention mechanism, and then dynamically adjust the difficulty distribution of the questions using the attention weights, increase the difficulty of the questions corresponding to the knowledge points whose attention weights exceed the weight threshold, and decrease the difficulty of the questions corresponding to the knowledge points whose attention weights are lower than the weight threshold, to obtain the adjusted difficulty distribution of the questions; S4. The default test questions of the target students, the mastery scores of the target student users on all knowledge points, the knowledge point coverage, and the adjusted question difficulty distribution are combined with the prompt word template to construct a prompt word text. The prompt word text is input into a personalized question setting model obtained by fine-tuning the large language model on the question setting task. The personalized question setting model outputs a new test question set and pushes it to the target student users.

2. The personalized dynamic question-setting method based on a large language model as claimed in claim 1, characterized in that: The learning direction guidance information includes the subjects that need to be studied emphatically and / or the knowledge points that need to be studied emphatically in each subject given to the target student user.

3. The personalized dynamic question-setting method based on a large language model as claimed in claim 1, characterized in that: The knowledge point mastery degree assessment model is obtained by performing supervised training of a multi-layer perceptron, a random forest, a support vector machine or a convolutional neural network on a labeled data set.

4. The personalized dynamic question-setting method based on a large language model as claimed in claim 1, characterized in that: The knowledge point coverage is calculated based on the expected distribution of knowledge point mastery levels predefined by the teacher user and the mastery scores of the target student users on all knowledge points. The calculation method is: for each knowledge point in the complete knowledge point distribution, the expected mastery score corresponding to the knowledge point in the expected distribution of knowledge point mastery levels is compared with the mastery score of the target student user on the knowledge point, and the smaller one is taken as the representation value. Then, the ratio of the sum of the representation values ​​of all knowledge points and the sum of the expected mastery scores of all knowledge points is calculated, and the calculated ratio is used as the current knowledge point coverage of the target student user.

5. The personalized dynamic question-setting method based on a large language model as claimed in claim 1, characterized in that: When the large language model is fine-tuned on the question-setting task, the model parameters of the large language model are optimized with the maximization of the sum of knowledge point coverage and knowledge point difficulty matching as the optimization goal to obtain a personalized question-setting model.

6. The personalized dynamic question-setting method based on a large language model as claimed in claim 5, characterized in that: The knowledge point difficulty matching degree is the mean square error between the adjusted question difficulty distribution of the input model and the question difficulty distribution of the new test question set output by the model.

7. The personalized dynamic question-setting method based on a large language model as claimed in claim 5, characterized in that: In actual application, the personalized question-setting model needs to construct an incremental learning data set by collecting feedback data, take maximizing the sum of knowledge point coverage and knowledge point difficulty matching as the optimization goal, and regularly update and optimize the model parameters of the personalized question-setting model through the Bayesian optimization algorithm.

8. A personalized dynamic question setting system based on a large language model, characterized in that: include: Information collection module, used to record multi-dimensional information data of each student user on the online learning platform; The target user selection module is used for users to select target student users who need to receive personalized test questions. The personalized push module is used to push personalized test questions for target student users selected by the user according to the personalized dynamic question-setting method based on a large language model as described in any one of claims 1 to 7 based on the multi-dimensional information data recorded in the information collection module.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the personalized dynamic question-setting method based on a large language model as described in any one of claims 1 to 7 is implemented.

10. A computer electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is used to implement the personalized dynamic question-setting method based on a large language model as described in any one of claims 1 to 7 when executing the computer program.

Images