Topic recommendation method, apparatus and device, readable storage medium and product

Abstract

The invention provides a question recommendation method and device, equipment, a readable storage medium and a product, and the method comprises the steps: screening out a plurality of questions matched with a weak knowledge point or a specific error cause from a recommendation question pool through combining the weak knowledge point and the specific error cause reflected by a wrong question which is wrongly answered by a user; actual knowledge mastering weak links of the user are accurately positioned, multi-dimensional scoring is carried out on each matched question according to question attributes and user answering conditions, and then exercises highly matched with the actual error condition of the user are selected according to a final multi-dimensional scoring result to be recommended. Therefore, the finally recommended exercises meet the current learning condition and the exercise demand of the user, the question recommendation accuracy is improved, the purpose of targeted training based on wrong questions is achieved, and the learning efficiency and the learning effect are improved.

Description

Technical Field

[0001] The present application relates to the field of educational technology, and in particular to a topic recommendation method, device, equipment, readable storage medium and product. Background Art

[0002] In the current field of educational technology, question practice-related products are usually equipped with a wrong question book function, which collects and analyzes questions that users have answered incorrectly in order to improve the user's answer accuracy and mastery of knowledge points.

[0003] When recommending exercises, the existing wrong answer book function usually collects the knowledge points associated with the wrong questions answered by the user, and then recommends exercises that are strongly related to the collected knowledge points for the user to practice repeatedly to improve learning effects. However, this recommendation method relies on the classification of the knowledge points involved in the wrong questions, and the recommended results are too general and lack sufficient accuracy, and cannot accurately locate the user's actual weak points and provide targeted exercises. Summary of the invention

[0004] In view of this, in order to solve the above technical problems, the present application provides a topic recommendation method, device, equipment, readable storage medium and product.

[0005] Specifically, the present application is implemented through the following technical solutions:

[0006] According to a first aspect of an embodiment of the present application, a topic recommendation method is provided, the method comprising:

[0007] In response to a question recommendation request initiated by a user based on a wrongly answered question, a plurality of matching questions matching at least one of a weak knowledge point and a wrong cause are obtained from a recommended question pool; wherein the weak knowledge point at least includes a knowledge point that the user has a weak grasp of, determined from the knowledge point to which the wrong question belongs and its preceding knowledge; the wrong cause indicates the reason why the wrong question was answered incorrectly by the user;

[0008] For each matching question, calculate the score of the matching question in each scoring dimension to obtain the total score of the matching question in all scoring dimensions; wherein the scoring dimensions are obtained based on the attributes of the question and the user's answer situation;

[0009] N questions are selected as recommended practice questions based on the total scores of multiple matching questions.

[0010] Optionally, the scoring dimension includes at least a knowledge point category and a question error reason classification; and the step of calculating the score of the matching question on each scoring dimension includes:

[0011] When the knowledge point to which the matching question belongs belongs to the weak knowledge point, based on the importance of the knowledge point to which the matching question belongs, determining the score of the matching question in the scoring dimension of the knowledge point category;

[0012] The score of the matching question on the scoring dimension of question error cause classification is determined based on the error cause categories included in the matching question and the matching with the error causes of the wrong question answered incorrectly by the user.

[0013] Optionally, the scoring dimension also includes at least: at least one of the difficulty level of the question, the number of wrong answers by the user, and the type of wrong answer by the user; wherein the type of wrong answer by the user includes at least the wrong answer by the user and the user not knowing how to answer.

[0014] Optionally, the reasons why the wrong question is answered incorrectly by the user include multiple levels of reasons that are refined layer by layer;

[0015] Determining the score of the matching question on the scoring dimension of the classification of the error cause of the matching question based on the error cause category included in the matching question and the matching situation with the error cause of the wrong answer of the wrong question by the user includes:

[0016] According to the hierarchical structure of error causes, the error cause categories included in the matching questions and the error causes of the wrong questions answered by the user are matched level by level, and the level score of the level is determined based on whether the error causes of the two levels are the same;

[0017] According to the pre-set weights of each level, the level scores of all levels are weighted and calculated to obtain the scores of the matching questions on the scoring dimension of question error classification.

[0018] Optionally, the method further comprises the step of determining the cause of the error:

[0019] The wrong questions and the user's answers to the wrong questions are matched with the error cause data set to determine the error cause of the wrong answer to the wrong question; wherein the error cause data set includes a data set with each exercise question and the wrong answer to the exercise question as the division dimensions, and annotating the error cause of the user answering with the wrong answer for each wrong answer.

[0020] Optionally, the method further comprises the step of determining weak knowledge points:

[0021] Retrieving the preceding knowledge points of the knowledge point to which the wrong question belongs from a pre-constructed knowledge graph; the knowledge graph refers to a network graph composed of knowledge points of various subjects and used to represent the dependency relationship between subject knowledge points;

[0022] Based on the knowledge point to which the wrong question belongs and the user's answer situation under the previous knowledge point, the knowledge point that the user has weak grasp of is determined as the weak knowledge point.

[0023] According to a second aspect of an embodiment of the present application, a topic recommendation device is provided, the device comprising:

[0024] The question matching module is used to respond to a question recommendation request initiated by a user based on a wrongly answered question, and obtain a plurality of matching questions matching at least one of a weak knowledge point and a wrong cause from a recommended question pool; wherein the weak knowledge point at least includes a knowledge point that the user has a weak grasp of, which is determined from the knowledge point to which the wrong question belongs and its preceding knowledge; the wrong cause indicates the reason why the wrong question is answered incorrectly by the user;

[0025] A matching question evaluation module is used to calculate the score of each matching question in each scoring dimension, and obtain the total score of the matching question in all scoring dimensions; wherein the scoring dimensions are obtained based on the attributes of the question and the user's answer situation;

[0026] The topic recommendation module is used to select N topics as recommended practice topics based on the total scores of multiple matching topics.

[0027] Optionally, the scoring dimension includes at least knowledge point category and question error cause classification; the matching question evaluation module is specifically used for:

[0028] When the knowledge point to which the matching question belongs belongs to the weak knowledge point, based on the importance of the knowledge point to which the matching question belongs, determining the score of the matching question in the scoring dimension of the knowledge point category;

[0029] The score of the matching question on the scoring dimension of question error cause classification is determined based on the error cause categories included in the matching question and the matching with the error causes of the wrong question answered incorrectly by the user.

[0030] Optionally, the scoring dimension also includes at least: at least one of the difficulty level of the question, the number of wrong answers by the user, and the type of wrong answer by the user; wherein the type of wrong answer by the user includes at least the wrong answer by the user and the user not knowing how to answer.

[0031] Optionally, the reasons why the wrong question is answered incorrectly by the user include multiple levels of reasons that are refined layer by layer; and the matching question evaluation module is specifically used to:

[0032] According to the hierarchical structure of error causes, the error cause categories included in the matching questions and the error causes of the wrong questions answered by the user are matched level by level, and the level score of the level is determined based on whether the error causes of the two levels are the same;

[0033] According to the pre-set weights of each level, the level scores of all levels are weighted and calculated to obtain the scores of the matching questions on the scoring dimension of question error classification.

[0034] Optionally, the device further comprises a step of determining the cause of the error:

[0035] The wrong questions and the user's answers to the wrong questions are matched with the error cause data set to determine the error cause of the wrong answer to the wrong question; wherein the error cause data set includes a data set with each exercise question and the wrong answer to the exercise question as the division dimensions, and annotating the error cause of the user answering with the wrong answer for each wrong answer.

[0036] Optionally, the device further comprises a step of determining weak knowledge points:

[0037] Retrieving the preceding knowledge points of the knowledge point to which the wrong question belongs from a pre-constructed knowledge graph; the knowledge graph refers to a network graph composed of knowledge points of various subjects and used to represent the dependency relationship between subject knowledge points;

[0038] Based on the knowledge point to which the wrong question belongs and the user's answer situation under the previous knowledge point, the knowledge point that the user has weak grasp of is determined as the weak knowledge point.

[0039] According to a third aspect of an embodiment of the present application, an electronic device is provided, comprising: a memory and a processor; the memory is used to store a computer program; the processor is used to execute the above-mentioned question recommendation method by calling the computer program.

[0040] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned question recommendation method is implemented.

[0041] The technical solution provided by the embodiments of the present application may have the following beneficial effects:

[0042] In the technical solution provided by the present application, by combining the weak knowledge points and specific reasons for the wrong answers reflected by the user, multiple questions matching the weak knowledge points or reasons for the wrong answers are screened out from the recommended question pool, accurately locating the weak links in the user's actual knowledge mastery, and performing multi-dimensional scoring on each matching question based on the question attributes and the user's answer situation, and then selecting exercises that are highly matched with the user's actual error situation for recommendation based on the final multi-dimensional scoring results, so that the final recommended exercises are in line with the user's current learning situation and practice needs, thereby improving the accuracy of question recommendations, achieving the purpose of targeted training based on wrong questions, and improving learning efficiency and learning effects.

[0043] It should be understood that the above general description and the following detailed description are only exemplary and explanatory and cannot limit the present application. In addition, any embodiment in the present application does not need to achieve all the above effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0045] Figure 1A It is a flowchart of a topic recommendation method shown in an exemplary embodiment of the present application;

[0046] Figure 1B This is a schematic diagram of processing steps for an exemplary embodiment of the present application, taking a math problem A in which a user answers incorrectly as an example;

[0047] Figure 2A It is a flowchart of determining the score of a matching question based on multi-level error causes in the scoring dimension of question error cause classification, shown in an exemplary embodiment of the present application;

[0048] Figure 2B This is an example diagram of a process of determining the score of a matching question in the scoring dimension of question error cause classification based on multi-level error cause matching, shown in an exemplary embodiment of the present application;

[0049] Figure 3 is an example diagram of the content of an error cause data set shown in an exemplary embodiment of the present application;

[0050] Figure 4 It is a partial schematic diagram of a knowledge graph using the knowledge point of solving a quadratic equation as an example, shown in an exemplary embodiment of the present application;

[0051] Figure 5A It is a schematic diagram of the overall process of a topic recommendation method shown in an exemplary embodiment of the present application;

[0052] Figure 5B is a schematic diagram of multi-service module interaction involved in a topic recommendation method shown in an exemplary embodiment of the present application;

[0053] Figure 5C It is an example of topic attribute information in a recommended topic pool shown in an exemplary embodiment of the present application;

[0054] Figure 5D It is a schematic diagram of a multi-dimensional scoring strategy shown in an exemplary embodiment of the present application;

[0055] Figure 6 is a structural schematic diagram of a topic recommendation device shown in an exemplary embodiment of the present application;

[0056] Figure 7 It is a hardware schematic diagram of an electronic device shown in an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0057] Here, exemplary embodiments will be described in detail, and examples thereof are shown in the accompanying drawings. When the following description relates to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are only examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims. It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other.

[0058] In the field of educational technology, question practice related products are usually equipped with a wrong question book function, which is used to collect and analyze questions that users answered incorrectly, in order to improve the user's answer accuracy and the degree of mastery of knowledge points. The working principle of the existing wrong question book function is to collect the knowledge points associated with the questions that users answered incorrectly, and recommend exercises that are strongly related to the collected knowledge points for users to practice repeatedly, so as to help users master the relevant knowledge points and improve learning effects.

[0059] However, the existing practice question recommendation method that relies on knowledge point classification has significant limitations. The recommended results are often too general and lack sufficient accuracy, and cannot accurately locate and conduct targeted exercises for users' actual weak links. For example, in questions related to calculating RMB, users may make mistakes due to various reasons, such as lack of proficiency in decimal system, unfamiliarity with RMB calculation units, or bad question-solving habits (such as incorrectly filling in units). Based on this, simply relying on knowledge points to recommend questions that are strongly related to knowledge points cannot guarantee the relevance and pertinence of the recommended exercises, thus affecting the improvement of learning effects.

[0060] In view of this, the present application proposes a more accurate and personalized question recommendation method to better adapt to and meet the user's actual question answering situation. In the question recommendation process, this method takes into account the error reasons why the user answered the wrong question incorrectly, and the user's weak knowledge points reflected by the wrong answer to the wrong question. Questions are matched from the question pool based on the error reasons and the user's weak knowledge points, and each matched question is further evaluated and scored using pre-set multiple evaluation dimensions to measure the degree of suitability of the question with the user's current practice needs, thereby screening out questions suitable for the user to conduct targeted practice.

[0061] The method can be applied to a terminal or server related to question practice, or to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. The terminal may include but is not limited to various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices may include smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. The portable wearable devices may include smart watches, smart bracelets, head-mounted devices, etc. The server may include an independent server or a server cluster consisting of multiple servers.

[0062] See also Figure 1A The method step flow chart shown exemplarily shows that the topic recommendation method provided by the present application can be implemented by at least the following steps:

[0063] S101, in response to a question recommendation request initiated by a user based on a wrongly answered question, a plurality of matching questions matching at least one of a weak knowledge point and a wrong cause are obtained from a recommended question pool; wherein the weak knowledge point at least includes a knowledge point determined from the knowledge point to which the wrong question belongs and its preceding knowledge and which the user has a weak grasp of; and the wrong cause indicates the reason why the wrong question was answered incorrectly by the user;

[0064] Wrong questions include questions that users failed to answer correctly when answering questions. Based on this, the recommendation request indicates that the user has initiated a request to recommend similar questions based on the questions that he answered incorrectly. By using the user's learning problems exposed by the wrong questions, the user can select appropriate questions from the recommended question pool, so that the user can conduct targeted question training to avoid making the same mistakes again, thereby optimizing the learning effect.

[0065] Weak knowledge points represent knowledge content that the user has not mastered well in the learning process, and the user's historical question answering situation under the corresponding knowledge point, such as the answer accuracy rate, can be used to quantitatively evaluate the user's mastery of the corresponding knowledge point. In this embodiment, the weak knowledge point can at least include: the knowledge point directly involved in the wrong question, that is, the knowledge point to which the wrong question belongs, and the pre-order knowledge point that should be mastered before the knowledge point to which the wrong question belongs, and the knowledge point that the user has a poor mastery of; wherein, the pre-order knowledge point represents the basic knowledge point that needs to be mastered in advance before learning a certain specific knowledge point. For example, in the study of mathematics, to learn the knowledge point of quadratic equations, it is necessary to first master the relevant knowledge of linear equations. The knowledge of linear equations is the pre-order knowledge point of quadratic equations, which can include how to solve linear equations such as moving terms, combining like terms, and coefficients equal to 1, and understanding the properties of equations.

[0066] The error reason indicates the reason why the user made an error in answering the question, which may include but is not limited to misunderstanding of knowledge points, calculation errors, carelessness, confusion of concepts, poor problem-solving habits such as forgetting to bring units, etc. For example, in a math multiple-choice question such as "5*4=", and the answer options "①9; ②20; ③16" are given, if the user selects ① when answering, then the error reason for the user's incorrect answer to the question is that the punctuation mark is misrecognized, and if the user selects ③ when answering, then the error reason for the user's incorrect answer to the question is that the multiplication formula is remembered incorrectly.

[0067] The recommended question pool is a pre-prepared database or collection containing a large number of questions. The questions in the recommended question pool may include questions that have been recommended to users for answering. The questions in the recommended question pool cover different knowledge points, and are used to select suitable questions from them for users to practice in a targeted manner according to the user's learning situation. Each question in the recommended question pool is provided with a question attribute, and the question attribute may include the knowledge point to which the question belongs and the reason for the question under different preset answer options. Therefore, when a question recommendation request based on a wrong question is received, the reason for the wrong question and the weak knowledge point related to the wrong question can be obtained first, and then based on the reason for the wrong question and the weak knowledge point related to the wrong question, a question that meets at least one of the following conditions is obtained from the recommended question pool as a matching question, and the condition includes: the knowledge point to which the question from the recommended question pool belongs belongs to a weak knowledge point; or, the reason for the wrong question from the recommended question pool under different preset answer options has a reason that matches the reason for the wrong question.

[0068] like Figure 1B The processing step flow chart shown as an example takes math question A that the user answered incorrectly as an example, assuming that the cause of the error is "the user forgot to bring the unit when answering", the knowledge point to which the wrong question belongs is "unit conversion between yuan, jiao and fen", the predecessor knowledge point of the knowledge point to which the wrong question belongs is "the base relationship between yuan, jiao and fen", and the knowledge point to which the wrong question belongs is a knowledge point that the user has a weak grasp of (i.e., a weak knowledge point), then when matching questions from the recommended question pool based on the cause of the error and the weak knowledge point, if the cause of the error for question 1 in the recommended question pool includes that the user forgot to bring the unit when answering, or the knowledge point to which question 2 belongs is "unit conversion between yuan, jiao and fen" which belongs to a weak knowledge point, then both question 1 and question 2 can be selected as matching questions.

[0069] S102, for each matching question, calculating the score of the matching question in each scoring dimension, and obtaining the total score of the matching question in all scoring dimensions; wherein the scoring dimensions are obtained based on the attributes of the question and the user's answer situation;

[0070] The scoring dimension is used to evaluate the adaptability between the matching questions and the question recommendation requests initiated by the user based on the wrong questions. The scoring dimension can be divided according to the question attributes of the questions in the recommended question pool, such as the difficulty of the questions, the scope of knowledge points, and the user's answer to the question, such as the number of wrong answers by the user, the type of wrong answers by the user, etc.

[0071] Scoring matching questions based on multiple scoring dimensions can more accurately capture the correlation between the matching questions and wrong questions. In this embodiment, since the question recommendation request is initiated based on the wrong questions, it is used to recommend practice questions related to the wrong questions to the user to train the user to overcome the learning problems exposed by the wrong questions, that is, the question recommendation is to screen and evaluate the weak knowledge points and causes of the wrong questions. Therefore, the scoring dimensions can at least include two dimensions: knowledge point category and question cause classification.

[0072] Among them, the knowledge point category is used to evaluate the importance of the knowledge point of the matching question. For the scoring dimension of the knowledge point category, when calculating the score of the matching question in each scoring dimension, when the knowledge point to which the matching question belongs belongs to the weak knowledge point, based on the importance of the knowledge point to which the matching question belongs, the score of the matching question in the scoring dimension of the knowledge point category is determined. Among them, the importance of the knowledge point to which the matching question belongs is used to indicate the importance of the knowledge point to the user's learning. Different importance levels can be expressed by defining different knowledge point types such as basic knowledge points, extended knowledge points, review knowledge points, and test preparation knowledge points (the importance of the four types increases in sequence), and scoring strategies that conform to their importance levels are set for knowledge point types of different importance.

[0073] It is understandable that, since the question recommendation request is used to recommend practice questions related to wrong questions for users, when the importance of knowledge points is used to evaluate and score the matching questions, when the knowledge point to which the matching question belongs belongs to the weak knowledge point related to the wrong question, it means that the matching question is directly related to the user's weak link. In this case, the importance of the knowledge point to which the matching question belongs is used to determine the score on this evaluation dimension, and the score can reflect the closeness between the matching question and the user's current learning needs and the potential contribution of the question to the user's mastery of the knowledge point; when the knowledge point to which the matching question belongs does not belong to the weak knowledge point related to the wrong question, that is, the knowledge point to which the matching question belongs is completely different from the weak knowledge point determined based on the wrong question, from the perspective that the question recommendation is used to make up for the user's deficiencies in the weak knowledge points exposed by the wrong question and to conduct targeted training on the weak knowledge points, the knowledge point involved in the matching question is irrelevant to the weak knowledge point, which means that the matching question has a low value on this scoring dimension and cannot play a direct role in solving the user's weak problem. Therefore, the score of the matching question on the scoring dimension of the knowledge point category can be directly set to 0, or its score can be set lower than any score when the knowledge point to which the matching question belongs belongs to the weak knowledge point related to the wrong question.

[0074] The classification of the causes of wrong answers to questions is used to evaluate whether the matching questions include the causes of wrong answers to wrong questions, so as to screen out matching questions that can be designed for the reasons why users answer wrong questions incorrectly, thereby providing users with more accurate practice questions. For the scoring dimension of classification of causes of wrong answers to questions, when calculating the score of the matching question on each scoring dimension, the score of the matching question on the scoring dimension of classification of causes of wrong answers to questions is determined based on the category of causes of wrong answers included in the matching question and the matching situation with the causes of wrong answers to the wrong questions by the user. Regarding the matching situation, it can be divided into complete match and complete mismatch, and different scoring strategies are set for these two situations. Alternatively, in order to achieve more refined dimensional scoring, the scoring strategy can be set according to the three matching situations of complete match, partial match, and complete mismatch.

[0075] For example, if the cause of a wrong question is "confusion about the concept of domain of mathematical functions", then for the matching questions, it is necessary to check whether the matching questions have answer options set for the cause of "confusion about the concept of domain of mathematical functions". By identifying whether the matching questions contain the cause of the wrong question, different scores are given for different matching situations.

[0076] This embodiment predefines scoring strategies with different values ​​under each scoring dimension. Each question in the recommended question pool can set question attribute information as multiple attribute types according to the predefined scoring dimensions, so that for the matching questions selected from the recommended question pool based on the causes of wrong questions, that is, weak knowledge points, the score of the question under each scoring dimension can be determined according to its value under each scoring dimension.

[0077] The total score of the matching question in all scoring dimensions is used to indicate the degree of adaptation of the matching question to the user's current practice needs represented by the question recommendation request initiated by the user based on the wrong question. The scoring strategy with different values ​​under each scoring dimension can be defined by an evaluation method in which the total score is proportional to the degree of adaptation, or an evaluation method in which the total score is proportional to the degree of adaptation.

[0078] When obtaining the total score of the matching title in all scoring dimensions, the scores of the matching title in all scoring dimensions can be directly added together, and the added result can be used as the total score. Alternatively, based on the different importance of different scoring dimensions to the recommendation results, in order to more accurately measure the quality of the matching title, a weighted weight can be set for each scoring dimension to obtain the weighted total score of the matching title in all scoring dimensions; that is, after determining the score of the matching title in each scoring dimension, the score in each scoring dimension can be multiplied by the weighted weight of the scoring dimension to obtain the weighted result in each scoring dimension, and the weighted results of all scoring dimensions can be summed to obtain the total score of the matching title in all scoring dimensions.

[0079] like Figure 1B The step flow chart exemplarily shows that, for matching topics 1-matching topics m obtained from the recommended topic pool, the total score of each matching topic in all scoring dimensions is calculated respectively, so as to recommend suitable topics for users to practice based on the total score.

[0080] S103: Select N questions as recommended practice questions based on the total scores of the multiple matching questions.

[0081] The N questions selected in this step as recommended practice questions indicate that the N questions are highly compatible with the user's current practice needs as indicated by the question recommendation request initiated by the user based on the wrong questions. The N questions meet the user's current practice needs and can provide targeted training for the learning problems exposed by the wrong questions answered incorrectly by the user.

[0082] Based on this, when the total score of the matching questions indicates that the matching questions are proportional to the degree of adaptation of the user's current practice needs represented by the question recommendation request, that is, the matching questions with higher total scores are more in line with the practice needs, the top N questions with the highest total scores can be selected as the recommended practice questions. Similarly, when the total score is inversely proportional to the degree of adaptation of the user's current practice needs represented by the question recommendation request, that is, the matching questions with lower total scores are more in line with the practice needs, the top N questions with the lowest total scores can be selected as the recommended practice questions.

[0083] In the disclosed embodiment, by combining the weak knowledge points and specific error causes reflected by the wrong questions answered by the user, multiple questions matching the weak knowledge points or error causes are screened out from the recommended question pool, accurately locating the weak links in the user's actual knowledge mastery, and performing multi-dimensional scoring on each matching question according to the question attributes and the user's answer situation, and then selecting practice questions that are highly matched with the user's actual error situation for recommendation based on the final multi-dimensional scoring results, so that the final recommended practice questions meet the user's current learning situation and practice needs, improve the accuracy of question recommendations, and achieve the purpose of conducting targeted training based on wrong questions, so as to help users correct the learning problems exposed by wrong questions, deepen the user's understanding and mastery of the weak knowledge points related to the wrong questions, and improve learning efficiency and learning effects.

[0084] In some embodiments, in order to better evaluate the degree of adaptation of the matching questions to the current learning needs of the user represented by the recommendation request initiated based on the wrong questions, when evaluating the matching questions from the perspective of whether the matching questions include the causes of the wrong questions, that is, the scoring dimension of the classification of the causes of the wrong questions, this implementation introduces multiple layers of causes of error. The setting of multiple layers of causes of error can more carefully reflect the various causes of errors that occur in the process of answering, so as to more accurately understand the weak links of the user. Through the layer-by-layer detailed analysis of the causes of error, it is possible to more deeply explore the reasons behind the user's errors, such as calculation errors, concept confusion, understanding deviations, etc., and further subdivide these causes into more specific subcategories, so as to more accurately match the actual needs of the user when recommending questions, and provide more targeted exercises on wrong questions.

[0085] Based on this, the reasons why the user answered the wrong question incorrectly may include multiple levels of reasons that are gradually refined. Similarly, each question in the recommended question pool may be marked with the multiple levels of reasons involved. For the scoring dimension of the classification of the reasons for the questions described in the above embodiment, see Figure 2A The method step flow chart shown as an example, when determining the score of the matching question on the scoring dimension of the error cause classification according to the error cause category included in the matching question and the matching situation with the error cause of the wrong answer given by the user to the wrong question, can be implemented by the following steps:

[0086] S201, according to the hierarchical structure of error causes, matching the error cause categories included in the matching questions with the error causes of the wrong questions answered by the user level by level, and determining the level score of the level based on whether the error causes of the two levels are the same;

[0087] The hierarchical structure of error causes is a data structure that classifies error causes according to their internal logic and relationships. As the levels go deeper, the detailed classification of error causes is realized layer by layer. It is understandable that in practical applications, the hierarchical structure of error causes and scoring rules can be customized and dynamically adjusted according to specific learning fields, user groups, and recommendation needs.

[0088] This step matches the error cause categories contained in the matching questions with the error causes when the user answers incorrectly, layer by layer, that is, starting from the highest level (such as the first level error cause), the error causes of the matching questions and the user's wrong questions are compared one by one to see if they are the same or highly similar. If the error cause of the matching question at a certain level is the same as the error cause of the user's wrong question, it is considered that the error cause at that level is matched successfully, and the level score of the successful error cause matching is assigned to that level. On the contrary, if the error cause of the matching question at that level is different from the error cause of the user's wrong question, it is considered that the error cause at that level is matched unsuccessfully, and the level score of the failed error cause matching is assigned to that level. Among them, if either the matching question or the wrong question does not include the error cause of the current error cause level, it can be considered that the error cause at the current level is matched unsuccessfully, and the level score of the failed error cause matching is assigned to that level. For each error cause level, different scoring strategies are set for error cause matching failure / error cause matching failure; the level scoring strategies between different error cause levels can be set independently, that is, different scoring strategies can be adopted for different error cause levels. For example, for the first-level error cause, the score is set to 0.4 for the same error cause and 0.1 for different error causes. For the second-level error cause, the score is set to 0.4 for the same error cause and 0.25 for different error causes.

[0089] like Figure 2B An example diagram shows a process of determining the score of a matching question in the scoring dimension of question error cause classification based on multi-level error cause matching. Suppose a user makes an error in answering a math question B, and the reason for the wrong answer belongs to the first-level error cause "insufficient mastery of knowledge points" shown in the figure. After screening out multiple matching questions from the recommended question pool based on weak knowledge points and error causes, assuming that the first-level error causes of matching question 1 include "question review problems" and "insufficient mastery of knowledge points", and the second-level error causes include "question review problems: inability to understand the question, misread the question" and "insufficient mastery of knowledge points: confusion of concepts", then the hierarchical score of the matching question on the first-level error cause and the hierarchical score on the second-level error cause are calculated respectively.

[0090] Based on the fact that both the matching question 1 and the incorrectly answered question B contain the first-level error cause "inadequate mastery of knowledge points", the first-level error cause matching is successfully matched to determine the matching question 1's score F1 at the first-level error cause level; similarly, based on the fact that both the matching question 1 and the incorrectly answered question B contain the second-level error cause "inadequate mastery of knowledge points: confusion of concepts", the second-level error cause matching is successfully matched to determine the matching question 1's score F2 at the second-level error cause level.

[0091] S202, according to the pre-set weights of each level, weighted calculation is performed on the level scores of all levels to obtain the score of the matching question on the scoring dimension of question error cause classification.

[0092] A level weight is pre-set for each error cause level to reflect the relative importance of the error cause level in the scoring dimension of question error cause classification. It can be understood that the level weights of different error cause levels can be the same, or can be set separately for each error cause level.

[0093] After obtaining the level score of the matching question at each error cause level, the level score of each error cause level can be multiplied by its corresponding level weight to obtain the weighted score; then, the weighted scores of all levels are added together to obtain the final score of the matching question on the scoring dimension of question error cause classification.

[0094] For example, still Figure 2B Taking the matching question 1 shown as an example, after obtaining the hierarchical scores F1 and F2 of the matching question 1 on the first-level error causes and the second-level error causes, assuming that the hierarchical weight of the first-level error cause is 0.3 and the hierarchical weight of the second-level error cause is 0.4, the score F of the matching question 1 on the scoring dimension of question error cause classification can be expressed as F=0.3*F1+0.4*F2.

[0095] In the embodiment of the present disclosure, by introducing a multi-layer error cause structure, the error causes are subdivided into multiple levels, each level represents a different type of error cause, such as calculation error, concept confusion, understanding deviation, etc., and these causes are further subdivided into more specific subcategories to more carefully reflect the various error causes that occur in the user's answering process; when recommending questions, the error cause categories included in the matching questions are matched with the error causes of the user's incorrect answers level by level, and the level scores of each error cause level are determined based on whether the error causes of the two at the same level are the same, which can more accurately understand the user's weak links, and the level scores of each error cause level are weighted to obtain the final score of the matching question in the scoring dimension of question error cause classification, considering the contribution of different error cause levels to the evaluation results, so that the final score of the matching question in the scoring dimension of question error cause classification can more accurately capture the correlation between the matching question and the wrong question answered by the user, thereby improving the accuracy and pertinence of the recommended questions and providing users with a more personalized learning experience.

[0096] In some embodiments, in order to more comprehensively evaluate the degree of adaptation of the matching questions to the user's current learning needs represented by the recommendation request initiated based on the wrong questions, the scoring dimensions may also include at least one of: the difficulty level of the question, the number of wrong answers by the user, and the type of wrong answers by the user.

[0097] Among them, the difficulty level of the question indicates the difficulty of the question based on a comprehensive assessment of factors such as the knowledge content of the question, the complexity of the problem-solving steps, and the depth of thinking required. The difficulty level of the question is proportional to the complexity of the question. The higher the difficulty level of the question, the higher the knowledge and ability requirements for the user.

[0098] The number of wrong answers by a user indicates the number of wrong answers a user has given to a question in the past, which can reflect the user's mastery of a specific knowledge point and the difficulty of learning it. By considering the number of wrong answers by a user, we can identify the knowledge points or question types that the user frequently makes mistakes in, and thus recommend relevant questions for intensive practice.

[0099] The user's wrong answer type indicates different types of errors that occurred during the user's answering process. In this embodiment, the user's wrong answer type may include at least two types: user's wrong answer and user's inability to answer. Among them, the user's wrong answer indicates that the user has tried to answer but answered incorrectly, reflecting the user's incomplete understanding of the knowledge points, insufficient problem-solving skills and other problems. The user's inability to answer indicates that the user has not tried to answer at all or clearly stated that he cannot do the question, reflecting that the user is completely unfamiliar with the relevant knowledge points or lacks the necessary problem-solving ability. Distinguishing the types of user's wrong answers helps to more accurately locate the user's learning difficulties and needs.

[0100] In this embodiment, by comprehensively considering scoring dimensions such as the difficulty level of the question, the number of wrong answers by the user, and the type of wrong answers by the user, the degree of adaptation of the matching questions to the user's current learning needs can be more comprehensively evaluated, thereby providing the user with more personalized and effective learning recommendations.

[0101] In some embodiments, the reasons for the wrong answers given by the user to the wrong questions described in the above embodiments can be determined by matching based on the error cause data set. The error cause data set includes a data set that uses each exercise question and the wrong answer to the exercise question as a division dimension, and annotates the reasons for the user's wrong answers for each wrong answer. The annotated reasons can cover various types such as calculation errors, concept confusion, understanding deviations, memory ambiguity, carelessness, etc., depending on the detailed division when the error cause data set is constructed. The error cause data set is dynamically modified with the creation or editing of the question.

[0102] The error cause data set may include all questions that users can practice or answer, and the error cause information of the practice questions is generated together after the practice questions are created or edited. In order to further improve the retrieval efficiency and accuracy of the error cause data set, a unique question identification ID (Identification) can be configured for each practice question, so that the error cause information related to a specific question can be quickly found through ID matching.

[0103] For example, participate Figure 3 A schematic diagram of part of the contents of an error cause data set is shown as an example. When creating the exercise question - multiple-choice question T1 "Question: 5*4=(). Answer: ①9; ②20; ③15; ④16", its identification ID is 0902, and the error cause is set for each preset answer option. For example, the error cause set for preset answer option ① includes: confusion between addition and multiplication symbols, and misreading of operation symbols. The error causes set for preset answer options ③④ include: lack of proficiency in multiplication formulas. Multiple-choice question T1 and the error causes set for its preset answer options constitute a piece of question data in the error cause data set. Based on a similar principle, for other produced or created exercise questions, the error causes when the user answers the exercise question with a preset wrong answer are marked to form a complete error cause data set.

[0104] Based on this, when it is necessary to determine the reason why a wrong question was answered incorrectly by a user, the wrong question and the user's answer to the wrong question can be matched with the error cause data set to determine the reason why the wrong question was answered incorrectly. That is, the error cause data item about the wrong question can be quickly located from the error cause data set according to the title identification ID of the wrong question, and then the error cause marked for the user's answer to the wrong question can be found from the error cause data item according to the user's answer to the wrong question, and it can be used as the reason why the wrong question was answered incorrectly by the user.

[0105] For example, for the exercise question "5*4=()" with question identification ID 0902, the available answer options are "①9; ②20; ③15; ④16", and the user's answer is ①, then based on the question identification ID "0902" + the user's answer "①", the reasons for the user's incorrect answer to this question can be quickly matched from the error cause data set to determine that the reasons include: confusion between addition and multiplication symbols, and misreading of operation symbols.

[0106] In the disclosed embodiment, the error reason for the wrong answer of the wrong question by the user is determined by matching the error reason data set, thereby improving the accuracy of error reason identification. At the same time, each exercise question is configured with a unique question identification ID, which can quickly locate the error reason information related to the specific question, thereby improving the retrieval efficiency.

[0107] In some embodiments, for the weak knowledge points described in the aforementioned embodiments, it is known that the weak knowledge points at least include knowledge points that are determined from the knowledge points to which the wrong questions belong and their preceding knowledge, and that the user has weak grasp of. Therefore, when responding to a question recommendation request initiated by a user based on a wrong question, the knowledge points to which the wrong questions belong and the preceding knowledge points required to master the knowledge points to which the wrong questions belong are first determined. Based on the principle of gradual progress in knowledge learning, a detailed and comprehensive knowledge graph can be pre-constructed. The knowledge graph is composed of knowledge points from various disciplines. Each knowledge point is regarded as a node, and the connection between nodes represents the dependency or sequential relationship between knowledge points. The graph not only contains the knowledge points themselves, but is also used to represent the dependency relationship between subject knowledge points. The sequence and logical relationship between knowledge points are clearly depicted through the connection of nodes, forming a complete knowledge system framework.

[0108] For example, see Figure 4 A partial schematic diagram of a knowledge graph showing a knowledge point of solving a quadratic equation as an example. The solutions to quadratic equations include matching method, formula method, and factoring method. Before learning the solution to a quadratic equation, you need to first understand the definition of a quadratic equation and the standard form of a quadratic equation. Before learning a quadratic equation, you need to first understand the basic knowledge of algebra, including knowledge points such as variables and constants, algebraic expressions, equations and inequalities, and factoring. Among them, algebraic expressions can be subdivided into sub-knowledge points: terms and coefficients, simplification of algebraic expressions, equations and inequalities can be divided into sub-knowledge points: linear equations, linear inequalities, and factoring can be divided into sub-knowledge points: common factor extraction, square difference formula, and perfect square formula. For the knowledge point of solving a quadratic equation, the definition of a quadratic equation, the standard form of a quadratic equation, and factoring can all be used as its predecessor knowledge points. As shown in the knowledge graph, the connection relationship annotation can be added through node connections to represent the predecessor knowledge points.

[0109] Based on this, the previous knowledge points of the knowledge points to which the wrong questions belong can be quickly retrieved and determined through the knowledge graph. The knowledge points to which the wrong questions belong can be located in the knowledge graph, and then the previous knowledge points of the knowledge points to which the wrong questions belong can be retrieved from the pre-constructed knowledge graph; based on the user's answering of the knowledge points to which the wrong questions belong and the previous knowledge points, the knowledge points that the user is weak in mastering can be determined as the weak knowledge points.

[0110] That is, after determining the knowledge point to which the wrong question belongs and its preceding knowledge point, the user's knowledge mastery of the knowledge point to which the wrong question belongs and its preceding knowledge point can be quantitatively evaluated by the user's answers to the questions under the knowledge point to which the wrong question belongs and its preceding knowledge point. For example, the correct rate of questions under the knowledge point to which the user answered the wrong question and the correct rate of questions under the preceding knowledge point to which the user answered the wrong question can be obtained respectively, and the degree of weakness of the user's mastery of the corresponding knowledge point can be reflected by the level of the correct rate. For example, a correct rate threshold can be set. When the correct rate of questions under the knowledge point to which the user answered the wrong question / preceding knowledge point is lower than the correct rate threshold, it is considered that the user's mastery of the knowledge point to which the wrong question belongs / preceding knowledge point is weak, and the knowledge point belongs to a weak knowledge point.

[0111] In the disclosed embodiment, when determining the weak knowledge points related to the wrong questions, the knowledge points directly involved in the wrong questions are taken into consideration, and the predecessor knowledge points of the knowledge points are quickly determined through the constructed knowledge graph, and then by analyzing the user's answer performance on the knowledge points to which the wrong questions belong and their predecessor knowledge points, the learner's learning weaknesses are more accurately located, thereby providing a data basis for subsequent question recommendations, so that the final recommended practice questions can provide users with targeted training based on wrong questions, thereby optimizing learning efficiency.

[0112] Next, in order to enable those skilled in the art to better understand the question recommendation method provided by the present application, this embodiment takes a question recommendation system including a knowledge graph and an error cause data set as an example to illustrate the overall flow and specific implementation process of the method.

[0113] See also Figure 5A The overall flow diagram of the question recommendation method is shown as an example. When a user answers a question that is answered incorrectly, the error cause determination and weak knowledge point retrieval steps can be performed simultaneously. Among them, the error cause determination can be determined by matching the question identification ID with the user's answer and the error cause data set, so as to determine the reason why the wrong question was answered incorrectly by the user; the weak knowledge point retrieval includes determining the knowledge point to which the wrong question belongs, and retrieving the previous knowledge point of the knowledge point to which the wrong question belongs from the knowledge graph. The knowledge point to which each question belongs belongs to known information when the question is created, and then by querying the correct rate of the user's answer to the knowledge point to which the wrong question belongs and the previous knowledge point, the weak knowledge point is determined. When determining the weak knowledge point, the knowledge point to which the wrong question belongs can be directly determined as a weak knowledge point, and according to the answer situation of the question under the previous knowledge point of the knowledge point to which the wrong question belongs, it is determined whether the previous knowledge point belongs to the knowledge point that the user has a weak grasp of. If so, the weak knowledge point includes both the knowledge point to which the wrong question belongs and its previous knowledge point, otherwise the weak knowledge point includes the knowledge point to which the wrong question belongs.

[0114] After determining the cause of the error and the weak knowledge point, the cause of the error and the weak knowledge point are used as matching basis, and multiple questions matching at least one of the matching basis are screened out from the recommended question pool as matching questions, and then through algorithm evaluation of multiple scoring dimensions, multiple questions are determined for recommendation to the user for practice.

[0115] In order to more clearly describe the topic recommendation method, take different services implemented by different service modules as an example, see Figure 5B The schematic diagram of the interaction of multiple service modules involved in the question recommendation method is shown as an example, the front end 501 is used to display questions to users and receive answers from users, and when the user completes answering a question, the question information and answer information are reported to the learning gateway service 502;

[0116] In the case where the question is answered incorrectly, the learning gateway service 502 requests the content basic service 503 to query the knowledge point to which the wrong question belongs and its preceding knowledge points, and requests to query the reason why the wrong question is answered incorrectly by the user, so that the content basic service 503 can quickly retrieve and determine the knowledge point to which the wrong question belongs and its preceding knowledge points based on the question identification ID and the stored knowledge graph, and match the question identification ID and the user's answer with the error cause data set to determine the reason why the wrong question is answered incorrectly by the user, and return the error cause, as well as the knowledge point to which the wrong question belongs and its preceding knowledge points to the learning gateway service 502;

[0117] The learning gateway service 502 may continue to request the learning record service 504 to query the user's mastery of the knowledge point to which the wrong or correct question belongs and its preceding knowledge point, so that the learning record service 504 can calculate the mastery of the knowledge point based on the correct rate of the user's answers to the questions under the relevant knowledge point and return it;

[0118] After obtaining the reasons for the wrong questions and the weak knowledge points, the learning gateway service 502 can request the recommendation service 505 according to the reasons for the wrong questions and the weak knowledge points, so that the recommendation service 505 can select matching questions from the recommended question pool based on the reasons for the wrong questions and the weak knowledge points, and evaluate each matching question in multiple scoring dimensions to obtain the questions recommended for the user to practice, and return them to the learning gateway service 502;

[0119] The learning gateway service 502 displays the received recommended practice questions on the front-end page for the user to practice.

[0120] For any of the aforementioned embodiments, multiple questions matching at least one of the matching criteria are screened out from the recommended question pool as matching questions, and then each matching question is evaluated by an algorithm with multiple scoring dimensions to determine multiple questions for recommendation to the user for practice, this embodiment uses the scoring dimensions including knowledge point category, number of wrong answers by the user, type of wrong answer by the user, question difficulty level, and classification of reasons for wrong answers as an example to illustrate multi-dimensional scoring.

[0121] In the recommended topic pool including multiple topics for recommendation, each topic may include topic ID, knowledge point ID, knowledge point type, topic difficulty level, number of wrong answers by the user, type of wrong answers by the user, error cause ID and specific error cause category and other attribute information, corresponding to the set multiple scoring dimensions. A topic includes a knowledge point ID and may include multiple error cause IDs and their error cause categories at the same time. This attribute information is used for multi-dimensional scoring to determine the degree of adaptation between the matching topic and the user's current practice needs represented by the topic recommendation request.

[0122] For example, see Figure 5C An example of topic attribute information in the recommended topic pool is shown exemplarily. For each topic, attribute values ​​of different attributes are determined respectively. The attribute values ​​are used to evaluate the suitability of the topic in conjunction with a multi-dimensional scoring strategy.

[0123] Different scoring strategies can be set for different scoring dimensions, and dimension weights can be set for each scoring dimension. Figure 5D A schematic diagram of a multi-dimensional scoring strategy is shown as an example, including multiple dimensions such as knowledge point type, number of wrong answers by users, type of wrong answers by users, difficulty level of questions, and classification of wrong causes of questions. Different scoring strategies are set for each scoring dimension according to different dimension values. After screening out multiple matching questions, the score of each matching question on the scoring dimension is determined according to its value on each scoring dimension. Among them, when the knowledge point category of the scoring dimension belongs to a determined weak knowledge point, that is, the knowledge point ID of the knowledge point to which the matching question belongs belongs to one of the knowledge point ID sets of the weak knowledge point, the score of the matching question on the scoring dimension of the knowledge point category is determined based on the knowledge point type of the knowledge point to which the matching question belongs. The scoring dimension of the classification of wrong causes of questions can set a scoring strategy of multi-level wrong causes that are progressive layer by layer, such as the first-level wrong causes and the second-level wrong causes shown in the figure. Whether the wrong causes ID are the same can be used to determine whether the wrong causes included in the matching question are the same as the wrong causes of the wrong questions answered by the user, and the score on the scoring dimension of the classification of wrong causes of questions is determined based on the matching of the wrong causes ID.

[0124] by Figure 5DTaking the scoring strategy of the multi-dimensional scoring shown as an example, the weak knowledge points and error causes related to the wrong questions are matched in the recommended question pool. For the matched questions, if the knowledge points to which they belong are the test preparation knowledge points, 0.8*40% is the score of the matched question in the scoring dimension of knowledge point category; if the user has not answered the matched question, the score in the scoring dimension of the number of wrong answers by the user is 0.2*10%; if the user has completed the answer but made a mistake, the score in the scoring dimension of the type of wrong answer by the user is 0.6*10%; if the difficulty of the matched question is medium, the score in the scoring dimension of the difficulty level of the question is 0.6*10%; if the first-level error cause of the matched question has an intersection with the first-level error cause of the original wrong question, the score of the matched question in the first-level error cause is 0.4*15%, and the second-level error cause score is similar, and its score in the scoring dimension of the error cause classification of the question is obtained, and finally the total score of the matched question in all scoring dimensions is determined, and the recommended practice questions are determined based on the total scores of each matched question.

[0125] For example, there is a wrong question A, whose knowledge point category is exam preparation knowledge point, and the previous knowledge points are well mastered. In the recommended wrong question pool, question A and question B are matched. The knowledge points of question A are the same as those of question A, but the error reasons are not intersecting. The first-level and second-level error reasons of question B are the same as those of question A, but the knowledge points are different. A has been answered incorrectly 0 times before, and B has been answered incorrectly 2 times before and the answers are wrong. The difficulty of A and B is medium. Then, according to the formula for calculating the score, it can be known that the total score of A is 0.8*40%+0.2*10%+0*10%+0.1%15%+0.2*15%+0.6*10%=0.445, and the total score of B is 0*40%+0.4*10%+0.6*10%+0.4*15%+0.4*15%+0.6*10%=0.28. If only the most matching questions are recommended, question A will be pushed for users to practice.

[0126] In the embodiments of the present disclosure, by considering the causes of wrong questions and the knowledge points to which the wrong questions belong and their preceding knowledge points, the user's learning weaknesses can be identified and located more accurately, so that when recommending questions, the causes of wrong questions and the weak knowledge points related to the wrong questions are considered for question screening, and the degree of suitability of the screened questions to the user's current practice needs is evaluated from different dimensions, thereby providing the user with targeted question recommendations regarding the user's learning weaknesses, achieving the purpose of conducting targeted training on the user's learning weaknesses, and optimizing the user's learning effect and learning efficiency.

[0127] Corresponding to the embodiment of the above-mentioned topic recommendation method, see Figure 6 As shown, the present application also provides an embodiment of a topic recommendation device, the device comprising:

[0128] The question matching module 601 is used to respond to a question recommendation request initiated by a user based on a wrongly answered question, and obtain a plurality of matching questions matching at least one of a weak knowledge point and a wrong cause from a recommended question pool; wherein the weak knowledge point at least includes a knowledge point that the user has a weak grasp of, determined from the knowledge point to which the wrong question belongs and its preceding knowledge; the wrong cause indicates the reason why the wrong question was answered incorrectly by the user;

[0129] The matching question evaluation module 602 is used to calculate the score of each matching question in each scoring dimension to obtain the total score of the matching question in all scoring dimensions; wherein the scoring dimensions are obtained based on the attributes of the question and the answer of the user;

[0130] The topic recommendation module 603 is used to select N topics as recommended practice topics based on the total scores of multiple matching topics.

[0131] In some embodiments, the scoring dimension includes at least knowledge point category and question error reason classification; the matching question evaluation module is specifically used to:

[0132] When the knowledge point to which the matching question belongs belongs to the weak knowledge point, based on the importance of the knowledge point to which the matching question belongs, determining the score of the matching question in the scoring dimension of the knowledge point category;

[0133] The score of the matching question on the scoring dimension of question error cause classification is determined based on the error cause categories included in the matching question and the matching with the error causes of the wrong question answered incorrectly by the user.

[0134] In some embodiments, the scoring dimension also includes at least: at least one of the difficulty level of the question, the number of incorrect answers by the user, and the type of incorrect answer by the user; wherein the type of incorrect answer by the user includes at least the incorrect answer by the user and the user not knowing how to answer.

[0135] In some embodiments, the reasons why the user answered the wrong question incorrectly include multiple levels of reasons that are refined layer by layer; the matching question evaluation module is specifically used to:

[0136] According to the hierarchical structure of error causes, the error cause categories included in the matching questions and the error causes of the wrong questions answered by the user are matched level by level, and the level score of the level is determined based on whether the error causes of the two levels are the same;

[0137] According to the pre-set weights of each level, the level scores of all levels are weighted and calculated to obtain the scores of the matching questions on the scoring dimension of question error classification.

[0138] In some embodiments, the device further comprises a step of determining the cause of the error:

[0139] The wrong questions and the user's answers to the wrong questions are matched with the error cause data set to determine the error cause of the wrong answer to the wrong question; wherein the error cause data set includes a data set with each exercise question and the wrong answer to the exercise question as the division dimensions, and annotating the error cause of the user answering with the wrong answer for each wrong answer.

[0140] In some embodiments, the apparatus further comprises a step of determining weak knowledge points:

[0141] Retrieving the preceding knowledge points of the knowledge point to which the wrong question belongs from a pre-constructed knowledge graph; the knowledge graph refers to a network graph composed of knowledge points of various subjects and used to represent the dependency relationship between subject knowledge points;

[0142] Based on the knowledge point to which the wrong question belongs and the user's answer situation under the previous knowledge point, the knowledge point that the user has weak grasp of is determined as the weak knowledge point.

[0143] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0144] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application. Ordinary technicians in this field can understand and implement it without creative work.

[0145] The present application also provides an electronic device. The structural diagram of the electronic device is as follows: Figure 7 As shown, the electronic device 700 includes at least one processor 701, a memory 702 and a bus 703, and the at least one processor 701 is electrically connected to the memory 702; the memory 702 is configured to store at least one computer-executable instruction, and the processor 701 is configured to execute the at least one computer-executable instruction, thereby performing the steps of any one of the topic recommendation methods provided in any one of the embodiments or any one of the optional implementation modes in the present application.

[0146] Furthermore, the processor 701 may be a Field-Programmable Gate Array (FPGA) or other devices with logic processing capabilities, such as a Microcontroller Unit (MCU) or a Central Process Unit (CPU).

[0147] An embodiment of the present application also provides another readable storage medium storing a computer program, which is used to implement the steps of any topic recommendation method provided in any embodiment or any optional implementation manner of the present application when executed by a processor.

[0148] The readable storage medium provided in the embodiments of the present application includes, but is not limited to, any type of disk (including floppy disk, hard disk, optical disk, CD-ROM, and magneto-optical disk), ROM (Read-Only Memory), RAM (Random Access Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic card or optical card. That is, the readable storage medium includes any medium that can store or transmit information in a readable form by a device (e.g., a computer).

[0149] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some implementations, multitasking and parallel processing may be advantageous.

[0150] Although this specification includes many specific implementation details, these should not be interpreted as limiting the scope of any invention or the scope of protection claimed, but are mainly used to describe the features of the specific embodiments of specific inventions. Certain features described in multiple embodiments in this specification may also be implemented in combination in a single embodiment. On the other hand, the various features described in a single embodiment may also be implemented separately in multiple embodiments or in any suitable sub-combination. In addition, although features may work in certain combinations as described above and even initially claim protection, one or more features from the claimed combination may be removed from the combination in some cases, and the claimed combination may point to a sub-combination or a variation of a sub-combination.

[0151] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A topic recommendation method, characterized in that: The method comprises: In response to a question recommendation request initiated by a user based on a wrongly answered question, a plurality of matching questions matching at least one of a weak knowledge point and a wrong cause are obtained from a recommended question pool; wherein the weak knowledge point at least includes a knowledge point that the user has a weak grasp of, determined from the knowledge point to which the wrong question belongs and its preceding knowledge; the wrong cause indicates the reason why the wrong question was answered incorrectly by the user; For each matching question, calculate the score of the matching question in each scoring dimension to obtain the total score of the matching question in all scoring dimensions; wherein the scoring dimensions are obtained based on the attributes of the question and the user's answer situation; N questions are selected as recommended practice questions based on the total scores of multiple matching questions.

2. The method according to claim 1, characterized in that The scoring dimensions at least include knowledge point categories and question error cause classification; The step of calculating the score of the matching question in each scoring dimension includes: When the knowledge point to which the matching question belongs belongs to the weak knowledge point, based on the importance of the knowledge point to which the matching question belongs, determining the score of the matching question in the scoring dimension of the knowledge point category; The score of the matching question on the scoring dimension of question error cause classification is determined based on the error cause categories included in the matching question and the matching with the error causes of the wrong question answered incorrectly by the user.

3. The method according to claim 1 or 2, characterized in that: The scoring dimensions include at least: at least one of the difficulty level of the question, the number of incorrect answers by the user, and the type of incorrect answers by the user; wherein the type of incorrect answers by the user includes at least the incorrect answer by the user and the user not knowing how to answer.

4. The method according to claim 2, characterized in that: The reasons why the user answered the wrong question incorrectly include multiple levels of reasons that are refined layer by layer; Determining the score of the matching question on the scoring dimension of the classification of the error cause of the matching question based on the error cause category included in the matching question and the matching situation with the error cause of the wrong answer of the wrong question by the user includes: According to the hierarchical structure of error causes, the error cause categories included in the matching questions and the error causes of the wrong questions answered by the user are matched level by level, and the level score of the level is determined based on whether the error causes of the two levels are the same; According to the pre-set weights of each level, the level scores of all levels are weighted and calculated to obtain the scores of the matching questions on the scoring dimension of question error classification.

5. The method according to claim 1, characterized in that The method also includes the step of determining the cause of the error: The wrong questions and the user's answers to the wrong questions are matched with the error cause data set to determine the error cause of the wrong answer to the wrong question; wherein the error cause data set includes a data set with each exercise question and the wrong answer to the exercise question as the division dimensions, and annotating the error cause of the user answering with the wrong answer for each wrong answer.

6. The method according to claim 1, characterized in that The method also includes the step of determining weak knowledge points: Retrieving the preceding knowledge points of the knowledge point to which the wrong question belongs from a pre-constructed knowledge graph; the knowledge graph refers to a network graph composed of knowledge points of various subjects and used to represent the dependency relationship between subject knowledge points; Based on the knowledge point to which the wrong question belongs and the user's answer situation under the previous knowledge point, the knowledge point that the user has weak grasp of is determined as the weak knowledge point.

7. A topic recommendation device, characterized in that: The device comprises: The question matching module is used to respond to a question recommendation request initiated by a user based on a wrongly answered question, and obtain a plurality of matching questions matching at least one of a weak knowledge point and a wrong cause from a recommended question pool; wherein the weak knowledge point at least includes a knowledge point that the user has a weak grasp of, which is determined from the knowledge point to which the wrong question belongs and its preceding knowledge; the wrong cause indicates the reason why the wrong question is answered incorrectly by the user; A matching question evaluation module is used to calculate the score of each matching question in each scoring dimension, and obtain the total score of the matching question in all scoring dimensions; wherein the scoring dimensions are obtained based on the attributes of the question and the answer of the user; The topic recommendation module is used to select N topics as recommended practice topics based on the total scores of multiple matching topics.

8. An electronic device, characterized in that: include: Memory, processor; The memory is used to store computer programs; The processor is used to call the computer program to implement the method according to any one of claims 1-6.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images