A knowledge blind spot classification recommendation method based on personality knowledge graph

Abstract

The present invention relates to the technical field of knowledge blind spot classification recommendation, and discloses a knowledge blind spot classification recommendation method based on a personalized knowledge graph, including: obtaining historical question set data of a target user, analyzing and forming wrong question data, constructing a wrong question knowledge graph and a question bank knowledge graph, and pushing exercises to the target user for practice. The knowledge blind spot classification recommendation method based on the personalized knowledge graph timely obtains different question classifications and expands the question bank knowledge graph, so that the questions pushed to the target user in the process of answering questions are richer, so that the target user can learn from one example and apply it to other situations. The knowledge blind spots of the target user are obtained by establishing a wrong question knowledge graph, and the proportion of exercises of different test point types of the pushed exercises is calculated, so that the exercises are pushed to the target user in a targeted manner, and it is judged whether the target user has mastered the test point content, so that exercises are pushed to the target user multiple times, so as to achieve the effect of consolidating the test points and strengthening the exercises.

Description

Technical Field

[0001] The present invention relates to the technical field of knowledge blind spot classification recommendation, and specifically to a knowledge blind spot classification recommendation method based on a personality knowledge graph. Background Art

[0002] The knowledge blind spot classification recommendation method based on the personalized knowledge graph mainly builds a personalized knowledge graph based on the user's personal interests and knowledge background, and then recommends some knowledge fields that the user may be interested in but has never been exposed to before, namely the so-called knowledge blind spots, by analyzing the user's personalized knowledge graph. It includes: collecting user interests and knowledge background, building a personalized knowledge graph, recommending knowledge blind spots in new fields, and filtering and sorting knowledge blind spots. This method can more accurately recommend personalized knowledge for each user and improve the efficiency of users in learning and mastering new knowledge.

[0003] The existing knowledge blind spot classification recommendation method based on personalized knowledge graph cannot obtain more questions in time to expand the question bank, so that the questions pushed to the target user in the process of answering questions are relatively single. It is impossible to establish a wrong question knowledge graph based on the target user's historical question set data to obtain the target user's knowledge blind spots, and calculate the proportion of pushed practice exercises according to different test points and the number of wrong questions, so as to push practice exercises to the target user in a targeted manner. It is impossible to judge whether the target user has mastered the test point based on the target user's correct or incorrect answers to wrong questions and exercises of the same type as wrong questions, so as to push exercises to the target user again to achieve the purpose of consolidating test points and strengthening exercises. Its practicality also has certain limitations. Summary of the invention

[0004] The present invention provides a knowledge blind spot classification recommendation method based on a personality knowledge graph, which is used to promote the solution of the problems raised in the background technology.

[0005] The present invention provides the following technical solution: a knowledge blind spot classification recommendation method based on a personality knowledge graph, comprising:

[0006] Collection unit: obtains historical question set data of target users;

[0007] The historical question set data includes question type, question score, and question score value;

[0008] Analysis unit: Analyze and generate wrong question data based on the historical question set data of the target user;

[0009] Construction unit: Based on the wrong question data and historical question set data, construct the wrong question knowledge graph and question bank knowledge graph respectively;

[0010] Push unit: Push exercises to target users for practice based on the wrong question knowledge graph and question bank knowledge graph.

[0011] As an optional solution of the knowledge blind spot classification recommendation method based on the personalized knowledge graph of the present invention, wherein: the analysis and formation of wrong question data based on the historical question set data of the target user is specifically as follows:

[0012] Obtain historical question set data of target users;

[0013] Extract the exercise score of each question in the historical question set data, recorded as Z;

[0014] Extract the exercise score of each question in the historical question set data, record it as Y;

[0015] Set the judgment value;

[0016] If the exercise score Y-exercise score Z=judgment value, it is judged as the first correct data and the historical correct set is set;

[0017] All the questions in the historical question set data that are determined to be the first correct data are included in the historical correct set;

[0018] If the exercise score Y-exercise score Z>judgment value, it is judged as the first error data and the historical error set is set;

[0019] All questions in the historical question set data that are determined to be first error data are included in the historical error set;

[0020] Set the questions in the historical error collection as target wrong questions.

[0021] As an optional solution of the knowledge blind spot classification recommendation method based on the personalized knowledge graph of the present invention, wherein: the wrong question knowledge graph and the question bank knowledge graph are constructed respectively according to the wrong question data and the historical question set data, specifically:

[0022] Obtain the question types of all target wrong questions, and define the question types as wrong question types;

[0023] Get the number of all wrong question types;

[0024] According to the wrong question types, the target wrong questions are respectively integrated into a number of wrong question type triplets, where the number of wrong question type triplets is the number of wrong question types, that is, <wrong question, including, wrong question type>;

[0025] For each wrong question type triple, several wrong question triplets are formed, namely <wrong question type, including, target wrong question>;

[0026] Integrate the wrong question type triples and the wrong question triples to construct a wrong question knowledge graph.

[0027] As an optional solution of the knowledge blind spot classification recommendation method based on the personalized knowledge graph of the present invention, wherein: the wrong question knowledge graph and the question bank knowledge graph are constructed respectively according to the wrong question data and the historical question set data, and further includes:

[0028] Get all the questions and their exercise types in the historical question set data;

[0029] Get the number of all exercise types;

[0030] According to the exercise type, all questions are integrated into several exercise type triplets, where the number of exercise type triplets is the number of exercise types, i.e. <exercise, including, exercise type>;

[0031] For each exercise type triple, several exercise triples are formed, namely <exercise type, including, topic>;

[0032] Integrate the exercise type triples and exercise triples to construct the first question bank knowledge graph.

[0033] As an optional solution of the knowledge blind spot classification recommendation method based on the personalized knowledge graph of the present invention, wherein: the wrong question knowledge graph and the question bank knowledge graph are constructed respectively according to the wrong question data and the historical question set data, and further includes:

[0034] Get all the question types in the historical question set data and set them as the target question types;

[0035] Obtain all questions of the target search question type on the Internet, and define the question as the search question;

[0036] Get the number of all target search question types;

[0037] According to the target search problem type, all search problems are integrated into a number of search type triplets, where the number of search type triplets is the number of target search problem types, i.e. <problems, including, target search problem types>;

[0038] For each search type triple, several search question triples are formed, namely <target search question type, including, search question>;

[0039] Integrate the search type triples and the search question triples to construct a second question bank knowledge graph;

[0040] Merge the question type triples and the search type triples in the first question bank knowledge graph and the second question bank knowledge graph to form a target type triple;

[0041] Merge all the question triples and search question triples in the first question bank knowledge graph and the second question bank knowledge graph to form a target question triple;

[0042] Integrate the target type triples and target exercise triples to construct a question bank knowledge graph.

[0043] As an optional solution of the knowledge blind spot classification recommendation method based on the personalized knowledge graph of the present invention, wherein: the pushing of exercises to the target user for practice based on the wrong question knowledge graph and the question bank knowledge graph specifically includes:

[0044] Obtain the knowledge graph of wrong questions;

[0045] Get the question bank knowledge graph;

[0046] Extract the wrong question types and target wrong questions from the wrong question knowledge graph;

[0047] Extract the exercise type corresponding to the wrong question type in the question bank knowledge graph, and set the exercise type as the target practice exercise type;

[0048] Extract the target exercise triples corresponding to the target exercise types in the question bank knowledge graph;

[0049] Implement the strategy of pushing exercises on wrong questions.

[0050] The present invention has the following beneficial effects:

[0051] 1. The knowledge blind spot classification recommendation method based on the personalized knowledge graph obtains all test point types according to the historical question set data, searches the Internet, and establishes a question bank knowledge graph to obtain more questions in time to classify and expand the question bank knowledge graph, so that the questions pushed to the target users in the process of answering questions are richer, so that the target users can learn from one example and master the content of the test point.

[0052] 2. The knowledge blind spot classification recommendation method based on the personalized knowledge graph establishes a wrong question knowledge graph according to the historical question set data of the target user, thereby obtaining the target user's knowledge blind spots, and calculates the proportion of exercises of different test point types to be pushed according to the different test points and the number of wrong questions, and judges the target user's mastery of different test points, so as to push exercises to the target user in a targeted manner.

[0053] 3. The knowledge blind spot classification recommendation method based on the personalized knowledge graph determines whether the target user has mastered the test point according to the target user's correct or incorrect answers to the questions and exercises of the same type as the wrong questions, and pushes exercises to the target user multiple times at different times to achieve the purpose of consolidating the test points and strengthening the exercises. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a block diagram of the knowledge blind spot classification recommendation method based on the personality knowledge graph of the present invention;

[0055] Figure 2 A schematic diagram of a wrong question knowledge graph of the present invention;

[0056] Figure 3 This is a schematic diagram of the first question bank knowledge graph of the present invention;

[0057] Figure 4 This is a schematic diagram of the second question bank knowledge graph of the present invention;

[0058] Figure 5 It is a schematic diagram of the knowledge graph of the question bank of the present invention. DETAILED DESCRIPTION

[0059] Embodiment 1, a knowledge blind spot classification recommendation method based on personality knowledge graph, see Figure 1 ,include:

[0060] Collection unit: obtains historical question set data of target users;

[0061] The historical question set data includes question types, question scores, and question scores, wherein the question types are test point types. Taking mathematics as an example, the test point types include functions, number series, trigonometric functions, plane vectors, conic sections, solid geometry, derivatives, etc., that is, the question types are functions, number series, trigonometric functions, plane vectors, conic sections, solid geometry, derivatives, etc.;

[0062] Analysis unit: Analyze and generate wrong question data based on the historical question set data of the target user;

[0063] Construction unit: Based on the wrong question data and historical question set data, construct the wrong question knowledge graph and question bank knowledge graph respectively;

[0064] Push unit: Push exercises to target users based on the wrong question knowledge graph and question bank knowledge graph to consolidate knowledge.

[0065] Through the above-mentioned unit, the above-mentioned method can timely obtain different question classifications and expand the question bank knowledge graph, so that the questions pushed to the target user in the process of answering questions are richer, so that the target user can learn by analogy and thus master the test content. By establishing a knowledge graph of wrong questions, the target user's knowledge blind spots are obtained, and the proportion of exercises of different test point types in the pushed exercises is calculated, so as to push practice exercises to the target user in a targeted manner, judge whether the target user has mastered the test content, and push exercises to the target user multiple times at different times, so as to achieve the effect of consolidating test points and strengthening exercises.

[0066] Embodiment 2: This embodiment is an improvement made on the basis of embodiment 1. The knowledge blind spot classification recommendation method based on the personalized knowledge graph analyzes and forms wrong question data based on the historical question set data of the target user, specifically:

[0067] Obtain historical question set data of target users;

[0068] Extract the exercise score of each question in the historical question set data, recorded as Z;

[0069] Extract the exercise score of each question in the historical question set data, record it as Y;

[0070] A determination value is set, the determination value is 0, and is used to determine whether the exercise score Y is equal to the exercise score Z, so as to determine whether the target user's answer to the question is correct. If the exercise score Z = the exercise score Y, the answer to the question is correct; if the exercise score Z < the exercise score Y, the answer to the question is incorrect;

[0071] If the exercise score Y-exercise score Z=judgment value, it is judged as the first correct data and the historical correct set is set;

[0072] All the questions in the historical question set data that are determined to be the first correct data are included in the historical correct set;

[0073] If the exercise score Y-exercise score Z>judgment value, it is judged as the first error data and the historical error set is set;

[0074] All questions in the historical question set data that are determined to be first error data are included in the historical error set;

[0075] Set the questions in the historical error collection as target wrong questions.

[0076] Embodiment 3: This embodiment is an improvement made on the basis of embodiment 2. Figure 2-Figure 5 ,This scheme takes into consideration that it is necessary to construct a ,knowledge graph of the question bank and the knowledge graph of the wrong ,questions in order to discover as many knowledge blind spots of the target ,users as much as possible, so as to conduct exercises to achieve the purpose of consolidating the ,knowledge points.

[0077] In this embodiment, the wrong question knowledge graph and the question bank knowledge graph are constructed respectively according to the wrong question data and the historical question set data, specifically:

[0078] Obtain the exercise types of all target wrong questions, and define the exercise types as wrong question types. Taking mathematics as an example, the exercise types of the target wrong questions include functions, number series, trigonometric functions, conic sections, solid geometry, and derivatives, which are defined as wrong question type 1, wrong question type 2, wrong question type 3, wrong question type 4, wrong question type 5, and wrong question type 6 respectively;

[0079] Get the number of all wrong question types, that is, get the number of target wrong questions under wrong question type 1, wrong question type 2, wrong question type 3, wrong question type 4, wrong question type 5, and wrong question type 6 respectively. For example, if there are 10 target wrong questions sorted from 1 to 10, and there are test points of wrong question type 1 in target wrong question 1, target wrong question 3, target wrong question 4, and target wrong question 7, then the number of wrong question type 1 is 4;

[0080] According to the wrong question types, the target wrong questions are respectively integrated into several wrong question type triplets, where the number of wrong question type triplets is the number of wrong question types, that is, <wrong question, including, wrong question type>. For example, if there are 6 wrong question types in total, the number of wrong question type triplets is 6;

[0081] For each wrong question type triple, several wrong question triplets are formed, namely <wrong question type, including, target wrong question>;

[0082] Integrate the wrong question type triples and the wrong question triples to construct a wrong question knowledge graph, such as Figure 2 shown.

[0083] This embodiment also provides that the step of constructing a wrong question knowledge graph and a question bank knowledge graph respectively according to the wrong question data and the historical question set data further includes:

[0084] Obtain all questions and their exercise types in the historical question set data. Taking mathematics as an example, all exercise types in the historical question set data include functions, series, trigonometric functions, plane vectors, conic sections, solid geometry, and derivatives, and are respectively defined as exercise type 1, exercise type 2, exercise type 3, exercise type 4, exercise type 5, exercise type 6, and exercise type 7;

[0085] Get the number of all question types, that is, get the number of questions under question type 1, question type 2, question type 3, question type 4, question type 5, question type 6, and question type 7 respectively. For example, if there are 100 questions in the historical question set data, and 30 of them have test points of question type 1, then the number of question type 1 is 30;

[0086] According to the exercise type, all questions are integrated into several exercise type triplets, where the number of exercise type triplets is the number of exercise types, that is, <exercise, including, exercise type>. For example, if there are 7 exercise types in total, the number of exercise type triplets is 7;

[0087] For each exercise type triple, several exercise triples are formed, namely <exercise type, including, topic>;

[0088] Integrate the exercise type triples and exercise triples to construct the first question bank knowledge graph, such as Figure 3 shown.

[0089] This embodiment also provides that the step of constructing a wrong question knowledge graph and a question bank knowledge graph respectively according to the wrong question data and the historical question set data further includes:

[0090] Obtain all exercise types in the historical exercise data and define them as target search exercise types. Take mathematics as an example. For example, all exercise types in the historical exercise data include functions, series, trigonometric functions, plane vectors, conic sections, solid geometry, and derivatives, and are defined as exercise type 1, exercise type 2, exercise type 3, exercise type 4, exercise type 5, exercise type 6, and exercise type 7, respectively. Then, these 7 exercise types are used for search. Therefore, the target search exercise types are functions, series, trigonometric functions, plane vectors, conic sections, solid geometry, and derivatives, and are defined as target search exercise type 1, target search exercise type 2, target search exercise type 3, target search exercise type 4, target search exercise type 5, target search exercise type 6, and target search exercise type 7, respectively.

[0091] Obtain all questions of the target search question type on the Internet, and define the question as the search question;

[0092] Get the number of all target search question types, that is, get the number of search questions under target search question type 1, target search question type 2, target search question type 3, target search question type 4, target search question type 5, target search question type 6, and target search question type 7 respectively. For example, if there are 1000 search questions in total, and 500 of them have test points of target search question type 1, then the number of target search question type 1 is 500;

[0093] According to the target search problem type, all search problems are integrated into several search type triplets, where the number of search type triplets is the number of target search problem types, i.e. <problem, including, target search problem type>. For example, if there are 7 target search problem types in total, the number of search type triplets is 7.

[0094] For each search type triple, several search question triples are formed, namely <target search question type, including, search question>;

[0095] Integrate the search type triples and the search question triples to construct the second question bank knowledge graph, such as Figure 4 As shown;

[0096] The question type triples and search type triples in the first question bank knowledge graph and the second question bank knowledge graph are merged to form a target type triple. Since the question type in the question type triple is the same as the target search question type in the search type triple, the question type in the merged target type triple is still the question type in the question type triple.

[0097] Merge all the question triples and search question triples in the first question bank knowledge graph and the second question bank knowledge graph to form a target question triple;

[0098] Integrate the target type triples and target exercise triples to construct a question bank knowledge graph, such as Figure 5 shown.

[0099] Embodiment 4: This embodiment is an improvement made on the basis of Embodiment 3. In this embodiment, according to the wrong question knowledge graph and the question bank knowledge graph, exercises are pushed to the target user for practice, specifically including:

[0100] Get the wrong question knowledge graph, such as Figure 2 As shown;

[0101] Get the question bank knowledge graph, such as Figure 5 As shown;

[0102] Extract the wrong question type and target wrong question from the wrong question knowledge graph, such as wrong question type 1;

[0103] Extract the exercise type corresponding to the wrong question type in the question bank knowledge graph, and set the exercise type as the target exercise type. Taking mathematics as an example, if the wrong question type 1 is a function, extract the exercise type corresponding to the function in the question bank knowledge graph. If the exercise type corresponding to the function is exercise type 1, then the target exercise type is function;

[0104] Extract the target exercise triple corresponding to the target exercise type in the question bank knowledge graph, that is, extract the target exercise triple corresponding to exercise type 1 in the question bank knowledge graph, such as Figure 5 The corresponding exercise types are shown in Questions 11 to 1n and Search Exercises 11 to Search Exercises 1n;

[0105] Implement the strategy of pushing exercises on wrong questions.

[0106] The strategy for pushing wrong question exercises is as follows:

[0107] S1. Obtain the target user's historical intervals between questions, so as to determine the time for initially pushing the target wrong questions and the questions in the target exercise triple corresponding to the wrong question type to the target user. For example, if the target user does questions every 3 days during the learning process to achieve the purpose of practice, the historical intervals between questions are 3 days.

[0108] S2. Obtain the answer time of each target wrong question in the historical error set, and set the time as the push start time point, such as the push start time point is 20:00 on Monday;

[0109] S3, calculate the push termination time point, which is the push termination time point = push start time point + historical question interval time, and the push termination time point is Thursday 20:00, that is, at Thursday 20:00, the target wrong questions or target push exercises will be pushed to the target user again;

[0110] S4, setting an initial push time interval, wherein the initial push time interval starts at the push start time point and ends at the push end time point. At this time, the initial push time interval is (20:00 on Monday and 20:00 on Thursday), so as to push wrong questions and exercises of the same test point type as the wrong questions to the target user for practice;

[0111] S5. Execute the topic push strategy within the initial push time interval;

[0112] S6. Obtain the exercise score of the target pushed exercise, recorded as X;

[0113] S7, obtaining the exercise score of the target user after answering the target pushed exercise, recorded as W;

[0114] S8. If the exercise score X-exercise score W=the judgment value, then the first push strategy is executed, that is, the target user answers the target pushed exercise correctly;

[0115] S9. If the exercise score X-exercise score W>the judgment value, the second push strategy is executed, that is, the target user answers the target pushed exercise incorrectly.

[0116] The execution of the topic push strategy specifically includes:

[0117] Get the number of target wrong questions in the historical error set, recorded as B;

[0118] Set the number of exercises to be pushed, the number of exercises to be pushed = set multiple × number B, the set multiple is the multiple set by the target user according to his or her own preferences, for example, if the set multiple is 2, the number of target wrong questions B is 10, then the number of exercises to be pushed is 20, that is, a total of 20 wrong questions of the type of exercises are pushed to the target user;

[0119] Obtain the knowledge graph of wrong questions;

[0120] Extract the types of wrong questions. Taking mathematics as an example, the types of exercises of the target wrong questions include functions, number series, trigonometric functions, conic sections, solid geometry, and derivatives, which are respectively defined as wrong question type 1, wrong question type 2, wrong question type 3, wrong question type 4, wrong question type 5, and wrong question type 6;

[0121] Extract the target number of wrong questions under each wrong question type, recorded as C(C1, ..., Cn). For example, the target number of wrong questions under 6 wrong question types are 4, 4, 4, 4, 3, and 1 respectively;

[0122] Calculate the push ratio of exercises for each wrong question type respectively, exercise push ratio = quantity C ÷ quantity B, that is, the push ratios of exercises for the six wrong question types are 4 / 10, 4 / 10, 4 / 10, 4 / 10, 3 / 10, and 1 / 10 respectively, so as to push questions in the question bank knowledge graph that have the same test point type as each target wrong question for practice. According to the ratio of the number of target wrong questions under each wrong question type to the total number of target wrong questions, determine the target user's mastery of different test point types, so as to strengthen the practice of the test point types with high error rates and low mastery of the target users, so that the target users can deepen their impression of the test point type until they master the test point type;

[0123] Get the target exercise triples;

[0124] According to the wrong question type, the exercise push ratio, and the number of exercise pushes, the number of exercise pushes of each wrong question type in the target exercise triple is calculated respectively, and the push number is set as the target exercise push number. Target exercise push number = exercise push number × exercise push ratio. For example, if the number of exercise pushes is 20, and the exercise push ratios of the 6 wrong question types are 4 / 10, 4 / 10, 4 / 10, 4 / 10, 3 / 10, and 1 / 10 respectively, then the target exercise push numbers of the 6 wrong question types are 8, 8, 8, 8, 6, and 2 respectively, that is, the 20 exercises randomly pushed to the target user for practice are exercises of the 6 wrong question types, and the numbers of the 6 wrong question types pushed for practice are 8, 8, 8, 8, 6, and 2 respectively;

[0125] Randomly extract wrong question types with the target number of exercises to be pushed from the target exercise triplet and push them to the target user, and define the question as the target pushed exercise, that is, extract the above 6 wrong question types, a total of 20 questions, from the question bank knowledge graph, and the numbers of exercises of the 6 wrong question types to be pushed for practice are 8, 8, 8, 8, 6, and 2 respectively, and push exercises to the target user. The target pushed exercises are the 20 questions to be pushed for practice to the target user.

[0126] Among them, the execution of the first push strategy specifically includes the steps executed after the target user correctly answers the target pushed exercise. The target user correctly answers the target pushed exercise. At this time, the target pushed exercise is converted into data in the correct answer set in the historical question set data. At the same time, other target wrong questions under the exercise type of the target pushed exercise are pushed to the target user again at the end time point of the initial push time interval for practice, and according to the answer result of the target wrong question, the time of pushing the practice exercise next time is changed until the target user answers the target wrong question multiple times. Correct, the target user is deemed to have mastered the target wrong question, and the target wrong question is no longer pushed to the target user. As the number of questions the target user has done in the question bank knowledge graph increases, the historical question set data also changes accordingly, and the wrong question knowledge graph also changes accordingly. When the target user is deemed to have mastered the target wrong question, the target wrong question is moved to the wrong question knowledge graph. Taking mathematics as an example, if wrong question type 1 is a function, the target user answers the target pushed exercises of wrong question type 1 of the pushed exercises correctly, and the initial push time interval of the target user is (Monday 20:00, Thursday 20:00), then At 20:00 on Thursday, push the target wrong questions of wrong question type 1 to the target user for practice, and determine whether the target wrong questions are answered correctly. If the answer is correct, the next push of the target wrong questions will be 6 days later, that is, the extended push time interval is (Thursday 20:00, Wednesday 20:00). If the answer is wrong, the next push of the target wrong questions will be 1.5 days later, that is, the shortened push time interval is (Thursday 20:00, Saturday 8:00), and continue to push wrong question type 1 to the target user in the extended or shortened push time interval. Other questions in the unfinished set of the question bank for practice The target wrong question is learned, and the above steps are repeated for judgment. When the number of correct answers given by the target user to the target wrong question reaches the judgment threshold, such as 3 times, the target wrong question will no longer be pushed to the target user. At this time, the target wrong question in the wrong question knowledge graph is removed. For example, the target wrong question is the target wrong question 1 in the wrong question type 1. In the original wrong question knowledge graph, the wrong question type 1 includes the target wrong question 1, the target wrong question 3, the target wrong question 4, and the target wrong question 7. When the target user correctly answers the target wrong question 1 3 times, the wrong question type 1 in the wrong question knowledge graph includes the target wrong question 3, the target wrong question 4, and the target wrong question 7.

[0127] S20, obtaining the initial push time interval, such as (Monday 20:00, Thursday 20:00);

[0128] S21, extracting the exercise type corresponding to the target pushed exercise, and setting the exercise type as the target exercise type. Taking mathematics as an example, if the exercise type corresponding to the target pushed exercise is function, that is, exercise type 1, then the target exercise type is exercise type 1, that is, function;

[0129] S22, extracting the wrong question type corresponding to the target exercise type, and defining the wrong question type as the target wrong question type. Taking mathematics as an example, if the target exercise type is exercise type 1, i.e., function, and the wrong question type corresponding to function is wrong question type 1, then the target wrong question type is wrong question type 1, i.e., function;

[0130] S23, extracting all target wrong questions under the target wrong question type, and defining the target wrong question as the target pushed wrong question, that is, extracting all target wrong questions under wrong question type 1, which includes target wrong question 1, target wrong question 3, target wrong question 4, and target wrong question 7, and the target pushed wrong questions are target wrong question 1, target wrong question 3, target wrong question 4, and target wrong question 7;

[0131] S24, obtaining target push exercises and a second question bank knowledge graph;

[0132] S25, setting the question bank that has been collected and the question bank that has not been collected;

[0133] S26, all the target pushed exercises with exercise scores X-exercise score W=determination value are included in the completed set of the question bank, that is, the target pushed exercises answered correctly by the target user are included in the completed set of the question bank, and the exercises pushed again in the future are randomly extracted from the uncompleted set of the question bank;

[0134] S27, all exercises in the second question bank knowledge graph after removing all target pushed exercises and target pushed exercises with exercise score X-exercise score W> judgment value are included in the unfinished set of the question bank;

[0135] S28, when the push termination time point is reached, the target wrong questions are pushed to the target user;

[0136] S29, obtaining the exercise score and exercise grade of the target pushed wrong questions, which are recorded as V and U respectively;

[0137] S30, if the exercise score V-exercise score U=the judgment value, then the initial push time interval of the target push exercises and the target push wrong questions under the target exercise type is extended, and the extended initial push time interval is set as the extended push time interval, and the extended push time interval=2×initial push time interval, that is, if the target user answers the target push wrong questions correctly, if the initial push time interval is (Monday 20:00, Thursday 20:00), the interval is 3 days, then the next push of the target wrong questions will be 6 days later, that is, the extended push time interval is (Thursday 20:00, Wednesday 20:00);

[0138] S31, when the end time point of the extended push time interval is reached, the target pushed wrong questions and any one of the target wrong questions in the unfinished set of the question bank are pushed to the target user again, and steps S29 and S6-S9 are executed accordingly, and any one of the target wrong questions in the unfinished set of the question bank is defined as the unfinished target pushed exercise, wherein step S29 is executed to judge whether the target pushed wrong question is right or wrong, and steps S6-S9 are executed to judge whether any one of the target wrong questions in the unfinished set of the question bank is right or wrong, that is, to judge whether the unfinished target pushed exercise is right or wrong;

[0139] S32, if the exercise score V-exercise score U>determination value, then set a shortened push time interval, shortened push time interval = initial push time interval ÷ 2, that is, if the target user answers the wrong question, the target wrong question will be pushed. For example, if the initial push time interval is (Monday 20:00, Thursday 20:00), the interval is 3 days, then the next push of the target wrong question will be 1.5 days later, that is, the shortened push time interval is (Thursday 20:00, Saturday 8:00);

[0140] S33, when the end time point of the shortened push time interval is reached, the target pushed wrong questions and the unfinished target pushed exercises are pushed to the target user again, and steps S29 and S6-S9 are executed accordingly, wherein step S29 is executed to judge whether the target pushed wrong questions are right or wrong, and steps S6-S9 are executed to judge whether the unfinished target pushed exercises are right or wrong;

[0141] S34, setting a correct answer threshold, which is a threshold for the number of correct answers. If the correct answer threshold is reached, it is assumed that the target user has mastered the knowledge points of the target pushed wrong questions and target pushed exercises, and the target pushed wrong questions and target pushed exercises will no longer be pushed to the target user, for example, the correct answer threshold is 3 times;

[0142] S35, obtaining the number of correct answers given by the target user to the wrong questions pushed to the target and the questions not done to the target pushed, recorded as A;

[0143] S36, if the number of times A = the correct answer threshold, then the wrong questions and unanswered exercises will no longer be pushed;

[0144] S37. If the number of times A is less than the correct answer threshold, then execute steps S20-S36, that is, push the target wrong questions and the unanswered target push questions to the target user again.

[0145] Among them, the execution of the second push strategy specifically includes the steps executed after the target user gives an incorrect answer to the target pushed exercise. The target user gives an incorrect answer to the target pushed exercise. At this time, the target pushed exercise is converted into data in the set of incorrect answers in the historical question set data, that is, it is converted into a target wrong question. For example, if the target pushed exercise is a question of wrong question type 1, and there are 10 target wrong questions in the original wrong question knowledge graph, at this time, the target pushed exercise is converted into target wrong question 11. In the original wrong question knowledge graph, there are target wrong question 1, target wrong question 3, target wrong question 4, and target wrong question 7 under wrong question type 1. When the target user gives an incorrect answer to the target pushed exercise, If the target user answers the question incorrectly, the target wrong question 1, target wrong question 3, target wrong question 4, target wrong question 7, and target wrong question 11 will be listed under wrong question type 1 in the wrong question knowledge graph. At the same time, the target wrong question under the exercise type where the target pushed question is located will be pushed to the target user for practice. Based on the answer result of the target wrong question, the time for pushing the next practice question will be changed. Until the target user answers the target wrong question correctly many times, it is determined that the target user has mastered the target wrong question, and the target wrong question will no longer be pushed to the target user. As the number of questions the target user answers in the question bank knowledge graph increases, the data of its historical question set will also change accordingly. The wrong question knowledge graph also changes accordingly. When it is determined that the target user has mastered the target wrong question, the target wrong question is moved to the wrong question knowledge graph. Taking mathematics as an example, if the wrong question type 1 is a function, and the target user answers the wrong target pushed exercise of wrong question type 1 of the pushed exercise incorrectly, the target wrong question of wrong question type 1 will be pushed to the target user immediately, instead of pushing the target wrong question of wrong question type 1 to the target user at the end time of the initial push time interval (Monday 20:00, Thursday 20:00), and judging whether the target wrong question is answered correctly. If the answer is correct, the next push of the target wrong question will be 6 days later, that is, The extended push time interval is (20:00 on Thursday, 20:00 on Wednesday). If the answer is wrong, the next push of the target wrong question will be 1.5 days later, that is, the shortened push time interval is (20:00 on Thursday, 8:00 on Saturday), and the wrong question type 1 will continue to be pushed to the target user in the extended or shortened push time interval. Practice other questions in the unanswered set in the question bank, and repeat the above steps for judgment. When the number of correct answers to the target wrong question by the target user reaches the judgment threshold, such as 3 times, the target wrong question will no longer be pushed to the target user. At this time, the target wrong question in the wrong question knowledge graph is removed:

[0146] Extract the exercise type corresponding to the target pushed exercise, and define the exercise type as the target exercise type. Taking mathematics as an example, if the exercise type corresponding to the target pushed exercise is function, that is, exercise type 1, then the target exercise type is exercise type 1, that is, function;

[0147] Extract the wrong question type corresponding to the target exercise type, and define the wrong question type as the target wrong question type. Taking mathematics as an example, if the target exercise type is exercise type 1, i.e. function, and the wrong question type corresponding to function is wrong question type 1, then the target wrong question type is wrong question type 1, i.e. function;

[0148] Extract all target wrong questions under the target wrong question type, and arbitrarily extract one target wrong question as the target pushed wrong question, that is, extract all target wrong questions under wrong question type 1. The original wrong question type 1 includes target wrong question 1, target wrong question 3, target wrong question 4, and target wrong question 7. Since the target user gives an incorrect answer to the target pushed question, the target pushed question is transformed into target wrong question 11. Then the target pushed wrong questions are target wrong question 1, target wrong question 3, target wrong question 4, target wrong question 7, and target wrong question 11.

[0149] Push the target to push wrong questions immediately;

[0150] Execute step S29 to determine whether the target pushed wrong question is correct;

[0151] If the answer to the wrong question is correct, then the steps S30-S31 and S34-S37 are executed step by step, that is, the steps S30, S31, S34, S35, S36, S37 are executed in sequence;

[0152] If the answer to the target pushed wrong question is wrong, execute steps S32-S37.

[0153] In this embodiment, different question classifications are obtained in a timely manner and the question bank knowledge graph is expanded, so that the questions pushed to the target user in the process of answering questions are richer, so that the target user can learn from one example and thus master the test content. By establishing a wrong question knowledge graph, the target user's knowledge blind spots are obtained, and the proportion of exercises of different test types in the pushed exercises is calculated, so that the target user can be pushed exercises in a targeted manner, and it is judged whether the target user has mastered the test content, so that exercises are pushed to the target user multiple times at different times, so as to achieve the purpose of consolidating the test points and strengthening the exercises.

Claims

1. A knowledge blind spot classification recommendation method based on personality knowledge graph, characterized by: include: Collection unit: obtains historical question set data of target users; The historical question set data includes question type, question score, and question score value; Analysis unit: Analyze and generate wrong question data based on the historical question set data of the target user; Construction unit: Based on the wrong question data and historical question set data, construct the wrong question knowledge graph and question bank knowledge graph respectively; Push unit: Push exercises to target users for practice based on wrong question knowledge graph and question bank knowledge graph; The analysis and formation of wrong question data based on the historical question set data of the target user is specifically as follows: Obtain historical question set data of target users; Extract the exercise score of each question in the historical question set data, recorded as Z; Extract the exercise score of each question in the historical question set data, record it as Y; Set the judgment value; If the exercise score Y-exercise score Z=judgment value, it is judged as the first correct data and the historical correct set is set; All the questions in the historical question set data that are determined to be the first correct data are included in the historical correct set; If the exercise score Y-exercise score Z>judgment value, it is judged as the first error data and the historical error set is set; Add all questions in the historical question set data that are determined to be first error data into the historical error set; Set the questions in the historical error set as target wrong questions; According to the wrong question data and the historical question set data, the wrong question knowledge graph and the question bank knowledge graph are constructed respectively, specifically: Obtain the question types of all target wrong questions, and define the question types as wrong question types; Get the number of all wrong question types; According to the wrong question types, the target wrong questions are respectively integrated into a number of wrong question type triplets, where the number of wrong question type triplets is the number of wrong question types, that is, <wrong question, including, wrong question type>; For each wrong question type triple, several wrong question triplets are formed, namely <wrong question type, including, target wrong question>; Integrate the wrong question type triples and the wrong question triples to construct a wrong question knowledge graph; The method of constructing a wrong question knowledge graph and a question bank knowledge graph respectively according to the wrong question data and the historical question set data, further includes: Get all the questions and their exercise types in the historical question set data; Get the number of all exercise types; According to the exercise type, all questions are integrated into several exercise type triplets, where the number of exercise type triplets is the number of exercise types, i.e. <exercise, including, exercise type>; For each exercise type triple, several exercise triples are formed, namely <exercise type, including, topic>; Integrate the exercise type triples and the exercise triples to construct a first question bank knowledge graph; The method of constructing a wrong question knowledge graph and a question bank knowledge graph respectively according to the wrong question data and the historical question set data, further includes: Get all the question types in the historical question set data and set them as the target question types; Obtain all questions of the target search question type on the Internet, and define the question as the search question; Get the number of all target search question types; According to the target search problem type, all search problems are integrated into a number of search type triplets, where the number of search type triplets is the number of target search problem types, i.e. <problems, including, target search problem types>; For each search type triple, several search question triples are formed, namely <target search question type, including, search question>; Integrate the search type triples and the search question triples to construct a second question bank knowledge graph; Merge the question type triples and the search type triples in the first question bank knowledge graph and the second question bank knowledge graph to form a target type triple; Merge all the question triples and search question triples in the first question bank knowledge graph and the second question bank knowledge graph to form a target question triple; Integrate the target type triples and target exercise triples to construct a question bank knowledge graph; The pushing of exercises to target users for practice based on the wrong question knowledge graph and the question bank knowledge graph specifically includes: Obtain the knowledge graph of wrong questions; Get the question bank knowledge graph; Extract the wrong question types and target wrong questions from the wrong question knowledge graph; Extract the exercise type corresponding to the wrong question type in the question bank knowledge graph, and set the exercise type as the target practice exercise type; Extract the target exercise triples corresponding to the target exercise types in the question bank knowledge graph; Implement the strategy of pushing exercises for wrong questions; The strategy for pushing wrong question exercises is as follows: S1. Obtain the historical time interval between questions answered by the target user; S2. Obtain the answer time of each target wrong question in the historical error set, and set the time as the push start time point; S3, calculate the push end time point, push end time point = push start time point + historical question interval time; S4, setting an initial push time interval, wherein the initial push time interval takes the push start time point as the start time point and takes the push end time point as the end time point; S5. Execute the topic push strategy within the initial push time interval; S6. Obtain the exercise score of the target pushed exercise, recorded as X; S7, obtaining the exercise score of the target user after answering the target pushed exercise, recorded as W; S8. If the exercise score X-exercise score W=the judgment value, then execute the first push strategy; S9, if the exercise score X-exercise score W> the judgment value, execute the second push strategy; The execution of the topic push strategy specifically includes: Get the number of target wrong questions in the historical error set, recorded as B; Set the number of exercises to be pushed, the number of exercises to be pushed = set multiple × quantity B; Obtain the knowledge graph of wrong questions; Extract wrong question types; Extract the number of target wrong questions under each wrong question type, recorded as C (C1, ..., Cn); Calculate the push ratio of exercises for each wrong question type respectively, exercise push ratio = quantity C ÷ quantity B; Get the target exercise triples; According to the wrong question type, the exercise push ratio, and the exercise push quantity, the exercise push quantity of each wrong question type in the target exercise triple is calculated respectively, and the push quantity is set as the target exercise push quantity, and the target exercise push quantity = exercise push quantity × exercise push ratio; Randomly extract the wrong question type of the target exercise push number from the target exercise triple and push it to the target user, and set the question as the target push exercise; The executing the first push strategy specifically includes: S20, obtaining the initial push time interval; S21, extracting the exercise type corresponding to the target pushed exercise, and setting the exercise type as the target exercise type; S22, extracting the wrong question type corresponding to the target exercise type, and defining the wrong question type as the target wrong question type; S23, extracting all target wrong questions under the target wrong question type, and defining the target wrong questions as target pushed wrong questions; S24, obtaining target push exercises and a second question bank knowledge graph; S25, setting the question bank that has been collected and the question bank that has not been collected; S26, adding all the target pushed exercises with exercise score X-exercise score W=judgment value into the completed set of the question bank; S27, all exercises in the second question bank knowledge graph after removing all target pushed exercises and target pushed exercises with exercise score X-exercise score W> judgment value are included in the unfinished set of the question bank; S28, when the push termination time point is reached, the target wrong questions are pushed to the target user; S29, obtaining the exercise score and exercise grade of the target pushed wrong questions, which are recorded as V and U respectively; S30, if the exercise score V-exercise score U=the judgment value, then the initial push time interval of the push exercises of the target push exercises and the target push wrong exercises under the target exercise type is extended, and the extended initial push time interval is set as the extended push time interval, and the extended push time interval=2×initial push time interval; S31, when the end time point of the extended push time interval is reached, the target pushed wrong questions and any one of the target wrong questions of the type in the unfinished set of the question bank are pushed to the target user again, and steps S29 and S6-S9 are executed accordingly, and any one of the target wrong questions of the type in the unfinished set of the question bank is defined as the unfinished target pushed exercises; S32, if the exercise score V-exercise score U>determination value, then set to shorten the push time interval, shortened push time interval = initial push time interval ÷ 2; S33, when the end time point of the shortened push time interval is reached, the target pushed wrong questions and uncompleted target pushed exercises are pushed to the target user again, and steps S29 and S6-S9 are executed accordingly; S34, setting a correct answer threshold; S35, obtaining the number of correct answers given by the target user to the wrong questions pushed to the target and the questions not done to the target pushed, recorded as A; S36, if the number of times A = the correct answer threshold, then the wrong questions and unanswered exercises will no longer be pushed; S37, if the number of times A is less than the correct answer threshold, then execute steps S20-S36; The executing the second push strategy specifically includes: Extracting the exercise type corresponding to the target pushed exercise, and defining the exercise type as the target exercise type; Extract the wrong question type corresponding to the target question type, and define the wrong question type as the target wrong question type; Extract all target wrong questions under the target wrong question type, and randomly extract one target wrong question as the target pushed wrong question; Push the target to push wrong questions immediately; Execute step S29; If the answer to the wrong question is correct, then the steps S30-S31 and S34-S37 are executed step by step; If the answer to the target pushed wrong question is wrong, execute steps S32-S37.

Images