Learning path determination method, device, equipment and storage medium

Abstract

The embodiments of the present application provide a method, device, computing device and storage medium for determining a learning path. The method includes: inputting the acquired first learning situation data of the target user into a weak point diagnosis model to obtain the weak knowledge points of the target user; sorting the weak knowledge points of the target user to obtain the weak knowledge point sorting of the target user; and determining the target user's learning path for the weak knowledge points based on the weak knowledge point sorting. That is, in the present application, by processing the first learning situation data of the target user through a weak point diagnosis model, it is possible to achieve accurate diagnosis of the weak knowledge points of the target user, sort the diagnosed weak knowledge points, identify the weak knowledge points that need to be solved first and the weak knowledge points that need to be solved later, and then adapt an efficient learning path for the target user to improve the learning efficiency of the target user.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer technology, and in particular, to a method, apparatus, device and storage medium for determining a learning path. Background Art

[0002] Teaching students in accordance with their aptitude has always been a highly respected concept in the field of education. To achieve teaching students in accordance with their aptitude, it is necessary to find students' weak knowledge points and strengthen their learning of these weak knowledge points so as to improve students' learning outcomes without increasing their burden.

[0003] As online learning scenarios become more and more popular, various products that combine cognitive intelligence with traditional education have been put into practical use. For example, they can help students recommend homework, push similar questions to wrong questions in exams, or count the knowledge points of wrong questions to determine students' weaknesses, and then push practice questions based on them.

[0004] However, the current methods fail to accurately diagnose students' weak knowledge points, and are unable to further plan reasonable learning paths based on students' individual circumstances, which in turn leads to the problem of low learning efficiency. Summary of the invention

[0005] The embodiments of the present application provide a method, apparatus, device and storage medium for determining a learning path to improve students' learning efficiency.

[0006] In a first aspect, an embodiment of the present application provides a method for determining a learning path, including:

[0007] Inputting the acquired first learning situation data of the target user into a weak point diagnosis model to obtain the weak knowledge points of the target user;

[0008] Sorting the weak knowledge points of the target user to obtain the weak knowledge point ranking of the target user;

[0009] According to the ranking of the weak knowledge points, a learning path for the weak knowledge points of the target user is determined.

[0010] In a second aspect, an embodiment of the present application provides a learning path determination device, including:

[0011] A diagnosis unit, used for inputting the acquired first learning situation data of the target user into a weak point diagnosis model to obtain weak knowledge points of the target user;

[0012] A sorting unit, used to sort the weak knowledge points of the target user to obtain a sorting of the weak knowledge points of the target user;

[0013] A processing unit is used to determine the target user's learning path for the weak knowledge points according to the ranking of the weak knowledge points.

[0014] In a third aspect, a computing device is provided, comprising a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory to execute the method in the first aspect or its implementations.

[0015] In a fourth aspect, a chip is provided for implementing the method in any one of the first to second aspects or their respective implementations. Specifically, the chip includes: a processor for calling and running a computer program from a memory, so that a device equipped with the chip executes the method in any one of the first aspect or their respective implementations.

[0016] In a fifth aspect, a computer-readable storage medium is provided for storing a computer program, wherein the computer program enables a computer to execute the method in any aspect of the first aspect or its various implementations.

[0017] In a sixth aspect, a computer program product is provided, comprising computer program instructions, wherein the computer program instructions enable a computer to execute the method in any aspect of the first aspect or its various implementations.

[0018] In a seventh aspect, a computer program is provided, which, when executed on a computer, enables the computer to execute the method in any aspect of the first aspect or its implementation.

[0019] In summary, the present application obtains the weak knowledge points of the target user by inputting the acquired first learning situation data of the target user into the weak point diagnosis model; sorts the weak knowledge points of the target user to obtain the weak knowledge point sorting of the target user; and determines the learning path of the target user for the weak knowledge points according to the weak knowledge point sorting. That is, in the embodiment of the present application, the first learning situation data of the target user is processed by the weak point diagnosis model, so as to realize the accurate diagnosis of the weak knowledge points of the target user, sort the diagnosed weak knowledge points, identify the weak knowledge points to be solved first and the weak knowledge points to be solved later, and then adapt an efficient learning path for the target user to improve the learning efficiency of the target user. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] 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.

[0021] Figure 1 A schematic diagram of an application scenario of a method for determining a learning path applied to a terminal device provided in an embodiment of the present application;

[0022] Figure 2A schematic diagram of an application scenario of a learning path determination method applied to a server provided in an embodiment of the present application;

[0023] Figure 3 A flowchart of a method for determining a learning path provided in an embodiment of the present application;

[0024] Figure 4 A schematic diagram of a learning path determination process involved in an embodiment of the present application;

[0025] Figure 5 A schematic diagram of a learning path determination process involved in an embodiment of the present application;

[0026] Figure 6 A schematic diagram of a learning path determination process involved in an embodiment of the present application;

[0027] Figure 7 A flowchart of a method for determining a learning path provided in an embodiment of the present application;

[0028] Figure 8 A schematic diagram of the structure of a device for determining a learning path provided in an embodiment of the present application;

[0029] Fig. 9 It is a schematic block diagram of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0031] It should be understood that in the embodiment of the present invention, "B corresponding to A" means that B is associated with A. In one implementation, B can be determined based on A. However, it should also be understood that determining B based on A does not mean determining B based only on A, and B can also be determined based on A and / or other information.

[0032] In the description of the present application, unless otherwise specified, “plurality” means two or more than two.

[0033] In addition, in order to clearly describe the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish the same items or similar items with substantially the same functions and effects. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit the difference.

[0034] The technical solution proposed in this application is mainly used in the field of educational technology.

[0035] With the implementation of the national "double reduction" policy, local governments are required to control the total amount and duration of homework. Although homework is an important part of school education and teaching management and a necessary supplement to classroom teaching activities, some schools have too much homework, poor quality, and alienated functions, which not only fails to achieve the effect of reviewing old knowledge and learning new things, but also occupies students' normal exercise, rest, and entertainment time. Therefore, in recent years, the call for personalized homework has become louder and louder, that is, the release of homework needs to directly hit students' weak knowledge points, so as to achieve twice the result with half the effort.

[0036] Online learning scenarios are becoming more and more popular, with tens of millions of students taking classes, doing homework and taking exams online. Not only that, more and more schools have established academic databases for each student. With the accumulation of massive academic data, various products combining cognitive intelligence with traditional education have been put into practical use. Based on traditional teaching, these products provide more targeted supplements to improve students' academic performance. It can be said that they can reduce students' burden and pressure while improving their academic performance.

[0037] "Teaching students in accordance with their aptitude" is a profound educational thought that has been condensed in my country for more than two thousand years. Realizing teaching students in accordance with their aptitude is not only the educational dream of mankind, but also the unremitting pursuit of educators for thousands of years. This application combines important technical fields such as educational evaluation, natural language processing, machine learning and deep learning. On the basis of traditional education, it fully absorbs the teaching concept of "teaching students in accordance with their aptitude", and realizes the complete process from academic diagnosis to learning path planning, truly understanding the strengths and weaknesses of each student's academic situation, so that students can play to their strengths and avoid their weaknesses in the learning process and build interest in learning. The method of this application is applicable to online learning scenarios. In some embodiments, the student's learning path can be adjusted in real time in combination with the student's real-time answer results to help students quickly solve their weaknesses.

[0038] Through the research on existing education market applications and papers, it is found that the current solutions mainly include the following two types:

[0039] In some solutions, in the current education market, the main scenario is based on school exams, using each student's wrong test questions to search for similar questions for consolidation and re-practice. This solution has been widely used in schools, and these products are very popular because it is in line with the idea of ​​consolidating and re-practice wrong questions in pedagogy, and it allows teachers and students to give up time to find questions and quickly consolidate wrong questions. However, the main shortcomings of this type of method are also recognized by everyone. It cannot be recommended without an exam and it cannot find test questions that are not covered by the exam, especially for points with low assessments. Such a solution that relies heavily on exams will prevent students from overcoming weaknesses that are not covered by the exam. Even if the weaknesses solved this time are not covered in the next exam, the rollback consolidation effect cannot be achieved. In addition, this method of pushing questions by question cannot adapt to learning path planning, and cannot recommend test questions of difficulty suitable for students based on their learning situation. Therefore, the recommended questions may not play a role in solving weaknesses, but instead give students panic when answering.

[0040] In other solutions, some tutoring institutions will establish their own relevant knowledge point system, attribute relevant questions to relevant knowledge points according to the students’ practice in the tutoring institutions, and conduct rough probability statistics, which can also achieve the effect of diagnosing some weak points, thereby assisting the tutors in experience analysis, and the teachers further make relevant teaching plans based on the poorly mastered knowledge points. However, this type of solution is still in the auxiliary teaching link, and has not been derived to the student side to help students practice independently to overcome weaknesses. In simple terms, assuming that the accuracy of the statistical weaknesses is correct, but the teacher knows the weaknesses, the teacher needs to assign personalized homework to each student, correct each personalized homework in time, and adjust the exercises / give explanations according to the answers. Although this is in full compliance with the teaching principle of "teaching students in accordance with their aptitude", the actual operation is very burdensome for teachers, and it is impossible to automatically push learning paths to students or teachers.

[0041] As can be seen from the above, most of these products currently only help students recommend homework / test questions and push similar questions (such as "personalized learning manuals"), or judge students' weaknesses by counting wrong questions and knowledge points, and then push some exercises. In essence, none of them accurately diagnose students' weaknesses based on the complete learning situation and further plan a reasonable learning path based on the individual situation of students, so that each student can efficiently solve their weaknesses within a limited time.

[0042] In view of the limitations of existing solutions, this application provides a method for determining a learning path. Different from the traditional test question to knowledge point statistics solution, this application uses tens of millions of complete student learning data to train an accurate weak point diagnosis model, which can quickly construct a portrait of each student's learning situation and knowledge mastery (i.e., knowledge graph), which can help students accurately find weak links. Then, each student's weak points are individually sorted to identify priority and delayed solution points. Finally, the student's learning path is determined based on the sorted weak knowledge points, ensuring the rationality of the determined learning path to improve students' learning efficiency.

[0043] In order to facilitate the understanding of the technical solution provided by the embodiment of the present application, before introducing the detailed solution of the embodiment of the present application, the following Figure 1 and Figure 2 The application scenarios provided by the embodiments of the present application are exemplarily introduced.

[0044] Figure 1 A schematic diagram of an application scenario of a method for determining a learning path applied to a terminal device provided in an embodiment of the present application. Figure 1 In the application scenario shown, the terminal device 102 executes the learning path method provided in the embodiment of the present application. Figure 1 As shown, when user 101 triggers a learning path determination request on terminal device 102, terminal device 102 receives the learning path determination request, and obtains the target user's learning path for weak knowledge points by executing the method provided in the embodiment of the present application. Specifically, user 101 sends the target user's first learning situation data to terminal device 102, or user 101 sends a learning path request to terminal device 102, and the learning path request includes the target user's identification information. Terminal device 102 obtains the target user's first learning situation data from a server or other platform based on the target user's identification information. Next, terminal device 102 inputs the obtained first learning situation data of the target user into the weak point diagnosis model to obtain the target user's weak knowledge points, and determines the target user's learning path for weak knowledge points based on the weak knowledge points.

[0045] In addition, the training process of each model involved in the embodiment of the present application is also implemented by the terminal device 102.

[0046] Optionally, the terminal device 102 may save the trained weak point diagnosis model into its own storage space.

[0047] Optionally, the terminal device 102 may also save the trained weak point diagnosis model to a server corresponding to the terminal device 102, and read the weak point diagnosis model from the server when executing the method of the embodiment of the present application.

[0048] Figure 2A schematic diagram of an application scenario of a learning path determination method applied to a server provided in an embodiment of the present application. Figure 2 In the application scenario shown, when user 201 triggers a learning path determination request on terminal device 202, terminal device 202 receives the learning path determination request and forwards the learning path determination request to server 203, so that server 203 obtains the target user's learning path for weak knowledge points by executing the method provided in an embodiment of the present application.

[0049] In addition, the training process of each model involved in the embodiment of the present application is also implemented by the server 203.

[0050] Optionally, the server 203 may save the trained weak point diagnosis model to its own storage space.

[0051] Optionally, the server 203 may also save the trained weak point diagnosis model to other devices, such as a distributed storage system. When executing the method of the embodiment of the present application, the server 203 reads the weak point diagnosis model from the distributed storage system.

[0052] As mentioned above, the terminal device 102 and the terminal device 202 may include a mobile phone, a tablet computer, a laptop computer, a PDA, a mobile internet device (MID) or other terminal devices with data processing functions.

[0053] The server 203 may be a computing device such as a rack server, a blade server, a tower server or a cabinet server, etc. The server may be an independent server or a server cluster composed of multiple servers.

[0054] In this embodiment, the terminal device 202 is connected to the server 203 via a network. The network may be a wireless or wired network such as an intranet, the Internet, a global system of mobile communications (GSM), a wideband code division multiple access (WCDMA), a 4G network, a 5G network, Bluetooth, Wi-Fi, a call network, or the like.

[0055] It should be noted that the method for determining the learning path provided in the embodiment of the present application can be applied not only to Figure 1 or Figure 2 The application scenario shown can also be applied to other application scenarios where a learning path needs to be determined, and the embodiments of the present application do not specifically limit this.

[0056] The technical solutions of the embodiments of the present application are described in detail below through some embodiments. The following embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0057] Figure 3 A flow chart of a method for determining a learning path provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, including:

[0058] S301. Input the acquired first learning situation data of the target user into a weak point diagnosis model to obtain the weak knowledge points of the target user.

[0059] The execution subject of the embodiment of the present application can be Figure 1 The terminal device 102 shown may also be Figure 2 The server 203 is shown. Optionally, the execution subject of the embodiment of the present application may also be a computing device with other specific data processing functions.

[0060] For ease of description, the following embodiments are described by taking the execution subject as a computing device as an example.

[0061] For the convenience of description, in the embodiment of the present application, the learning status data of the target user is recorded as the first learning status data.

[0062] Exemplarily, the first learning data of the target user includes the answer record of at least one of the weekly test papers, monthly test papers, midterm test papers, final test papers, joint test papers, homework and classroom data in school within a preset time (for example, one semester), and the answer record includes test questions and the scores of the test questions.

[0063] It should be noted that the first learning situation data of the target user obtained in the embodiment of the present application can be understood as the full amount of learning situation data, that is, obtaining as much learning situation data of the target user as possible.

[0064] The weak point diagnosis model of the embodiment of the present application can be based on the first learning data of the target user to achieve the description of the learning profile of the target user and quickly locate the weak knowledge points of the target user. The biggest advantage of the scheme of determining weak knowledge points through the weak point diagnosis model in the embodiment of the present application compared with the traditional statistical method is that it can not only reflect the obvious weak knowledge points of individual students (i.e., the knowledge points involved in frequently wrong questions), but also build associations between knowledge points, conduct knowledge point association analysis, and even give early warnings for possible weaknesses in the future.

[0065] The embodiments of the present application do not limit the specific network structure of the weakness diagnosis model. For example, it can be a timing model such as RNN (Recurrent Neural Network), end-to-end Transformer (a machine translation model), etc.

[0066] The weak point diagnosis model of the embodiment of the present application is obtained through training with massive learning data. Some existing tutoring institutions / education companies use self-built learning knowledge systems and statistical methods to diagnose students' weaknesses and assist teachers in teaching. The biggest problem is that few institutions can obtain complete semester data for school students, especially important test data. Because the exam is in the most serious scenario, and the test questions have reasonable indicators such as high reliability and good discrimination, it is essential to portray the students' learning situation. The embodiment of the present application is based on online databases such as personalized online reading data and precise teaching test papers to obtain massive learning data of students, and the data quality and quantity are guaranteed. In this way, when using massive learning data to train the weak point diagnosis model, the weak point diagnosis model can be fully trained, ensuring the prediction accuracy of the weak point diagnosis model.

[0067] In the embodiment of the present application, the weak point diagnosis model is used to diagnose the weak knowledge points of the user, wherein the specific output content of the weak point diagnosis model can be set.

[0068] In case 1, the weak point diagnosis model of the embodiment of the present application can directly output the weak knowledge points of the target user, for example, directly output: knowledge point A, knowledge point B, knowledge point C, etc., so that the weak knowledge points of the target user can be determined as knowledge point A, knowledge point B, knowledge point C, etc. In other words, the embodiment of the present application can limit the output of the weak point diagnosis model, so that the weak point diagnosis model only outputs at least one knowledge point that the target user has a poor grasp of.

[0069] In this case 1, in addition to directly outputting the target user's weak knowledge points, the above-mentioned weak point diagnosis model can also output the mastery of each weak knowledge point, which can be represented by a scalar, and the size of the scalar can be used to represent the target user's mastery of the weak knowledge point, that is, the degree of weakness. In this way, when recommending subsequent learning materials, more learning materials can be recommended for knowledge points with greater weaknesses for intensive learning, and less learning materials can be recommended for knowledge points with less weaknesses for intensive learning.

[0070] In case 2, the weak point diagnosis model of the embodiment of the present application outputs the target user's mastery of multiple preset knowledge points. For example, the weak point diagnosis model outputs the target user's mastery of each of the preset 100 knowledge points. Exemplarily, the mastery of the knowledge point can be represented by a scalar. For example, the larger the scalar corresponding to the knowledge point, the better the mastery of the knowledge point, that is, it is not a weak knowledge point. The smaller the scalar corresponding to the knowledge point, the worse the mastery of the knowledge point, that is, it is a weak knowledge point. Optionally, the scalar can also be used to indicate the weakness of the knowledge point. For example, the larger the value of the scalar, the weaker the target user is with respect to the knowledge point. The smaller the value of the scalar, the better the target user's mastery of the knowledge point, and the smaller the value of the scalar, the better the target user's mastery of the knowledge point, and it is not a weak knowledge point.

[0071] In this situation 2, the target user's mastery of each knowledge point output by the weak point diagnosis model can be screened, and the knowledge points with a mastery level less than a preset value can be determined as the target user's weak knowledge points.

[0072] Based on the network structure of the weak point diagnosis model, the methods for obtaining the weak knowledge points of the target user through the weak point diagnosis model in S301 above include but are not limited to the following:

[0073] Method 1: directly input the first learning situation data of the target user into the weak point diagnosis model, and the weak point diagnosis model outputs the weak knowledge points of the target user based on the first learning situation data of the target user, or outputs the mastery of the target user on each preset knowledge point. That is to say, in method 1, the first learning situation data of the target user is not preprocessed before being input into the weak point diagnosis model, but is directly input into the weak point diagnosis model.

[0074] Method 2: In order to reduce the data processing amount of the weak point diagnosis model, in this method 2, before the first learning situation data of the target user is input into the weak point diagnosis model, the first learning situation data of the target user is first processed, for example, converted into a vector representation. Then, the processed data is input into the weak point diagnosis model, which reduces the data processing amount of the weak point diagnosis model, thereby improving the speed of determining the weak knowledge points of the target user.

[0075] In the second method, if Figure 4 As shown, the above S301 includes the following steps S301-A1 to S301-A3:

[0076] S301-A1. Determine the question vector representation and score vector representation of each question in the first learning situation data.

[0077] The methods for determining the question vector representation and score vector representation of each question in the first learning situation data in S301-A1 include but are not limited to the following examples:

[0078] Example 1, the first learning situation data of the target user includes multiple test questions, and the target user's score for each test question. Based on this, the first learning situation data of the target user can be input into a pre-trained neural network model A, and the neural network model A can simultaneously output the test question vector representation of the test question, and the user's score vector representation for the test question. In this way, the neural network model A can output the test question vector representation of each test question in the first learning situation data, and the target user's score vector representation for each test question. That is to say, in this Example 1, the test question vector representation and the score vector representation can be obtained simultaneously through a neural network model A.

[0079] Example 2: Input each test question in the first learning situation data into the test question representation model to obtain a test question vector representation of each test question; process the score of the target user on each test question in the first learning situation data to obtain a score vector representation of each test question.

[0080] The above-mentioned test question representation model can be a Text CNN (Text Convolutional Neural Network) type model, or it can be a currently popular large corpus pre-training model, such as BERT (Bidirectional Encoder Representation from Transformers) and ELMo (Embeddings from Language Models). The embodiment of the present application does not limit the specific type of the test question representation model.

[0081] That is to say, in Example 2, the process of determining the question vector representation and the process of determining the score vector representation are two processes, wherein the characteristic information of the question is extracted through a neural network model, such as a question representation model, to obtain the question vector representation. Since the score is a numerical scalar, the score vector representation can be obtained by performing data processing on the score.

[0082] For example, the scores are directly represented as score vectors.

[0083] For another example, the score of the target user on each test question in the first learning situation data is normalized to obtain a score vector representation of each test question.

[0084] The embodiment of the present application does not limit the manner in which the scores are normalized.

[0085] In one example, a zero-mean normalization method can be used to normalize the score of each test question in the first learning data of the target user to obtain a score vector representation of each test question. For example, the score vector representation of a test question is obtained by the following formula (1):

[0086]

[0087] Among them, x* is the target score vector representation for question i, xi is the target user's score for question i, and σ is the standard deviation of the target user's scores for all questions in the first learning situation data.

[0088] In another example, the minimum-maximum normalization method can be used to normalize the score of each test question in the first learning data of the target user to obtain the score vector representation of each test question. For example, the score vector representation of a test question is obtained by the following formula (2):

[0089] x*=xi-x min / x max -x min (2)

[0090] Among them, x* is the target score vector representation for question i, xi is the target user's score for question i, and x min and x max They are respectively the minimum score and the maximum score of the target user on all test questions in the first learning situation data.

[0091] Optionally, other normalization methods may be used to normalize the scores to obtain a score vector representation, where the score vector representation is a numerical value.

[0092] S301-A2. Integrate the question vector representations and score vector representations of all the questions in the first learning situation data to obtain an integrated vector representation.

[0093] According to the above steps, after obtaining the question vector representation and the score vector representation of each question, the question score vector representation and the score vector representation are integrated to obtain an integrated vector representation.

[0094] It should be noted that the integrated vector representation is a vector representation, that is, the question vector representation and the score vector representation of all the questions in the first learning situation data are integrated into a vector representation.

[0095] In some embodiments, since the learning process of knowledge points is sequential, for example, first learn knowledge point A, and then learn knowledge point B based on mastering knowledge point A. Therefore, in order to improve the diagnostic accuracy of the weak point diagnosis model for weak knowledge points, the embodiment of the present application integrates the test vector representation and the score vector representation of all test questions in the first learning data according to the learning order of the knowledge points corresponding to each test question in the first learning data to obtain the integrated vector representation. For example, if the learning order of the knowledge points corresponding to the test question is at the front, the test vector representation and the score vector representation of the test question are integrated first.

[0096] The embodiment of the present application does not limit the integration method of the test question vector representation and the score vector representation.

[0097] Possible implementation method 1, first concatenate the question vector representation and the score vector representation of each question, and then concatenate them to obtain the integrated vector representation. For example, the first learning situation data of the target user includes n questions, where the question vector representation of question i is question vector representation i, and the score vector representation of question i is score vector representation i. According to this method, the concatenation is performed to obtain the integrated vector representation: {question vector representation 1, score vector representation 1, question vector representation 2, score vector representation 2, ..., question vector representation i, score vector representation i, ..., question vector representation n, score vector representation n}.

[0098] Possible implementation method 2 is to perform dot product on the question vector representation and the score vector representation of each question to obtain the dot product result corresponding to each question, and then concatenate the dot product results corresponding to all questions in the first learning situation data to obtain an integrated vector representation.

[0099] For example, the first learning situation data of the target user includes n test questions, wherein the test question vector representation of test question i is test question vector representation i, and the score vector representation of test question i is score vector representation i. The test question vector representation i of test question i and the score vector representation i are dot-multiplied. Since the score vector representation i is a numerical value, the dot-multiplication result of the test question vector representation i and the score vector representation i is the multiplication of each dimension of the test question vector representation i by the score vector representation i. Assuming that the test question vector representation i is a 10-dimensional vector and the score vector representation i is 0.2, the dot-multiplication result of the test question vector representation i and the score vector representation i is the multiplication of each of the 10 dimensions of the test question vector representation i by 0.2, and a 10-dimensional vector is obtained. According to the above method, the above steps for test question i are performed on each of the n test questions to obtain the dot-multiplication result corresponding to each test question. Next, the dot-multiplication results corresponding to the n test questions are spliced ​​to obtain an integrated vector representation, for example, the dimension of the integrated vector representation is n*10.

[0100] Possible implementation method 3 is to perform dot product on the question vector representation and the score vector representation of each question to obtain the dot product result corresponding to each question, and take the average of the dot product results corresponding to all questions in the first learning situation data in each dimension to obtain the integrated vector representation.

[0101] In the embodiment of the present application, the dot product result is also a vector, and the number of elements included in the vector is called dimension. For example, if the vector includes 10 elements, then the dimension of the vector is determined to be 10.

[0102] Continuing with the example of the target user's first learning situation data including n test questions, the test question vector representation of test question i is test question vector representation i, and the score vector representation of test question i is score vector representation i. The test question vector representation i and the score vector representation i of test question i are dot-multiplied. Since the score vector representation i is a numerical value, the dot product result of the test question vector representation i and the score vector representation i is the multiplication of each dimension of the test question vector representation i by the score vector representation i. Assuming that the test question vector representation i is a 10-dimensional vector and the score vector representation i is 0.2, the dot product result of the test question vector representation i and the score vector representation i is the multiplication of each of the 10 dimensions of the test question vector representation i by 0.2, to obtain a 10-dimensional vector. According to the above method, perform the above steps for test question i for each of the n test questions to obtain the dot product result corresponding to each test question. Next, take the average of the dot product results corresponding to the n test questions in each dimension. For example, the dimension of the dot product result corresponding to each test question is 10. For each of the 10 dimensions, take the average of the dot product results corresponding to the n test questions in that dimension, and use the average value as the integrated value of that dimension, thereby obtaining an integrated 10-dimensional vector representation.

[0103] S301-A3, input the integrated vector representation into the weak point diagnosis model to obtain the weak knowledge points of the target user.

[0104] According to the above method, an integrated vector representation of the test question vector representation and the score vector representation is obtained, and then the integrated vector representation is input into the weak point diagnosis model to obtain the weak knowledge points of the target user.

[0105] In some embodiments, if it is necessary to obtain the target user's mastery of a specific knowledge point, the above S301 includes the following steps:

[0106] S301-B, inputting the first learning situation data of the target user and the preset knowledge point range into the weak point diagnosis model to obtain the weak knowledge points of the target user.

[0107] In the embodiment of the present application, based on the first learning situation data of the target user, an accurate diagnosis of the weak knowledge points of the target user is achieved through a weak point diagnosis model. Next, the following steps S302 and S303 are executed to determine the learning path of the target user.

[0108] S302: Sort the weak knowledge points of the target user to obtain the weak knowledge point ranking of the target user.

[0109] According to the above, the weak knowledge points of the target users are accurately found, but the sorting of the weak knowledge points themselves is still weak, that is, there is a certain learning order for weak knowledge points for different students, and this order theory directly affects the learning effect.

[0110] Based on this, after obtaining the weak knowledge points of the target user, the embodiment of the present application sorts the weak knowledge points of the target user to obtain the sorting of the weak knowledge points of the target user, and determines the learning path based on the sorted weak knowledge points of the target user.

[0111] The methods for sorting the weak knowledge points of the target user in S302 include but are not limited to the following:

[0112] Method 1: In the process of student learning, the criticality of all knowledge points is also different. Some knowledge points are key knowledge points, which every student must master, and they account for a large proportion in the exam. Some knowledge points are not key knowledge points, such as extended knowledge points, which do not need to be mastered by every student, and they account for a small proportion in the exam. Therefore, in method 1, the weak knowledge points of the target users can be sorted according to whether the weak knowledge points of the target users are key knowledge points, so that the target users can prioritize the key knowledge points and postpone the non-key knowledge points, thereby improving the learning efficiency of students.

[0113] Specifically, key knowledge points are obtained, wherein the key knowledge points can be obtained from the teaching syllabus, from the teacher's teaching materials, or by counting the key knowledge points from history test papers. For example, knowledge points that appear frequently in the test paper and account for a large proportion of the score are determined as key knowledge points.

[0114] Next, the weak knowledge points that are key knowledge points among the weak knowledge points of the target user are regarded as the weak knowledge points to be solved first and are arranged at the front. The weak knowledge points that are not key knowledge points among the weak knowledge points of the target user are regarded as the weak knowledge points to be solved later and are arranged after the weak knowledge points to be solved first, thereby obtaining the ranking of the weak knowledge points of the target user.

[0115] Method 2: By using the theory of the zone of proximal development, the weak knowledge points of the target user are ranked according to the mastery of the weak knowledge points by users with similar scores to the target user. Specifically, the above S302 includes the following steps S302-A1 to S302-A3:

[0116] S302-A1, obtaining the test ranking of the target user, and determining at least one user within a preset range above the test ranking of the target user as a role model user of the target user;

[0117] S302-A2, comparing the weak knowledge points of the target user with the weak knowledge points of the model user, determining the weak knowledge points that the model user has mastered among the weak knowledge points of the target user as the weak knowledge points that the target user should solve first, and determining the weak knowledge points that the model user has not mastered among the weak knowledge points of the target user as the weak knowledge points that the target user should solve later;

[0118] S302-A3. Obtain a ranking of the target user's weak knowledge points based on the target user's weak knowledge points that are prioritized and delayed.

[0119] In actual learning situations, students with poor academic performance and students with good academic performance have different degrees of acceptance of the same knowledge point. For example, the knowledge point "analytic geometry = comprehensive application of conic sections" is a weak knowledge point for both student A with poor academic performance and student B with good academic performance. However, when the practice materials for this knowledge point are recommended to student B, student B is easy to accept and learn. However, for student A, this weak knowledge point is a panic zone, and no matter how much time is spent, the expected learning effect cannot be achieved. This is because student A has not yet mastered the relevant basic knowledge. Directly learning more comprehensive weak knowledge points will have a great impact on student A's mind, and student A may be unwilling to learn.

[0120] In order to solve this technical problem, the embodiment of the present application adopts the theory of the zone of proximal development to sort the weak knowledge points of the target user. Specifically, the test ranking of the target user is obtained, and the test ranking can be the most recent test ranking. The ranking of the target user is floated up by a preset interval. For example, if the target user is ranked 30th, the preset interval of the float is 10%, and the interval of the model user obtained is the user ranked between 20th and 29th, and at least one user ranked between 20th and 29th is used as the model user of the target user. In other words, the embodiment of the present application uses the first few students close to the target user in the test ranking as the model users of the target user. The advantage of this method is that it always benchmarks users who are slightly stronger than themselves. The weak knowledge points of the target user are compared with the weak knowledge points of the model user, and the knowledge points that the model user has mastered in the weak knowledge points of the target user are determined as the weak knowledge points that the target user solves first, and the knowledge points that the model user has not mastered in the weak knowledge points of the target user are determined as the weak knowledge points that the target user solves later or part of them as extended knowledge points, so as to achieve a step-by-step learning effect, prevent the target user from falling directly into the learning panic zone, and thus improve the learning interest and learning efficiency of the target user.

[0121] In some embodiments, the weak knowledge points that the target user needs to solve first can be directly arranged before the weak knowledge points that need to be solved later, so as to obtain the ranking of the weak knowledge points of the target user.

[0122] In some embodiments, due to regional differences in education itself, different knowledge points are difficult to master in a certain region, but are easier to master in other regions due to differences in examination methods or education methods. Therefore, in order to further improve the accuracy of the ranking of the target user's weak knowledge points, it is necessary to adjust the target user's weak knowledge points determined by the above method in combination with the weak knowledge points of the target region where the target user is located. Specifically, Figure 5 As shown, the above S302-A3 includes the following steps S302-A31 to S302-A33:

[0123] S302-A31. Obtain an initial ranking of the target user's weak knowledge points based on the target user's weak knowledge points that are prioritized and delayed.

[0124] For example, the weak knowledge points that the target user has priority to solve are arranged before the weak knowledge points that are delayed to solve, so as to obtain the initial ranking of the weak knowledge points of the target user.

[0125] S302-A32. Obtain second learning situation data of the target area where the target user is located, and based on the second learning situation data, obtain weak knowledge points and weights of the weak knowledge points in the target area.

[0126] For the convenience of description, the learning situation data of the target area where the target user is located is recorded as the second learning situation data.

[0127] The embodiment of the present application does not limit the method of obtaining weak knowledge points of the target area based on the second learning situation data.

[0128] For example, through statistical methods, the weak knowledge points of students in the target area can be obtained.

[0129] The weight of the weak knowledge point may be a preset value set according to whether the weak knowledge point is a key knowledge point. For example, if weak knowledge point A is a key knowledge point, a larger weight is set, and if weak knowledge point B is a non-key knowledge point, a smaller weight may be set.

[0130] S302-A33. According to the weak knowledge points of the target area and the weights of the weak knowledge points, the initial ranking of the weak knowledge points of the target user is adjusted to obtain the ranking of the weak knowledge points of the target user.

[0131] Specifically, according to the weak knowledge points of the target area and the weights of the weak knowledge points, a weighted method is adopted to fine-tune the initial ranking of the weak knowledge points of the target user to obtain the ranking of the weak knowledge points of the target user.

[0132] For example, suppose the initial ranking of the target user's weak knowledge points is: knowledge point A, knowledge point B, knowledge point C, where the model user's mastery of knowledge point A is 100%, the model user's mastery of knowledge point B is 90%, and the model user's mastery of knowledge C is 50%. According to the weak knowledge points and the weights of the weak knowledge points in the target area, in the target area, the weight of knowledge point A is 1, the weight of knowledge point B is 4, and the weight of knowledge point C is 1.2. In this way, the mastery degree and weight of each knowledge point are multiplied, and the results are sorted from large to small to obtain the ranking of the target user's weak knowledge points, that is, the ranking of the target user's weak knowledge points is: knowledge point B, knowledge point A, knowledge point C.

[0133] The embodiment of the present application combines the second learning situation data of the target area where the target user is located to realize the ranking of the target user's weaknesses. When the weak knowledge points are studied according to the ranking of the weaknesses, the learning enthusiasm of the target user and the learning of key knowledge points can be improved.

[0134] S303: Determine the target user's learning path for the weak knowledge points according to the sorting of the weak knowledge points.

[0135] In the embodiment of the present application, the learning path of the target user for the weak knowledge points can be understood as recommending the learning materials corresponding to each weak knowledge point to the target user according to the order of the weak knowledge points. For example, the weak knowledge points of the target user are ranked as: knowledge point B, knowledge point A, knowledge point C, then the learning path is: first recommend the relevant learning materials of knowledge point B, after the relevant materials of knowledge point B are learned, then recommend the relevant learning materials of knowledge point A, after the relevant materials of knowledge point B are learned, then recommend the relevant learning materials of knowledge point C.

[0136] The implementation methods of S303 include but are not limited to the following:

[0137] Method 1: According to the ranking of weak knowledge points, the learning materials corresponding to each weak knowledge point are searched one by one in the existing learning resource library to obtain the target user's learning path for the weak knowledge points. Specifically, the learning resources corresponding to the first weak knowledge point in the weak knowledge point ranking are recommended to the target user first, and then the learning resources corresponding to the second weak knowledge point in the weak knowledge point ranking are recommended, and so on, so that the target user can learn according to the ranking of weak knowledge points, thereby achieving a step-by-step learning effect, and will not fall directly into the learning panic zone, thereby improving learning efficiency.

[0138] Method 2: automatically integrate learning resources through the model to generate a learning path for target users' weak knowledge points. That is, the above S303 includes the following steps S303-A1 to S303-A3:

[0139] S303-A1. Determine the feature information output by an intermediate layer of the weakness diagnosis model as the learning situation profile of the target user.

[0140] For example, the feature information output by any intermediate layer of the weakness diagnosis model is determined as the learning status profile of the target user.

[0141] For another example, the feature information output by the penultimate network layer of the weakness diagnosis model is determined as the learning profile of the target user.

[0142] In some embodiments, the learning situation portrait can be called a learning situation representation, and the learning situation portrait is a vector.

[0143] S303-A2: Input the target user's learning profile, weak knowledge point ranking, and learning resources in the target area where the target user is located into the path representation model to obtain the target user's learning path representation.

[0144] The above learning resources include exercises, videos, knowledge cards, etc.

[0145] Optionally, the learning materials also include labels, which may be difficulty, knowledge point labels, regional attribute labels, etc.

[0146] In some embodiments, the target user's learning profile, weak knowledge point ranking, and learning resources in the target area where the target user is located are directly input into the path representation model to obtain the target user's learning path representation.

[0147] In some embodiments, the target user's learning profile, weak knowledge point ranking, and learning resources in the target area where the target user is located are first converted into vector representations and then input into the path representation model to obtain the target user's learning path representation. Among them, the target user's learning profile is a vector in this application, so no conversion is required. Specifically, the weak knowledge point ranking and the learning resources in the target area where the target user is located are converted into vector representations.

[0148] For example, Figure 6 As shown in Figure 1, the weak knowledge point ranking and learning resources are mapped into weak knowledge point vectors and learning resource vectors respectively through one-hot form. Then, the learning situation portrait, weak knowledge point vectors and learning resource vectors are input into the path representation model to obtain the learning path representation of the target user.

[0149] The embodiment of the present application does not limit the specific network structure of the above-mentioned path characterization model. Optionally, the path characterization model is a shallow neural network model.

[0150] S303-A3. Obtain the target user's learning path for weak knowledge points based on the target user's learning path representation and the preset learning path representations of N users, where N is a positive integer.

[0151] In some embodiments, based on the learning path representation of the target user and the learning path representations of the preset N users, the similarity between the learning path representation of the target user and each of the learning path representations of the N users is calculated. The learning path corresponding to the learning path representation with the highest similarity is determined as the learning path of the target user for the weak knowledge point.

[0152] In some embodiments, Figure 6 As shown, the above S303-A3 includes the following steps S303-A31 to S303-A33:

[0153] S303-A31, determining the similarity between the learning path representation of the target user and the learning path representation of each of the N users;

[0154] S303-A32. Obtain M users whose similarity meets a preset condition from N users, where M is a positive integer less than or equal to N.

[0155] For example, M users whose similarity is greater than a preset value are obtained from N users.

[0156] For another example, the top M users with the greatest similarity are obtained from N users.

[0157] S303-A33: Determine the learning path of one of the M users as the learning path of the target user for the weak knowledge point.

[0158] In the embodiment of the present application, the similarity between the learning path representation of the target user and the learning path representation of each user in N users is determined, and a group of dynamic isomorphic users are locked from the massive N users, that is, M users who are very similar to the target user in learning ability structure. Then, based on the learning paths of these M users, the learning path of the target user for weak knowledge points is determined.

[0159] In some embodiments, the learning path of any one of the M users is determined as the learning path of the target user for the weak knowledge point.

[0160] In some embodiments, the learning path of a user with the largest similarity value among the M users is determined as the learning path of the target user for the weak knowledge point.

[0161] In some embodiments, Figure 6 As shown, S303-A33 includes the following steps S303-A331 and S303-A332:

[0162] S303-A331, determining a target learning path representation among the learning path representations of the M users, which has the highest similarity to the learning path representation of the target user and the lowest cost;

[0163] S303-A332: Determine the learning path for the target user's weak knowledge points according to the learning path corresponding to the target learning path representation.

[0164] The embodiment of the present application does not limit the method for determining the target learning path representation that has the highest similarity to the learning path representation of the target user and the lowest cost among the learning path representations of M users.

[0165] Exemplarily, a dynamic time warping (DTW) method is used to determine a target learning path representation among the learning path representations of M users, which has the highest similarity to the learning path representation of the target user and the lowest cost.

[0166] For example, the target learning path representation that is most similar to the target user's learning path representation and has the lowest cost among the M users is calculated by DTW=min{sqrt(Σwk) / k, where wk=(i, j), i is the i-th element in the target learning path representation, j is the j-th element in the learning path representation of one of the M users, and k is the number of elements. The learning path corresponding to the target learning path representation can be considered to be the most suitable learning path for the target user and the most effective learning path for overcoming the target user's weak knowledge points, rather than the shortest path alone.

[0167] In some embodiments, the target learning path is characterized by a corresponding learning path, and the learning path of the target user for the weak knowledge point is determined.

[0168] In other embodiments, when there is real-time student interaction in the learning path, that is, students are answering and recommending scenarios at the same time, the selection of learning path resources / exercises can be adjusted in real time to make the learning path more effective. Figure 6 As shown, the learning path corresponding to the target learning path representation is determined as the initial learning path for the target user's weak knowledge points; the real-time answer results of the target user are obtained, and the real-time answer results are used as incentives to update the initial learning path through the reinforcement learning model; the updated initial learning path is determined as the learning path for the target user's weak knowledge points. That is, in the embodiment of the present application, the learning path is adaptively adjusted according to the student's online answers, so that the target user has different personalized homework and reduces ineffective repetitive training.

[0169] Among them, the above reinforcement learning model is trained with the historical students' academic data as training data. In other words, the training source of the reinforcement learning model is the historical students' academic data in school. By modeling the learning path process of tens of millions of historical students, the reinforcement learning model can quickly find the appropriate learning path for the user who is most similar to the target user's learning path state to quickly overcome the weak knowledge point based on the target user's initial learning path and answer feedback.

[0170] In some embodiments, the real-time answer results of the target user can also be obtained, and the real-time answer results are mapped into answer result vectors through the language model, and then the learning situation portrait, weak knowledge point ranking, learning resources and answer result vectors are input into the path representation model to obtain the learning path representation of the target user. Specifically, the learning situation portrait, weak knowledge point vector, learning resource vector and answer result vector are input into the path representation model to obtain the learning path representation of the target user.

[0171] Optionally, the real-time answer result can be mapped into an answer result vector through the language model BERT.

[0172] The application scenarios of the embodiments of the present application include but are not limited to the following scenarios:

[0173] Scenario 1: In the scenario where teachers assign personalized paper homework, by inputting the target user's first learning data and current learning scope, the weak point diagnosis model can automatically locate the target user's weak knowledge points within the learning scope and sort the weak knowledge points. Then, based on the target user's learning profile, weak knowledge point sorting, and the learning resources in the target area where the target user is located, the target user's learning path representation is determined. Based on the target user's learning path representation, the learning path of the most similar student (including knowledge ability structure and learning path structure) when overcoming the target user's weak knowledge points is found. That is, for each weak knowledge point of the target user, the special training that needs to be practiced is recommended at one time. The training questions for each weak knowledge point are selected by automatically selecting training questions of different difficulty and different examination aspects from the question bank based on the target user's learning ability.

[0174] Scenario 2 is a scenario in which students perform online exercises based on recommended weak points. Based on Scenario 1, Scenario 2 uses the target user's real-time answer results as feedback and updates the reinforcement learning model in real time. The reinforcement learning model adaptively adjusts the difficulty of the current test question, changes the question point, or calls relevant video resources / learning card resources based on the target user's learning situation, learning level (excellent, average, poor, etc.), and answer situation (scoring rate and errors), so as to quickly solve students' weak knowledge points.

[0175] The method for determining the learning path provided in the embodiment of the present application is to obtain the weak knowledge points of the target user by inputting the acquired first learning situation data of the target user into the weak point diagnosis model; sorting the weak knowledge points of the target user to obtain the weak knowledge point sorting of the target user; and determining the learning path of the target user for the weak knowledge points according to the weak knowledge point sorting. That is, in the embodiment of the present application, the first learning situation data of the target user is processed by the weak point diagnosis model, so that the weak knowledge points of the target user can be accurately diagnosed, and the weak knowledge points after diagnosis can be sorted, and the weak knowledge points that are to be solved first and those that are to be solved later can be identified, thereby adapting an efficient learning path for the target user and improving the learning efficiency of the target user.

[0176] The above introduces the method for determining the learning path provided in the embodiment of the present application, and the following introduces the training process of the above-mentioned weakness diagnosis model.

[0177] Figure 7 A flow chart of a method for determining a learning path provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, including:

[0178] S401. Obtain historical academic data of K students, where K is a positive integer.

[0179] The historical academic data of the K students include at least one of the weekly test papers, monthly test papers, midterm test papers, final test papers, joint test papers, homework and classroom data of the K students in school within a preset historical time period.

[0180] For example, the embodiment of the present application screens out the answer records of student examination papers of K students' complete weekly / monthly / midterm / final / joint examinations in school in the past five years, processes the student examination scores and the scores of each question, and obtains the historical academic data of K students.

[0181] Optionally, the obtained homework and classroom data can be used as an additional supplement to the historical learning data.

[0182] S402. Input the historical learning data of K students into the weak point diagnosis model to obtain the predicted weak knowledge points of each of the K students.

[0183] The weakness diagnosis model of the embodiment of the present application is a time series model, which can successfully capture the relationship between similar questions, group together questions under the same concept such as functions and geometry, and combine the answering time of the test questions to depict the relationship between students' various knowledge points at different times.

[0184] The implementation process of S402 is substantially the same as that of S301 , and is described with reference to the description of S301 , which will not be repeated here.

[0185] S403: Determine the loss between the predicted weak knowledge point of each of the K students and the actual weak knowledge point of each of the K students.

[0186] In some embodiments, the actual weak knowledge point of each student among the K students can be whether the student answers the test questions in the future exam correctly, and the test questions are mapped to the corresponding knowledge points, and then whether the knowledge points are weak knowledge points. For example, if the test question is answered correctly, it is determined that the knowledge point corresponding to the test question is not a weak knowledge point. If the test question is answered incorrectly, it is determined that the knowledge point corresponding to the test question is a weak knowledge point.

[0187] In some embodiments, manual labeling methods are directly used to mark whether students are weak in various knowledge points.

[0188] Optionally, the loss function used in the embodiment of the present application may be a cross entropy loss function.

[0189] S404: Train the weak point diagnosis model according to the loss.

[0190] According to the above method, after the loss is determined, the loss is used to update the parameters in the weak point diagnosis model, and it is determined whether the updated weak point diagnosis model meets the preset training end condition. If the training end condition is met, the above updated weak point diagnosis model will be used as the trained weak point diagnosis model to execute the method of the above embodiment. If the training does not meet the training end condition, the weak point diagnosis model will continue to be updated with reference to the above method until the training end condition is met. Among them, the model training end condition can be that the model's prediction result reaches a preset value, or the model training times reach a preset number of times.

[0191] In the embodiment of the present application, a massive amount of learning data is used to train a weakness diagnosis model to achieve sufficient training of the model, so that the trained model can accurately diagnose the user's weak knowledge points.

[0192] The preferred embodiments of the present application are described in detail above in conjunction with the accompanying drawings. However, the present application is not limited to the specific details in the above embodiments. Within the technical concept of the present application, the technical solution of the present application can be subjected to a variety of simple modifications, and these simple modifications all belong to the protection scope of the present application. For example, the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present application will not further explain various possible combinations. For another example, the various different embodiments of the present application can also be arbitrarily combined, as long as they do not violate the ideas of the present application, they should also be regarded as the contents disclosed in the present application.

[0193] It should also be understood that in the various method embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0194] Combination of the above Figures 3 to 7 , describes in detail the method embodiment of the present application, and the following is combined with Figures 8 to 9 , describe in detail the device embodiments of the present application.

[0195] Figure 8 A schematic diagram of a learning path determination device provided in an embodiment of the present application. The learning path determination device may be a computing device or a component of a computing device (e.g., an integrated circuit, a chip, etc.). The computing device may be Figure 2 The server shown can also be Figure 1 The terminal equipment shown.

[0196] like Figure 8 As shown, the learning path determination device 10 includes:

[0197] A diagnosis unit 11 is used to input the acquired first learning situation data of the target user into a weak point diagnosis model to obtain the weak knowledge points of the target user;

[0198] A sorting unit 12, used to sort the weak knowledge points of the target user to obtain a sorting of the weak knowledge points of the target user;

[0199] The processing unit 13 is used to determine the target user's learning path for the weak knowledge points according to the ranking of the weak knowledge points.

[0200] In some embodiments, the diagnosis unit 11 is specifically used to determine the question vector representation and score vector representation of each question in the first learning situation data; integrate the question vector representation and score vector representation of all questions in the first learning situation data to obtain an integrated vector representation; input the integrated vector representation into the weakness diagnosis model to obtain the weak knowledge points of the target user.

[0201] In some embodiments, the diagnostic unit 11 is specifically used to input each test question in the first learning situation data into a test question representation model to obtain a test question vector representation of each test question; and process the score of the target user on each test question in the first learning situation data to obtain a score vector representation of each test question.

[0202] In some embodiments, the diagnosis unit 11 is specifically used to normalize the score of the target user on each test question in the first learning situation data to obtain a score vector representation of each test question.

[0203] In some embodiments, the diagnostic unit 11 is specifically used to integrate the question vector representations and score vector representations of all questions in the first learning situation data according to the learning order of the knowledge points corresponding to each question in the first learning situation data to obtain the integrated vector representation.

[0204] In some embodiments, the diagnostic unit 11 is specifically used to perform a dot product between the question vector representation and the score vector representation of each test question to obtain the dot product result corresponding to each test question, and concatenate the dot product results corresponding to all test questions in the first learning situation data to obtain the integrated vector representation; or, perform a dot product between the question vector representation and the score vector representation of each test question to obtain the dot product result corresponding to each test question, and take the average of the dot product results corresponding to all test questions in the first learning situation data in each dimension to obtain the integrated vector representation.

[0205] In some embodiments, the sorting unit 12 is specifically configured to

[0206] Acquire the test ranking of the target user, and determine at least one user within a preset interval above the test ranking of the target user as a role model user of the target user;

[0207] The weak knowledge points of the target user are compared with the weak knowledge points of the model user, and the knowledge points that the model user has mastered among the weak knowledge points of the target user are determined as the weak knowledge points that the target user should solve first, and the knowledge points that the model user has not mastered among the weak knowledge points of the target user are determined as the weak knowledge points that the target user should solve later; according to the weak knowledge points that the target user should solve first and the weak knowledge points that should be solved later, the weak knowledge points of the target user are ranked.

[0208] In some embodiments, the sorting unit 12 is specifically used to obtain an initial sorting of the target user's weak knowledge points based on the target user's priority and delay in solving weak knowledge points; obtain second learning situation data of the target area where the target user is located, and based on the second learning situation data, obtain the weak knowledge points of the target area and the weights of the weak knowledge points; adjust the initial sorting of the target user's weak knowledge points based on the weak knowledge points of the target area and the weights of the weak knowledge points to obtain the sorting of the target user's weak knowledge points.

[0209] In some embodiments, the processing unit 13 is specifically used to determine the feature information output by an intermediate layer of the weakness diagnosis model as the learning profile of the target user; input the learning profile of the target user, the ranking of the weak knowledge points, and the learning resources of the target area where the target user is located into the path representation model to obtain the learning path representation of the target user; based on the learning path representation of the target user and the learning path representations of N preset users, obtain the learning path of the target user for the weak knowledge points, where N is a positive integer.

[0210] In some embodiments, the processing unit 13 is specifically used to determine the similarity between the learning path representation of the target user and the learning path representation of each of the N users; obtain M users whose similarity meets a preset condition from the N users, where M is a positive integer less than or equal to N; and determine the learning path of one of the M users as the learning path of the target user for the weak knowledge point.

[0211] In some embodiments, the processing unit 13 is specifically used to determine a target learning path representation among the learning path representations of the M users, which has the highest similarity with the learning path representation of the target user and the lowest cost; and determine the learning path of the target user for the weak knowledge points based on the learning path corresponding to the target learning path representation.

[0212] In some embodiments, the processing unit 13 is specifically used to represent the target learning path and determine the corresponding learning path as the target user's initial learning path for weak knowledge points; obtain the real-time answer result of the target user, and use the real-time answer result as an incentive to update the initial learning path through a reinforcement learning model; and determine the updated initial learning path as the target user's learning path for weak knowledge points.

[0213] In some embodiments, the reinforcement learning model is trained using historical student academic data as training data.

[0214] In some embodiments, the learning profile of the target user is feature information output by the penultimate network layer of the weakness diagnosis model.

[0215] In some embodiments, the processing unit 13 is specifically used to sort the weak knowledge points and the learning resources in a one-hot form, mapping them into weak knowledge point vectors and learning resource vectors respectively; inputting the learning situation portrait, the weak knowledge point vector and the learning resource vector into the path representation model to obtain the learning path representation of the target user.

[0216] In some embodiments, the processing unit 13 is specifically used to obtain the real-time answer results of the target user, and map the real-time answer results into answer result vectors through a language model; input the learning situation portrait, the weak knowledge point vector, the learning resource vector and the answer result vector into the path representation model to obtain the learning path representation of the target user.

[0217] In some embodiments, the diagnosis unit 11 is specifically used to input the first learning situation data of the target user and a preset knowledge point range into a weak point diagnosis model to obtain the weak knowledge points of the target user.

[0218] In some embodiments, the processing unit 13 is also used to obtain historical learning data of K students, where K is a positive integer; input the historical learning data of the K students into the weakness diagnosis model to obtain predicted weak knowledge points for each of the K students; determine the loss between the predicted weak knowledge points of each of the K students and the actual weak knowledge points of each of the K students; and train the weakness diagnosis model according to the loss.

[0219] In some embodiments, the historical academic data of the K students include at least one of the weekly test papers, monthly test papers, midterm test papers, final test papers, joint test papers, homework and classroom data of the K students in school within a preset historical time period.

[0220] Optionally, the weak point diagnosis model is a time series model.

[0221] It should be understood that the device embodiment and the method embodiment may correspond to each other, and similar descriptions may refer to the method embodiment. To avoid repetition, they will not be described here. Specifically, Figure 7 The device shown can execute the embodiments of the above method, and the aforementioned and other operations and / or functions of each module in the device are respectively for implementing the method embodiments corresponding to the computing device, which will not be described in detail here for the sake of brevity.

[0222] The above describes the device of the embodiment of the present application from the perspective of the functional module in conjunction with the accompanying drawings. It should be understood that the functional module can be implemented in hardware form, can be implemented by instructions in software form, and can also be implemented by a combination of hardware and software modules. Specifically, the steps of the method embodiment in the embodiment of the present application can be completed by the hardware integrated logic circuit and / or software form instructions in the processor, and the steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or a combination of hardware and software modules in the decoding processor to perform. Optionally, the software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, etc. The storage medium is located in a memory, and the processor reads the information in the memory, and completes the steps in the above method embodiment in conjunction with its hardware.

[0223] Fig. 9 It is a schematic block diagram of a computing device provided in an embodiment of the present application, and the computing device is used to execute the above method embodiment.

[0224] like Fig. 9 As shown, the computing device 30 may include:

[0225] The memory 31 and the processor 32, the memory 31 is used to store the computer program 33 and transmit the program code 33 to the processor 32. In other words, the processor 32 can call and run the computer program 33 from the memory 31 to implement the method in the embodiment of the present application.

[0226] For example, the processor 32 may be configured to execute the above method steps according to instructions in the computer program 33 .

[0227] In some embodiments of the present application, the processor 32 may include but is not limited to:

[0228] General-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc.

[0229] In some embodiments of the present application, the memory 31 includes but is not limited to:

[0230] Volatile memory and / or non-volatile memory. Among them, the non-volatile memory can be read-only memory (ROM), programmable ROM (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM) or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus random access memory (DR RAM).

[0231] In some embodiments of the present application, the computer program 33 may be divided into one or more modules, which are stored in the memory 31 and executed by the processor 32 to complete the method for recording pages provided by the present application. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 33 in the computing device.

[0232] like Fig. 9 As shown, the computing device 30 may also include:

[0233] The transceiver 34 may be connected to the processor 32 or the memory 31 .

[0234] The processor 32 may control the transceiver 34 to communicate with other devices, specifically, to send information or data to other devices, or to receive information or data sent by other devices. The transceiver 34 may include a transmitter and a receiver. The transceiver 34 may further include an antenna, and the number of antennas may be one or more.

[0235] It should be understood that the various components in the computing device 30 are connected via a bus system, wherein the bus system includes not only a data bus but also a power bus, a control bus and a status signal bus.

[0236] According to one aspect of the present application, a computer storage medium is provided, on which a computer program is stored, and when the computer program is executed by a computer, the computer can perform the method of the above method embodiment. In other words, the present application embodiment also provides a computer program product containing instructions, and when the instructions are executed by a computer, the computer can perform the method of the above method embodiment.

[0237] According to another aspect of the present application, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computing device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computing device executes the method of the above method embodiment.

[0238] In other words, when implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a server, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server, or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a digital video disc (digital video disc, DVD)), or a semiconductor medium (e.g., a solid state drive (solid state disk, SSD)), etc.

[0239] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0240] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the module is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0241] The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, 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 scheme of this embodiment. For example, each functional module in each embodiment of the present application may be integrated into a processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0242] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for determining a learning path, characterized in that: include: Inputting the acquired first learning situation data of the target user into a weak point diagnosis model to obtain the weak knowledge points of the target user; Sorting the weak knowledge points of the target user to obtain the weak knowledge point ranking of the target user; Determine the feature information output by an intermediate layer of the weakness diagnosis model as the learning status profile of the target user, wherein the learning status profile is a vector; Inputting the learning profile of the target user, the ranking of the weak knowledge points, and the learning resources of the target area where the target user is located into a path representation model to obtain a learning path representation of the target user; Determine the similarity between the learning path representation of the target user and the learning path representations of N preset users, obtain M users whose similarities meet preset conditions from the N users, and determine the learning path of one user among the M users as the learning path of the target user for the weak knowledge point, where N is a positive integer and M is a positive integer less than or equal to N.

2. The method according to claim 1, characterized in that The step of inputting the acquired first learning situation data of the target user into the weak point diagnosis model to obtain the weak knowledge points of the target user includes: Determine a test question vector representation and a score vector representation for each test question in the first learning situation data; Integrate the question vector representations and score vector representations of all the questions in the first learning situation data to obtain an integrated vector representation; The integrated vector representation is input into the weak point diagnosis model to obtain the weak knowledge points of the target user.

3. The method according to claim 2, characterized in that The step of determining a question vector representation and a score vector representation of each question in the first learning situation data includes: Input each test question in the first learning situation data into a test question representation model to obtain a test question vector representation of each test question; The score of the target user on each test question in the first learning situation data is processed to obtain a score vector representation of each test question.

4. The method according to claim 2, characterized in that: The step of integrating the question vector representations and the score vector representations of all the questions in the first learning situation data to obtain an integrated vector representation includes: According to the learning order of the knowledge points corresponding to each test question in the first learning situation data, the test question vector representations and the score vector representations of all test questions in the first learning situation data are integrated to obtain the integrated vector representation.

5. The method according to any one of claims 2 to 4, characterized in that: The step of integrating the question vector representations and the score vector representations of all the questions in the first learning situation data to obtain an integrated vector representation includes: Performing a dot product of the question vector representation and the score vector representation of each question to obtain a dot product result corresponding to each question, and concatenating the dot product results corresponding to all questions in the first learning situation data to obtain the integrated vector representation; or Perform a dot product on the question vector representation and the score vector representation of each question to obtain the dot product result corresponding to each question, and take the average of the dot product results corresponding to all questions in the first learning situation data in each dimension to obtain the integrated vector representation.

6. The method according to any one of claims 1 to 4, characterized in that: The step of sorting the weak knowledge points of the target user to obtain the sorting of the weak knowledge points of the target user includes: Acquire the test ranking of the target user, and determine at least one user within a preset interval above the test ranking of the target user as a role model user of the target user; Compare the weak knowledge points of the target user with the weak knowledge points of the model user, determine the knowledge points that the model user has mastered among the weak knowledge points of the target user as the weak knowledge points that the target user should solve first, and determine the knowledge points that the model user has not mastered among the weak knowledge points of the target user as the weak knowledge points that the target user should solve later; According to the weak knowledge points that are preferentially solved and the weak knowledge points that are delayed in being solved by the target user, a ranking of the weak knowledge points of the target user is obtained.

7. The method according to claim 6, characterized in that The step of obtaining a ranking of the target user's weak knowledge points according to the target user's weak knowledge points that are preferentially solved and the weak knowledge points that are delayed in being solved comprises: According to the weak knowledge points that the target user prioritizes to solve and the weak knowledge points that are delayed to solve, an initial ranking of the weak knowledge points of the target user is obtained; Acquire second learning situation data of the target area where the target user is located, and obtain weak knowledge points of the target area and weights of the weak knowledge points based on the second learning situation data; According to the weak knowledge points of the target area and the weights of the weak knowledge points, the initial ranking of the weak knowledge points of the target user is adjusted to obtain the ranking of the weak knowledge points of the target user.

8. The method according to claim 1, characterized in that The step of determining the learning path of one of the M users as the learning path of the target user for the weak knowledge point includes: Determine a target learning path representation among the learning path representations of the M users, which has the highest similarity to the learning path representation of the target user and the lowest cost; According to the learning path corresponding to the target learning path representation, the learning path of the target user for the weak knowledge point is determined.

9. The method according to claim 8, characterized in that The determining the learning path of the target user for the weak knowledge point according to the learning path corresponding to the target learning path representation includes: Determine the learning path corresponding to the target learning path representation as the initial learning path for the target user's weak knowledge points; Acquire the real-time answer result of the target user, and use the real-time answer result as an incentive to update the initial learning path through a reinforcement learning model; The updated initial learning path is determined as the learning path for the weak knowledge points of the target user.

10. The method according to claim 1, characterized in that The step of inputting the learning profile of the target user, the ranking of the weak knowledge points, and the learning resources of the target area where the target user is located into a path representation model to obtain the learning path representation of the target user includes: Obtaining the real-time answer result of the target user, and mapping the real-time answer result into an answer result vector through a language model; The learning situation portrait, the weak knowledge point ranking, the learning resources and the answer result vector are input into the path representation model to obtain the learning path representation of the target user.

11. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: Obtain historical academic data of K students, where K is a positive integer; Inputting the historical learning data of the K students into the weak point diagnosis model to obtain the predicted weak knowledge points of each of the K students; Determine the loss between the predicted weak knowledge point of each of the K students and the actual weak knowledge point of each of the K students; The weak point diagnosis model is trained according to the loss.

12. The method according to any one of claims 1 to 4, characterized in that: The weak point diagnosis model is a time series model.

13. A learning path determination device, characterized in that: include: A diagnosis unit, used for inputting the acquired first learning situation data of the target user into a weak point diagnosis model to obtain weak knowledge points of the target user; A sorting unit, used to sort the weak knowledge points of the target user to obtain a sorting of the weak knowledge points of the target user; A processing unit is used to determine the feature information output by an intermediate layer of the weakness diagnosis model as the learning profile of the target user, where the learning profile is a vector; input the learning profile of the target user, the weak knowledge point ranking, and the learning resources of the target area where the target user is located into a path representation model to obtain the learning path representation of the target user; determine the similarity between the learning path representation of the target user and the learning path representations of N preset users, obtain M users whose similarities meet preset conditions from the N users, and determine the learning path of one of the M users as the learning path of the target user for the weak knowledge points, where N is a positive integer and M is a positive integer less than or equal to N.

14. A computing device, characterized in that include: Memory, processor; The memory is used to store computer programs; The processor is used to execute the computer program to implement the method according to any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 12 when executed by a processor.

Images